阅读说明：本技术 光学制品和与其进行交互的系统 (Optical goods and the system interacted ) 是由 迈克尔·A·麦科伊 安妮·C·戈尔德 西尔维娅·盖乔娃-博罗沃娃·古特曼 格伦·E·卡斯纳 于 2017-09-27 设计创作，主要内容包括：本发明提供光学制品,该光学制品包括具有空间上限定的布置的多个数据丰富的回射元件,其中当在紫外光谱、可见光谱、近红外光谱、或它们的组合内观察时,多个回射元件包括相对于背景基材具有至少两种不同的回射特性和至少两种不同的光学对比度的回射元件。(The present invention provides optical goods, the optical goods include the retro-reflection element of multiple data riches with the arrangement spatially limited, wherein when observing in ultraviolet spectra, visible spectrum, near infrared spectrum or their combination, multiple retro-reflection elements include the retro-reflection element for having at least two different retroreflection characteristics and at least two different optical contrasts relative to background substrate.)

1. a kind of optical goods, the optical goods include the retroeflection member of multiple data riches with the arrangement spatially limited Part, wherein when being observed in ultraviolet spectra, visible spectrum, near infrared spectrum or their combination, the multiple retro-reflection element Including having the retroeflection of at least two different retroreflection characteristics and at least two different optical contrasts relative to background substrate Element.

2. product according to claim 1, wherein the retro-reflection element in the near infrared region in relative to the background base Material has at least two different optical contrasts.

3. the product according to any one of claim 2, wherein the difference of optical contrast with 0.05 bright coverage Coverage is become clear with the weighted average less than 50% to describe.

4. the product according to any one of claim 2, wherein the difference of optical contrast with 0.05 it is bright Coverage and the bright coverage of weighted average less than 25%.

5. the product according to any one of claim 2, wherein the difference of optical contrast with 0.05 it is bright Coverage and the bright coverage of weighted average less than 15%.

6. the product according to any one of claim 2, wherein the difference of optical contrast with 0.10 it is bright Coverage and the bright coverage of weighted average less than 25%.

7. the product according to any one of claim 2, wherein the difference of optical contrast with 0.10 it is bright Coverage and the bright coverage of weighted average less than 20%.

8. the product according to any one of claim 2, wherein the difference of optical contrast with 0.10 it is bright Coverage and the bright coverage of weighted average less than 15%.

9. the product according to any one of claim 2, wherein the difference of optical contrast with 0.15 it is bright Coverage and the bright coverage of weighted average less than 15%.

10. the product described in any one of according to claim 1, wherein the retroeflection under 0.2 °/5 ° of viewing angle/incidence angle The retroeflection coefficient of element is greater than 50cd/lux-m2, is greater than 100cd/lux-m2 or is greater than 200cd/lux-m2.

11. the product described in any one of according to claim 1, wherein the optical goods include the safe dress ornament of high-visibility.

12. the product described in any one of according to claim 1, wherein the retro-reflection element of the multiple data rich is with duplicate The arrangement spatially limited configures, so that can explain information when a part of the retro-reflection element is blocked.

13. the product described in any one of according to claim 1, wherein the optical goods are deformable optical goods.

14. the product described in any one of according to claim 1, wherein described be deformed at least one of contraction or expansion.

15. product according to claim 14, wherein the deformation leads at least two retroeflection in the retro-reflection element Spacing variation between element.

16. product according to claim 14, wherein it is described be deformed into it is reversible.

17. the product described in any one of according to claim 1, wherein at least two different retroreflection characteristics are not at least two not Same backreflected intensity value.

18. the product described in any one of according to claim 1, wherein at least two different retroreflection characteristics are in different perspectives Under at least two different backreflected intensity values when checking.

19. the product described in any one of according to claim 1, wherein the retroreflection characteristics change in response to condition and changed.

20. product according to claim 19, wherein condition variation is heat, moisture, mechanically deform or the change of radiation At least one of change.

21. the product described in any one of according to claim 1, wherein the retro-reflection element is individually sized and that This is separated, and each individually retro-reflection element is distinguished in the desired distance punishment away from the optical goods.

22. the product described in any one of according to claim 1, wherein the arrangement spatially limited includes geometrical arrangements side Formula, wherein the retro-reflection element, which is oriented the adjacent retro-reflection element with they, has a distance, and the wherein retroeflection member There is the periodicity from an element to another element in the arrangement that part limits on the space.

23. product according to claim 22, wherein the periodicity is regular periodicity.

24. product according to claim 22, wherein the periodicity is property irregular cycle.

25. the product described in any one of according to claim 1, wherein it is described spatially limit be arranged as rotating it is insensitive 's.

26. product according to claim 25, wherein geometrical arrangements mode needed for the arrangement each spatially limited Quantity depends on required fit quality.

27. product according to claim 25, wherein the retro-reflection element is oriented nearest adjacent time with them Element is penetrated at a distance of a characteristic distance.

28. product according to claim 27, wherein the retro-reflection element have size with to adjacent retro-reflection element away from From characteristic rate, the characteristic rate is not with visual angle change.

29. the product described in any one of according to claim 1, the product further includes printing layer, and the printing layer setting exists On at least part of outer surface of the retro-reflection element.

30. the product described in any one of according to claim 1, wherein the retroreflection characteristics are can be in the infrared spectroscopy It detects.

31. wherein at least two retro-reflection element is discontinuous according to claim 1 to product described in any one of 30.

32. the product described in any one of according to claim 1, wherein the multiple retro-reflection element is discontinuous.

33. the product described in any one of according to claim 1, wherein the optical goods are arranged on substrate, the substrate Selected from least one of infrastructure, wearable object and vehicle.

34. a kind of fabric, the fabric includes according to claim 1 to product described in any one of 31.

35. a kind of system, the system comprises:

A) according to claim 1 to product described in any one of 33；

B) optical system；With

C) inference engine, the inference engine is used to explain and the multiple retro-reflection element of classifying, wherein the optical system will Data are fed to the inference engine.

36. system according to claim 35, wherein the product is arranged in infrastructure, target, wearable object and vehicle At least one of on.

37. system according to claim 35, wherein a part of the optical system for vehicle, and other wherein institute State the input that vehicle uses the information as autonomous driving module.

38. the system according to claim 37, wherein the vehicle provides mankind's language to driver using the information Speech feedback.

39. the system according to claim 37, wherein the vehicle provides touching to the driver using the information Feel at least one of feedback, audio feedback or visual feedback.

40. system according to claim 35, wherein the retro-reflection element of the multiple data rich is in the infrared spectroscopy In be visible.

41. system according to claim 33, wherein the letter relevant to the retroreflective articles of the multiple data rich Breath includes expected road worker, expected pedestrian, expected construction worker, expected student, at least one in expected emergency answering worker Person.

42. system according to claim 35, wherein the inference engine is stored locally as the optical system Composition.

43. system according to claim 35, wherein the optical system is drawn using wireless communication protocol and the reasoning Hold up communication.

Technical field

This disclosure relates to optical goods.More specifically, this disclosure relates to being configured to be observed by optical system and the mankind The optical goods that both person and the system including such optical goods read and/or pay attention to.

Background technique

Optical system includes the method for acquiring, analyzing and understanding image.For example, the illustrative of these systems is answered With including robot, face recognition, picture search, machine vision, remote sensing, monitoring, autonomous vehicle and object detection.Object detection Some examples of application include vehicle vision system, autonomous vehicle and safety of workers.

In recent years, computer vision system has taken a variety of methods to detect interested object, such as pedestrian.Most of objects Body detecting method includes visible light or near-infrared video camera.Since optical goods and system can be located at (for example, daytime or night； It is sunny or cloudy；Urban district or city；Building etc.) environment complexity and diversity, the posture that they can take it is more The diversity of sample and their appearances based on size, clothing etc., and due to potential partial occlusion, object inspection The problem of survey is complicated.

Many methods in pedestrian detection method use the exhaustion scanning to whole image or the outline based on template, Body part matching.However, the diversity for the form that can be taken in the picture due to the mankind, these methods are extremely difficult, It is time-consuming and there is not satisfactory performance.

Equally, mankind observer detects and identifies that the difficult task of pedestrian results in drawing for high-visibility clothes at night Enter and adjusts.High-visibility clothes (that is, clothes with retroreflection material) are designed to by returning to more incident lights Light source and keep wearer more visible or aobvious so that easily the pattern in the form of other mankind can be identified as by human viewer Eye.Current optical system is based on collecting a large amount of training datas, is annotated by human expert to it, and then training pattern To detect interested certain objects.This collection of data and annotation are time-consuming and with high costs.

Even if considering the prior art relevant to optical goods, still have an opportunity to improve optical goods and including such optics The substrate of product, such as infrastructure, wearable object, vehicle and other products.

Summary of the invention

The disclosure provides many advantages compared with existing optical goods and the system being used therewith.The disclosure provides It easily can observe and/or communicate information in any condition the optics system of machine observer, human viewer, or both Product, to allow to identify and track such optical goods on user or object.

The disclosure provides a kind of optical goods, which includes multiple numbers with the arrangement spatially limited to configure According to retro-reflection element abundant, plurality of retro-reflection element includes different at least two different retroreflection characteristics and at least two Optical contrast retro-reflection element, wherein data rich refers to the information that can be easy to machine interpretation.In some cases, multiple The retro-reflection element of data rich is configured with the duplicate arrangement spatially limited so that even if retro-reflection element a part of quilt Also information can be explained when blocking.

In some cases, at least two different retroreflection characteristics are at least two different backreflected intensity values.Some In the case of, at least two different retroreflection characteristics are at least two different wavelength.In some cases, at least two is different Retroreflection characteristics have at least two different polarization states.In some cases, at least two different retroreflection characteristics have at least Two different phase delays.

In some cases, the arrangement spatially limited includes geometrical arrangements mode, wherein retro-reflection element be oriented with Their adjacent retro-reflection element has a distance, and has in the arrangement that wherein retro-reflection element spatially limits from a member Periodicity of the part to another element.It in some cases, is periodically regular periodicity.In some cases, periodically it is Property irregular cycle.In some cases, spatially limit be arranged as rotating it is insensitive.

In some cases, the quantity of geometrical arrangements mode needed for the arrangement each spatially limited depends on required Fit quality.In some cases, retro-reflection element be oriented the nearest adjacent retro-reflection element with them at a distance of a feature away from From.In some cases, retro-reflection element has size with to the characteristic rate at a distance from adjacent retro-reflection element, and this feature ratio is not With visual angle change.

The disclosure provides the fabric including aforementioned articles.

The disclosure further includes a kind of system, which includes any product, optical system and reasoning in aforementioned articles Engine, which is used to explain and multiple retro-reflection elements of classifying, and wherein optical system feeds data to inference engine.One In a little situations, product is arranged at least one of infrastructure, target, wearable object and vehicle.

In some cases, optical system is a part of vehicle, and in addition wherein vehicle use information is used as independently Drive the input of module.In some cases, vehicle use information to provide human language feedback to driver.In some cases Under, vehicle use information to provide at least one of touch feedback, audio feedback or visual feedback to driver.

In some cases, the retro-reflection element of multiple data riches uses computer vision system observable in infrared spectroscopy Or as it can be seen that human eye observable or as it can be seen that visible with both computer vision system and human eye is used in the visible spectrum.One In a little situations, information relevant to the retroreflective articles of multiple data riches includes expected road worker, expected pedestrian, expected building At least one of worker, expected student, expected emergency answering worker.

In some cases, inference engine is stored locally as the composition of optical system.In some cases, optical system System is communicated using wireless communication protocol with inference engine.In some cases, inference engine includes look-up table, which has Specified meaning associated with the specific pattern of the retro-reflection element of multiple data riches.In some cases, inference engine includes Look-up table.

Foregoing invention content is not intended to each embodiment of the description disclosure.One or more embodiment party of the disclosure The details of case also illustrates in the following description.According to specification and according to claims, the other feature of the disclosure, target With advantage it will be evident that.

Detailed description of the invention

This hair can be more fully understood in the following detailed description for being considered in conjunction with the accompanying various embodiments of the present invention It is bright, in which:

Fig. 1 a to Fig. 1 m shows the various patterns including the retro-reflection element in presently disclosed optical goods.

Fig. 2 a and Fig. 2 b show the presently disclosed optical goods being arranged on object.

Fig. 3 shows the rendering image of the environment of some embodiments according to presently disclosed system, which includes The interested region of shape in the presence of the chaff interferent generated by software and the shape from the image being synthetically generated (ROI) automatically extract.

Fig. 4 depicts description according to the data of some embodiments of presently disclosed system being synthetically generated for use Come the flow chart for the step of assessing the conspicuousness for inputting shape.

Fig. 5 depicts the image (carrier pattern) of interested object.The image bounding box annotation being located at around object, As shown in (a).Training machine learning model is to eliminate the carrier pattern and between the other objects found in the environment in (b) Ambiguity.

Fig. 6 depicts the exemplary modification to carrier pattern.Carrier pattern is shown in (a).Modification to pattern may include The gap among band or the repetition gap as shown in (c) in the form of dividing lop are introduced as shown in (b).

Fig. 7 depicts the image of the example of the different subclass of carrier pattern.

Fig. 8 a and Fig. 8 b are depicted for available image procossing in some embodiments in presently disclosed system Exemplary system.

Fig. 9 shows available possibility shape cloth in the algorithm in some embodiments that can be used for presently disclosed system Set the general view of the appraisal procedure of mode.

Figure 10 is shown in some embodiments in presently disclosed system after available 1484 generation genetic algorithm The design that fitting (or conspicuousness) score of generation is 0.752332.

Figure 11, which is shown, is rendered into 3D back by available algorithm in some embodiments in presently disclosed system The design of Figure 10 on heart model.

Figure 12 shows the exemplary functions that can be used for optimizing the design in presently disclosed system.

Figure 13 depicts available exemplary genetic algorithm in some embodiments in presently disclosed system.

Figure 14 depicts the workflow of available single image example in some embodiments in presently disclosed system Embodiment.

Figure 15 depicts the reality of the workflow of available input picture in some embodiments in presently disclosed system Apply scheme.

Figure 16 is the process program of the system for training and test for observing disclosed product.

Figure 17 a, Figure 17 b and Figure 17 c show the image of three products shown in embodiment 1.

Figure 18 a, Figure 18 b, Figure 18 c and Figure 18 d show the image of four different products shown in embodiment 2.

Figure 19 to Figure 33 shows the illustrative pattern of retro-reflection element shown in embodiment 5.

Figure 34 is to be oriented as being orientated close to " front " incidence angle with the figure for the panel tested for " yaw angle " of modification Picture.From panel to source/distance of video camera is about 152m.

Figure 35 is the enlarged drawing of the test panel of Figure 34, and which identify 0.025m × 0.025m retro-reflection elements.

Figure 36 be with relative to check direction at about 30 degree " yaw angle " orientation be orientated test panel image.

Figure 37 is the enlarged drawing of the test panel of Figure 36, and which identify 0.025m × 0.025m retro-reflection elements.

Figure 38 is the image of the vest of the embodiment 3a and CE3e that are shot in the dark with near-infrared system.

It should be appreciated that without departing from the scope of the invention, using embodiment and structure change can be carried out. Figure is not necessarily drawn to scale.Like numeral used in figure refers to similar composition.It will be appreciated, however, that in given figure Composition, which is referred to, using number is not intended to be limited to the composition marked in another figure with same numbers.

Specific embodiment

The light being incident on product is re-directed Hui Qiguang by optical goods such as retroeflection (retroreflective) product Source.The characteristic leads to many practical applications of the retroreflective articles in traffic and personal security field.Retroreflective articles are currently used in friendship The clothing that logical mark, license plate, pavement marker, building area cone and bucket and user (for example, worker, pedestrian, animal etc.) dress High-visibility material on clothes.

The system interacted with optical goods such as retroreflective articles includes observer's such as computer vision system, optical system And human viewer.Disclosed retroreflective articles are initially use and even can both be seen after extended use by these One or more observers in the person of examining see under various types and horizontal environmental condition.

Multiple retro-reflection elements include retro-reflection element, wherein when ultraviolet spectra, visible spectrum, near infrared spectrum or they Any combination in observation when, these multiple retro-reflection elements relative to background substrate have at least two different retroreflection characteristics and At least two different optical contrasts.

The ultimate challenge of the performance of retroreflective articles first is that can influence the huge of its performance and variation environmental factor.Example Such as, it is expected that retroreflective articles are in full striation part (such as full sun on daytime), full dark condition (such as night Cloudy conditions) and the two Between each situation under work.The tool that can be worked in terms of how particular retroreflective material or their combination can work Body factor may include brightness or its shortage sun of such as sun；Angle and/or position of the sun relative to product；The sun is opposite Angle and/or position in observer；The presence of cloud is not present or their combination；The presence of shade, be not present or they Combination；Angle and/or position of the shade relative to product；Angle and/or position of the shade relative to observer；Shade relative to The angle of product and/or position；The presence of light source in addition to the sun is not present or their combination；Light in addition to the sun Angle and/or position of the source relative to product；Angle and/or position of the light source relative to observer in addition to the sun；Product On moisture content existence or non-existence and intensity；And the existence or non-existence of mist.

Different types of retroreflective articles can have different retroeflection to respond for various environmental factors；Different types of retroeflection Product can have different optical contrasts to respond for various environmental factors relative to its background；And some retroreflective articles can Relative to its background there is different retroeflection responses and optical contrast to respond the two for various environmental factors.In illustrative field Jing Zhong, can have two environmental factors of maximum negative effect to the optical contrast observed may include the sun relative to system The angle of product, observer, or both and the presence of shade.For example, when the sun behind observer but is directly fallen in be viewed When on object, the direct sunlight on the clothes including retroreflective articles can make sensor saturation, and (eyes of human viewer are inhuman The sensor of class observer) so that certain combinations for retroreflective articles and background, may not be used between retroreflective articles and its background The contrast of observation.Similarly, when the sun is behind the object with retroreflective articles, the shade being incident upon on observable side can To be very dark, hence for certain combinations of retroreflective articles and background, it is not present again between retroreflective articles and its background Detectable contrast.When therefore, compared with other combinations under the conditions of such, the retroeflection of Additional optical contrast is provided The combination of product and background can be more advantageous and providing more steady detection under more lighting conditions.

Advantageously, for example, retroreflective articles can give full play to function under all lighting conditions, because for example when the sun exists When under cloud or across sky movement, the user of the retroreflective articles including clothes stops and changes the particular retroreflective product including clothes It is unpractical.

Optical contrast's difference

Disclosed optical goods include at least two retro-reflection elements, and include multiple examples in most of embodiments Such as more than two retro-reflection element.When being observed in ultraviolet spectra, visible spectrum, near infrared spectrum or any combination of them, Retro-reflection element has at least two different optical contrasts relative to background substrate.

Term " ultraviolet " refers to the energy at 10 nanometers (nm) to the wavelength within the scope of 400nm.Term " ultraviolet light Spectrum " refers to the wave-length coverage of 10nm to 400nm.

Term " visible " refers to the energy with usual macroscopic wavelength, and in some embodiments, is Refer to the energy with the wavelength within the scope of 400nm to 700nm.Term " visible spectrum " refers to the wavelength model of 400nm to 700nm It encloses.

Term " near-infrared " refers to the energy with the wavelength within the scope of 700nm to 2500nm.Term " near infrared light Spectrum " refers to the wave-length coverage of 700nm to 2500nm.

In some embodiments, when in ultraviolet spectra, visible spectrum, near infrared spectrum or any combination of them When observation, relative to background substrate two different optical contrasts can by measuring technique, observation or their combination come It determines.

In some embodiments, it when using eye-observation, can be determined by checking two different retro-reflection elements Two different optical contrasts relative to background substrate.In some embodiments, it is seen when using detector or imager When examining, two different optics relative to background substrate can be determined by measurement optical contrast or its at least some component Contrast.

In some embodiments, a kind of method measuring optical contrast includes using colorimeter.Colorimeter can determine " color " of two retro-reflection elements, and then colorimeter can determine whether they are different.Two retro-reflection elements can be used in this Primitive color or by the way that two retro-reflection elements are converted into gray scale and determine whether difference is completed for two kinds of colors.The method can be with It is useful, but in some cases, can be influenced by the difference of the pixel quantity in such as sensor.With pixel quantity Reduction, distinguish two different optical contrasts ability reduce.It in embodiments, can be by being used for using colorimeter The hardware of optical contrast is measured to determine different optical contrast angle value.

In some embodiments, a kind of method measuring optical contrast includes using brightness measurement.This method is available In being only intentional or especially advantageous retroreflective articles for night use, because the difference in brightness can be poor for optical contrast Different overwhelms sexual factor.In some embodiments, the luminous flux of measurable two retro-reflection elements is to determine optical contrast No difference.In embodiments, using luminous flux, different optical contrasts can be determined by the hardware for measuring luminous flux Angle value.

In some embodiments, the difference for being mainly due to the optical contrast of brightness can also be determined by human viewer For the brightness of two different levels.Wherein brightness be exemplary physical stimulation the intensity that is perceived with it of magnitude or intensity it Between a kind of relationship provided by the power law of Steven.The general formula of law is ψ (I)=kI^a, wherein I is the magnitude of physical stimulation, ψ (I) is the subjective magnitude sensed caused by stimulating, and a is the index of the type depending on stimulation, and k is depending on used Unit proportionality constant.Index a substantially indicate to perceive for average human observer its is brighter (for example, its With different optical contrasts) have to it is how bright.In some embodiments, when incentive condition is 5 ° of targets of dark place, 0.33 times of luminance difference can be perceived as brighter by human viewer；In some embodiments, when incentive condition is point source, 0.5 times of luminance difference can be perceived as brighter by human viewer；In some embodiments, when incentive condition is short flash When, 0.5 times of luminance difference can be perceived as brighter by human viewer；And in some embodiments, when incentive condition is When the point source of short flash, 1 times of luminance difference can be perceived as brighter by human viewer.

In some embodiments, a kind of method measuring optical contrast includes using bidirectional reflectance distribution function (BRDF).BRDF provides the reflectivity as illumination geometry and the target for the function for checking geometry.BRDF is depended on Wavelength, and being determined by the structural and optical characteristic on surface, including for example shadow-casting, multiple scattering, mutually masking, transmission, Reflection, absorption and surface element transmitting, facet distribution of orientations, facet density and their combination.BRDF is only with multiple Miscellaneous mode describe eye-observation to the case where.

Under real world conditions the visual contrast between retroreflection material and its background material is quantified can be to have A little complexity.Typical method may include measuring the color (as described above) of retroreflection material.The color of material is generally defined as Simple measured value is made of the combined measurement of definition brightness or intensity and coloration.The example of this measurement is CIE 1931Yxy color space, wherein Y corresponds to brightness, and wherein x and y is limited the chromaticity coordinate of the particular shade of color on fixation degree chart. Usually (such as duplication daylight and specific fluorescent spectral illumination are respectively used to corresponding to the specific illumination source of specific illumination environment D65 and F2) limit color.Coloration measuring color can be used, wherein color is in the clearly restriction for light source and reflected light Orientation under measure.

However, whether standard color measurement will provide good vision pair under the conditions of can not checking outdoors for retroreflection material Good fallout predictor is provided than degree.View under different daylight illumination conditions and orientation, between two reflection opaque articles Feel that contrast is usually simple more more complicated than the color parameter from colorimeter.The reflecting brightness of retroreflection material can be most heavy The factor wanted.However, the degree that light is reflected depends on the local orientation of viewer (or detector, in machine vision) And position of the light relative to the surface normal for illuminating lower material.When taking for any one of illumination, observer or reflecting material To when changing, the brightness of reflection will change.In addition, including depositing for the curvature on the clothes of the natural fringe margin of mankind's form And folding, abrasion or the presence of their combination, relative to human viewer or associated with machine vision can take the photograph Camera generates height local reflex geometry heterogeneous.

Bidirectional reflectance distribution function (BRDF) measurement can be carried out, is carried out if being distributed with the angle to the reflecting brightness from sample The quantization of dry incidence angle, to explore the reflection optical contrast under a variety of observation conditions.Such measurement is referred to alternatively as bright cover Cover degree measurement.For photometry, BRDF is defined as reflecting brightness divided by incoming illumination, and will be referred to as photoelectricity BRDF.BRDF has the unit of back projection's three-dimensional viewpoin, i.e., against sterad.Each BRDF data point and incident light direction and scattered It is associated to penetrate light direction.Scattering direction is specified using projection angle space, which is to have in an x-y plane The border circular areas of radius 1.The direction z corresponds to the direction perpendicular to sample.BRDF is integrated in entire scattering angle spatially to mention For total reflectivity (TR).For the arbitrary point in projection angle space, inclination angle (z-axis and scattering side are equal at a distance from origin Angle between) sine value, and put and provide azimuth direction of the scattering light relative to x-axis relative to the azimuth of x-axis. It is that the area in the region in projection angle space is equal to project stereoscopic angle using the advantages of projection angle space.Projection is vertical Body angle obtains the score for the incident light being reflected into the angular zone multiplied by the average BRDF in project stereoscopic angle.Projected angle Spending space has u_XAnd u_YCoordinate, wherein

u_X=u*sin (φ) u_Y=u*cos (φ) u=sin (θ)

θ (Xi Ta) is inclination angle, and φ (phi) is azimuth.Projection angle space will be referred to as u_X-u_YSpace.

For fixed incidence angle, it can measure and measure photoelectricity BRDF.By with u_X-u_YSpecified region (u < sin in space (70 °) and u_Y< θ -15 °) in predetermined critical threshold value is taken to photoelectricity BRDF, hundred of the specified region higher than threshold value can be calculated Divide than and " bright coverage " can be referred to as.

In some embodiments, when in u_X-u_YThe photoelectricity BRDF on sufficiently large region in space below about 0.1 ± 0.05 against sterad when, retroreflection material can be distinguished with background.The surface normal of retroreflection material can be around main body and with one day Time and change, so as to cause to u_X-u_YThe big region in space is sampled.If bright coverage is high, in the sun, The retroreflection material of larger proportion will be bright (higher brightness), and the dress materials for being more difficult to and becoming clear distinguish.

In some embodiments, the difference of the optical contrast of two different retro-reflection elements can with BRDF critical value come Description, the BRDF critical value is for calculating bright coverage.For example, 0.05 bright coverage defines the solid using 0.05 The bright coverage that the BRDF critical value of radian calculates.In some embodiments, when two retro-reflection elements with 0.05 it is bright When bright coverage is with weighted average less than 50%；When they are flat with 0.05 bright coverage and the weighting less than 25% When mean value；When they with 0.05 bright coverage with weighted average less than 15% when；When they with 0.10 it is bright When bright coverage is with weighted average less than 50%；When they are flat with 0.10 bright coverage and the weighting less than 25% When mean value；When they with 0.10 bright coverage with weighted average less than 15% when；Or when they have 0.15 Bright coverage with weighted average less than 15% when, they have different optical contrasts.

In some embodiments, retroreflection characteristics can also have at least two different colors.In some embodiments, At least some of retro-reflection element retro-reflection element is black, and at least some of retro-reflection element retro-reflection element is silver color 's.In some embodiments, at least some of retro-reflection element retro-reflection element is black, and in retro-reflection element at least Some retro-reflection elements are silver color, and the background that is located at of multiple retro-reflection elements or substrate are fluorescence.In some embodiment party In case, at least some of retro-reflection element retro-reflection element is black, and at least some of retro-reflection element retro-reflection element is Silver color, and the background that is located at of multiple retro-reflection elements or substrate are fluorescent orange or fluorescence lime yellow.

Retroreflection characteristics

Retro-reflection element has at least two different retroreflection characteristics.Term " retroeflection " refers to that energy (for example, light) is hit Surface and it is redirected to the phenomenon that returning to energy source.The retroeflection property of element can be by being caused with any energy source lighting.In In some embodiments, the retroeflection property of element can be caused by natural source (for example, the sun, moon etc.)；Or by non-natural origin (for example, headlight, street lamp etc.) causes.

Multiple retro-reflection elements include the element at least two different retroreflection characteristics.In some cases, at least two The different retroreflection characteristics of kind are at least two different backreflected intensity values.In some cases, at least two different retroeflection are special Property be at least two different reflection wavelengths.In some cases, at least two different retroreflection characteristics do not have at least two not Same polarization state.In some cases, at least two different retroreflection characteristics are at least two different phase delays.Some In embodiment, at least two different retroreflection characteristics are at least two different backreflected intensity values, at least two different anti- Ejected wave length, at least two different polarization states, at least two different phase delays or their some combinations.

It is observed in ultraviolet spectra, visible spectrum, near infrared spectrum, or both

In some embodiments, optical goods, its optical contrast or their some combinations can be by can be in purples Its system is observed in external spectrum to observe.In some embodiments, optical goods, its optical contrast or they one A little combinations can be observed by that can observe in the visible spectrum its system.In some embodiments, optical goods, its light Learning contrast or their some combinations can be observed by that can observe its system near infrared spectrum.In some realities It applies in scheme, optical goods, its optical contrast or their some combinations can be by can be in ultraviolet spectras, visible spectrum It is observed with its system is observed in some combinations of near infrared spectrum.In some embodiments, optical goods, its optics pair It can be by can be in visible spectrum and near infrared spectrum, visible spectrum or near infrared spectrum than degree or their some combinations Its system is observed in some combinations to observe.

Observe herein may include be perceived, visualize, checking, being imaged, detecting, monitoring or any combination of them.It sees Examining can be by human viewer；Machine or computer observer, imager, detector, visualization device or their combination；Or The combination of the person mankind and machine carries out.

Machine observer may include at least one video camera.The example of RGB video camera system includes such as FLIR Machine Vision (Point Grey originally)3 5MP colour TV camera (CM3-U3-50S5C- CS) (the FLIR Integrated Imaging Solutions company (FLIR of Canadian Richmond British Columbia Integrated Imaging Solutions Inc.,Richmond BC,CANADA)).Camera chain is available to have difference The camera lens of focal length is modified, including such as Edmund Optics 25mm C series fixed focus lens (New Jersey Ba Ling The edmond Optical Co., Ltd (Edmund Optics Inc.Barrington, NJ) paused).Example for RGB video camera system The property shown light source can be ambient lighting or automobile headlamp.The example of NIR camera chain includes such as FLIR Machine Vision (Point Grey originally)3 5MP colour TV cameras (CM3-U3-50S5C-CS) (Canada in scholar FLIR Integrated Imaging Solutions company (the FLIR Integrated of full British Columbia Imaging Solutions Inc.,Richmond BC,CANADA)).NIR camera chain is available have different focal length and/ Or the camera lens of optical filter is modified, including such as Edmund Optics 25mm C series fixed focus lens (New Jersey The edmond Optical Co., Ltd (Edmund Optics Inc.Barrington, NJ) of Barrington), the narrow near-infrared band logical of BN940 Optical filter (BN940-25.5) (Illinois para court of a feudal ruler Midwest Photonics Systems Inc. (Midwest Optical Systems, Inc.Palatine, IL)) or their combination.Illustrative light source for NIR camera chain may include coming From the 130mm of intelligent vision lamp company (state of Michigan maschi root) (Smart Vision Lights (Muskegon, MI)) Over Drive Ring Light940nm IR-“EZ Mount Ring Light”。

Disclosed retroreflective articles can be to be advantageous, because no matter whether a part of retroreflective articles is blocked, they are all Observation can be provided for both human viewer and machine observer under many different environmental conditions.

Angle sensibility

In some embodiments, multiple retro-reflection elements can also have at least two different angle sensibility.Different angles Sensibility is caused by the retroreflection characteristics of retroreflection material.Pulsatile effect can be generated using the various combinations of retro-reflection element, do not generated Pulsatile effect or their combination；It generates flicker effect, generate flicker effect or their combination；Or their combination. Pulsatile effect is partly caused by excessive luminous (overglow), because brighter returns when a retro-reflection element stops reflection Luminous seem (at the special angle) of excess for penetrating element becomes brighter, such as pulse.When the angulation change of incident light is (logical Cross product is mobile or light source is mobile, or both) when, one or more retro-reflection elements in retro-reflection element will have retroreflective change Change.In some embodiments, when with energy source lighting, multiple retro-reflection elements can produce pulsatile effect.Pulse can be Benefit, because it can distinguish an object and another object, the object from the mankind or two kinds of other types of objects It opens.

Different angle sensibility can be from different shapes, different sizes, different patterns or any combination of them group Close, to generate different types of effect, including such as pulse, flash, stroboscopic light effect, flashing.Because the difference of material is partially Angle retroreflection characteristics can produce the different appearances of monolithic article at different angles.For example, being reflected in all material good At positive angle, it is seen that full array, and at the angle of positive side (drift angle), have low drift angle retroreflective Material do not reflect very much, therefore such retroreflection material seems disappearance at drift angle.Have studied different pattern (such as chesses Disk, target center and column pattern), and can provide different effects, all those of findings in this example as follows.

When (by machine observer, human viewer, or both) observation is to infer position of the observer relative to fixed object Set, angle etc. whens, the different angle sensibility of at least two retro-reflection elements can also be used.For example, different angle sensibility is available In determining observer at a distance from object, because angle sensibility depends on distance, because effect (for example, flashing, pulse etc.) is only Occur in the distance range away from retro-reflection element.

The arrangement spatially limited

The disclosure provides a kind of optical goods, which includes the retro-reflection element of multiple data riches, multiple number It is configured according to retro-reflection element abundant with the arrangement that spatially limits.

As used herein, term " data rich " refers to the information for being easy to machine interpretation.

At least some of retro-reflection element retro-reflection element is discontinuous.In some embodiments, multiple retro-reflection elements In all retro-reflection elements be discontinuous.The so-called discontinuous edge for referring to retro-reflection element does not contact each other and retroeflection Element is not overlapped.

In some embodiments, the retro-reflection element of multiple data riches is matched with the duplicate arrangement spatially limited It sets, so that information can be explained when a part of retro-reflection element is blocked.In some embodiments, multiple retroeflection members At least some of part retro-reflection element is discontinuous.In some embodiments, most of retroeflection in multiple retro-reflection elements Element is discontinuous.In some embodiments, all retro-reflection elements in multiple retro-reflection elements are discontinuous.

Fig. 1 a shows the effect blocked with pattern duplication.Left upper quadrant shows the sample basis pattern of retro-reflection element. In this example, imagine, if two in circle circle be it is visible, pattern be it is identifiable；Right upper quadrant is shown The same basic pattern replicated with one is gone out；In the left lower quadrant of Fig. 1 a, the white object element in shielding patterns In an element.In this case, it blocks and leads to not detection pattern.In right lower quadrant, white object blocks again An element in the element of pattern, but due to duplication, however it remains enough elements by system such as vision for being examined Examining system detects the patterns of optical goods.The retro-reflection element of various patterns, such as Fig. 1 b to Fig. 1 m can be used in the disclosure Shown in exemplary design.

In some embodiments, can exist thereon using the arrangement specifically spatially limited to retroreflective articles Entity (for example, user or object) is classified.If the classifier in system such as vision detection system is based on for minimum The quantity of required retro-reflection element carrys out the quantity of the retro-reflection element in check pattern, then can be detected under partial occlusion including than The pattern of at least one more element of specified minimum.In contrast, when at least one retroeflection in the retro-reflection element in pattern When element is blocked, system, the classifier in this vision detection system is not steady when finding certain amount of retro-reflection element 's.

The disclosure, which also provides multiple retro-reflection elements, can have identical or different shape.The available shape of individual retro-reflection element Shape includes but is not limited to circle, star, square, polygon, curved shape and irregular shape etc..These individual retroeflection members Part can arrange that the mathematical way of shape is arranged, so that can carry out detection arrangement mode independently of individual formed shape, appoint The individual formed shape of selection of land can add additional information.Mathematics arrangement refers to carry out for the composition to gained optical goods Scale cun and scheme spaced apart.

These retro-reflection elements or gained optical goods can be independent or can be to be duplicate to increase to the steady of partial occlusion Strong property.If element or product are small, can need to repeat to be used for robust detection, if optical goods are big, since subset is It is visible, therefore it may be steady for partial occlusion.

Optionally, any amount of formed shape can be engineered with selectively reflect different wave length and/or partially The light of vibration.For example, in some embodiments, retro-reflection element, which has, is sufficient for adjustment criteria (for example, ANSI/ISEA 107- 2015' conformable material) characteristic but the subsets of optical goods be constructed such that it with special optical characteristics (for example, anti- The wavelength and/or polarization penetrated) so that system (such as computer vision system) can distinguish these sections and optics of optical goods The rest part of product or object mounted thereto.One example of the effectiveness of this construction can be, if to be adjusted, Then the submissive gap in retro-reflection element is necessarily less than Xmm, but computer vision system detection needs gap to be greater than Xmm.The two It is required that will clash, unless the construction permission computer vision system of retro-reflection element only sees the subset of retro-reflection element, but whole A (or at least part of optical goods or retro-reflection element) is sufficient for standard, because gained optical goods by the mankind to can be examined The light in spectrum measured has reflectivity.

In some embodiments, the quantity of the unique retro-reflection element in optical goods should to until retro-reflection element it is complete It blocks entirely or deformation that they start the degree merged relative to the density of bright pixels and visual angle change is steady.

The spacing and characteristic size of retro-reflection element (or shape) including optical goods will likely need to be included in excessive shine. One optional configuration of the disclosure may include retro-reflection element, the retro-reflection element by more than one level reflecting material construction so as to Reduce excessively luminous influence.For example, the outer edge of retro-reflection element can be by lower RA compared with the interior section of retro-reflection element Material construction.In some embodiments, the minimum measurement difference (such as at least 5%, 10%, 20%, 50% or more of RA Difference) it is useful.

Retro-reflection element can be manufactured by any amount of method, including but not limited to: silk-screen printing, braiding, suture etc..

In some embodiments, optical goods are deformable optical goods.In some cases, deformation be by shrink, Caused by expansion, or both.In some cases, deformation leads to the spacing between at least two retro-reflection elements in retro-reflection element Variation.In some cases, deformation is reversible.

In some cases, aforementioned retroreflection characteristics change in response to condition and are changed.For example, can lead to multiple retro-reflection elements The condition variation of at least one of retroreflection characteristics retroreflection characteristics variation can be heat, moisture, mechanically deform or the variation of radiation. Thermal change can be the variation of such as environment temperature.Exemplary moisture changes the variation including ambient humidity or uses optics wherein There are precipitation in the environment of product.Mechanically deform may include the corrugation of such as optical goods clothes mounted thereto.

In some cases, retro-reflection element is individually sized and is separated from each other, so that each individual retroeflection Element can be punished in the desired distance away from optical goods and be distinguished.

In some cases, the arrangement spatially limited includes geometrical arrangements mode, wherein retro-reflection element be oriented with Their adjacent retro-reflection element has a distance, and has in the arrangement that wherein retro-reflection element spatially limits from a member Periodicity of the part to another element.It in some cases, is periodically regular periodicity.In some cases, periodically it is Property irregular cycle.In some cases, spatially limit be arranged as rotating it is insensitive.

In some cases, the quantity of geometrical arrangements mode needed for the arrangement each spatially limited depends on required Fit quality.In some cases, retro-reflection element be oriented the nearest adjacent retro-reflection element with them at a distance of a feature away from From.In some cases, retro-reflection element has size with to the characteristic rate at a distance from adjacent retro-reflection element, and this feature ratio is not With visual angle change.

In some cases, optical goods further include printing layer, which is arranged at least part of retro-reflection element Outer surface on.In some cases, retroreflection characteristics can detect in infrared spectroscopy.

In some cases, optical goods be arranged on substrate, the substrate be selected from object, infrastructure, wearable object and At least one of vehicle.The disclosure provides the fabric including aforementioned articles.

System including retroreflective articles

The disclosure further includes a kind of system, which includes any product, optical system and reasoning in aforementioned articles Engine, which is used to explain and multiple retro-reflection elements of classifying, and wherein optical system feeds data to inference engine.One In a little situations, product is arranged at least one of object, infrastructure, target, wearable object and vehicle.

In some cases, optical system is a part of vehicle, and in addition wherein vehicle use information is used as independently Drive the input of module.In some cases, vehicle use information to provide human language feedback to driver.In some cases Under, vehicle use information to provide at least one of touch feedback, audio feedback or visual feedback to driver.

In some cases, the retro-reflection element of multiple data riches is visible to computer vision system in infrared spectroscopy. In some cases, information relevant to the retroreflective articles of multiple data riches includes expected road worker, expected pedestrian, expection At least one of construction worker, expected student, expected emergency answering worker.

In some cases, inference engine is stored locally as the composition of optical system.In some cases, optical system System is communicated using wireless communication protocol with inference engine.In some embodiments, inference engine and optical system may include as For method and system available in the disclosure with various features and step disclosed in lower curtate.

Presently disclosed system can be used for various applications.For example, presently disclosed system utilizes presently disclosed optics system Product simplify and enhance the detectability of system, such as computer vision pedestrian detection, which allows system to determine wearing light Learn position, identification and/or the posture of the individual for clothes, attachment or the other objects that product is disposed thereon.Multiple retro-reflection elements In the content of data rich facilitate the quantity by reducing chaff interferent to simplify the task of pedestrian detection, optical system needs The chaff interferent is assessed by following steps: being primarily based on the optical characteristics (intensity and/or color of the light such as returned Spectrum) threshold value is taken to image, and be then based on clothes, attachment or other products that optical goods are disposed thereon design and The possibility posture of wearer assesses those of significant shape (or pattern) cut zone.

Presently disclosed system includes at least one video camera, light source (such as, front lamp of vehicle or other visible NIR or FIR light source) and presently disclosed optical goods.Presently disclosed system utilizes the pattern of the light returned from optical goods Come a possibility that identifying object that optical goods are disposed thereon, infer posture, position, intersect etc..One possible embodiment party Case can utilize dress designing, those of shown in such as Fig. 2 a and Fig. 2 b.In this example, the front view of clothes, which has, is different from clothes The visible optical product of the quantity and pattern with multiple retro-reflection elements of the profile view of dress.If the optical goods on clothes With known size (for example, if all 6 inches of chevron in this case are long), system can be based on projection The relative distance and position of size and location estimating wearer and video camera.

The disclosure includes for automatic in the case where real world data (image/video) without collecting such shape Assessment design shape (optical goods being such as included therein and multiple retro-reflection elements) is for the conspicuousness of application environment System and method.

The step sequence for executing this method is described in Fig. 3 a and Fig. 3 b；And it is described here:

Input to system is interested shape, such as optical goods and the multiple retro-reflection elements being included therein.It is right In application environment, the one group of chaff interferent shape (or object) usually occurred in the environment is known, such as open-road applications, Chaff interferent collection may include highway information mark, speed(-)limit sign, cone, bucket etc..

Design shape (such as optical goods that interested object (infrastructure, clothes, attachment etc.) will be arranged With the multiple retro-reflection elements being included therein) and chaff interferent collection be input to the calculation of the generated data collection for generating image and video In method or software.This includes but is not limited to that the 3D model of use environment generates the Rendering software of rendering object in this context.This The data of the analog effect such as variation of illuminating effect, viewpoint, the mixed and disorderly, object of which movement of environment etc. will be generated.Fig. 3 a shows wearing The sample of the highway worker of high-visibility clothes renders image, wherein the exemplary optical article of the disclosure for he clothes just Design shape in the part of face.

Correspond to the sense of design shape (for example, optical goods and the multiple retro-reflection elements being included therein) from image zooming-out The region (ROI) of interest and chaff interferent.Fig. 3 b shows such example of the ROI from rendering image zooming-out.This method can It is automated using the knowledge about the 3D model provided for rendering contexts.

For the ROI of each extraction, the spy of their characteristic of computational representation (such as appearance, shape, texture, geometry) Sign, for example, Shape context, the histogram, the area that are orientated gradient etc..

Then calculated feature can be input in algorithm, the example of the algorithm is shown in FIG. 4, which can be directed to This group of chaff interferent shape generates the conspicuousness point of design shape (for example, optical goods and the multiple retro-reflection elements being included therein) Number.Significant assessment generates unique quantitative score of the design shape in this group of chaff interferent.

The disclosure also provides a kind of system and method, and the system and method are in interested object (such as infrastructure, clothes Dress, attachment etc.) on modify retroeflection shape (such as optical goods and the multiple retro-reflection elements being included therein) to provide additional letter Breath.In the present invention, interested object is also referred to as carrier pattern.Interested exemplary objects or carrier pattern include work Make in region worker wearing high-visibility safety vest, in roadside buildings area for mark navigate boundary bucket and Other infrastructure, clothes, attachment etc..There has been described the step sequences for executing this method:

Collect the annotating images of the carrier pattern of environment.These include the image of the object of different distance, posture and viewpoint. As an example, Fig. 5 includes the example of the retroeflection vest of the individual worker wearing in workspace.

Machine learning model can be trained for image patch to be classified as carrier pattern or without classification.For training The model provides the image patch of carrier pattern and background (not including the image patch of carrier pattern).These figures of computational representation As the characteristics of image of the appearance of patch, it is such as orientated the histogram of gradient (HOG) or Shape context.Then come using these features Training sorter model, such as support vector machines (SVM) or decision tree.The input of the model is the calculating feature of image patch, And exporting, which can be (but are not limited to), answers the existing Yes/No of the carrier pattern in input picture patch.

Given carrier pattern and the requirement based on system to environment, modify to the retroeflection shape of carrier pattern.Scheming Example is provided in 6a, Fig. 6 b and Fig. 6 c, wherein the H shape used in safety vest is partly modified to generate the two of pattern A additional sub-categories are other.Modification is not limited to size, and may also comprise the color for changing pattern.

The image of different subclass is collected in data collection experiment or through generated data generation module.In addition to individually receiving Except the image for collecting different subclass, carrier pattern image be also possible to included subclass example, and can be used poly- Class algorithm finds these examples.

Subclassification classifier is trained using the example of different subclass, as shown in Figure 7.

At runtime, system first looks for the presence of carrier pattern.Having detected that the carrier pattern in image patch In the case of, then pass through the subclass resume module image patch for subclass present in image.Example is in Fig. 8 a and figure It is provided in 8b.

In some embodiments, presently disclosed system also provides or including two kinds of algorithms, the algorithm is for 1) initial Change the boundary and 2 for the shape of optical goods being placed on interested object (such as clothes)) it limits and measures the boundary and match The objective function of the serviceability or fitness set.The space of the possible geometry of each algorithm search in algorithm and generation Optimize the geometry of the objective function.

Fig. 9 shows the method for assessing each possible geometry (parameter turns to one group of [x, y] point).In some embodiment party In case, one of algorithm algorithm is genetic algorithm, and another algorithm is the optimization algorithm based on numerical gradient.These algorithms In each algorithm sample geometry is generated using different technologies, they are assessed, and attempts further raw At the new arrangement with improved assessment score.

In some embodiments, multiple retro-reflection elements are placed with the configuration for generating design (such as dress designing), should Design is height (referring to Figure 10) outstanding to system (system that such as motor vehicle uses).As used herein, " height is prominent for term Refer to other entities or feature thing outstanding from environment out ".Objective function is applied to by that will be configured as texture The 3D model (for example, that vest worn by construction worker) of vest assesses the conspicuousness of design.3D modeling application program (for example, blender) is used to generate several different views of the 3D model (referring to Figure 11).Resulting view and one group is " dry Disturb object " shape is fed in clustering algorithm.Chaff interferent shape depends on application space.In some embodiments, chaff interferent shape Shape is the object that interested object can be confused in presently disclosed system and method.Clustering algorithm divides these inputs Group is cluster.

In some embodiments, cluster by these design in each design be accurately classified as a cluster, and And by each Shape Classification in chaff interferent shape be another cluster.This leads to 1.0 fitness.Fitting can pass through " Silhouette Score " quantifies, and the quality of one group of cluster is measured based on known ground true tag.In other words, Silhouette Score is used to measure the executive condition of clustering algorithm.There are the methods of other potentially usefuls to quantify a group The quality of collection.

In some embodiments, it can be used for generating design for the SciPy Optimization Toolbox of Python, as ours A part of Proof of Concept experiment, wherein using the objective function for generating circular shape.Scipy function is referred to as scipy.optimize.minimize.The starting configuration that the function is supplied with the 1) boundary of the polygonal shape of restriction design is (all Such as use multiple retro-reflection elements optical goods) [x, y] point list, 2) quantify the design specific configuration cost mesh Scalar functions, the lower intermediate value the better, and 3) it is the specification for the optimization method for optimizing and using and the column that 4) shape or size constrain Table.

In some embodiments, the list of option of the optimization method in document is (for example, sequence least square is compiled Journey).Constraint can be defined with any of constraint for listing in constraints graph 12 or all constraints.Genetic algorithm can be used for determining Possible data structure.Data structure is referred to alternatively as chromosome (container for being analogous to inhereditary material in biosystem).

Genetic algorithm generates multiple chromosomes (completely random or being changed at random by designing seed).Then it determines The fitness of each chromosome.It deletes the chromosome of fitness difference and is replaced with the copy of height fitting chromosome.Latest copy makes It is modified with mutation operator.Operator is mutated to the random variation of some values application in the value in chromosome.Copy can make It is generated with the operation for being referred to as exchange, wherein each filial generation obtains the inhereditary material from multiple parental generations, although simultaneously not always needing It exchanges.

In some embodiments, chromosome is the list of point.It includes the light with multiple retro-reflection elements that each point, which limits, Learn the vertex of the shape of product.The geometry that genetic algorithm is conducive to have high fitness is (for example, in this case, have There is the fitness closest to 1.0).Geometry with high fitness tends to remain in group, and due to selection operation, tool There is the geometry of low fitness to tend to be excluded except group.

Figure 13 describes genetic algorithm (GA).The group of chromosome can random initializtion or using the chromosome evolved in advance into Row initialization.Alternatively, the chromosome that the top n from one group of thousands of chromosome generated at random can be used most to be fitted is to group Body is initialized.Similar to numerical optimisation algorithms, genetic algorithm has used conspicuousness objective function.It can be with modified objective function To apply hard constraint or soft-constraint to design.Hard constraint ensures the consistency designed.GA will design " propulsion " using soft-constraint and arrive Desired attribute, but not fully exclude outlier design.

1) height and width

2) area (minimum and/or maximum-to meet supervision standard)

3) (that is, in order to meet ansi standard, there are materials for pressure on shoulder for presence of the retro-reflection element in some regions Material)

4) exposure mask is applied to design, to limit vest region

Chromosome with minimum fitness is replaced by the copy of the chromosome with highest fitness.Referring in Figure 13 Step C and D.This purpose can be completed in various ways.In some embodiments, using the single of the match size with 4 Match selection.This method needs for each chromosome to be assigned randomly to 4 group.Two chromosomes inferior are replaced in the group Two excellent heterosomal copies.These copies can be the accurate copy of two advanced parental generations, or can be used from every Some inhereditary materials of a parental generation create each filial generation.The later approach, which is referred to as, exchanges (referring to the step E in Figure 13).So Filial generation is mutated (referring to the step F in Figure 13) afterwards.In the case where our Proof of Concept embodiment, mutation is related at me Chromosome in the one or more vertex [x, y] of random perturbation.

Finally, it is determined whether met termination criteria (referring to the step G in Figure 13).It can instead of be completed afterwards in predetermined quantity Algorithm terminates.

The disclosure also provides using retroeflection the system and method for training object part detector.As used herein, term " object part detector " is to refer to find the separate part of object in image/video rather than find whole object itself Detector.

Optical goods with retroreflection characteristics seem bright in wherein light source projects image on them.Therefore, When the image of these optical goods takes intensity threshold, which can be rendered as the group of connection in resulting binary picture At.In the disclosure, which is used to the component of optical goods dividing (if present).There has been described execute the party The step sequence of method, and depict in Figure 14 the sample workflow of the single-instance of optical goods.

Input picture is provided.Image is annotated (such as with the bounding box position of entire interested object (such as optical goods) Shown in step (a) in Figure 14).Note that annotation does not include any information, for example, object component counting or position.Scheming It carries out taking intensity threshold and morphological operation such as closure as on (it includes expansion and corrodes).These provide binary picture (such as Gynophore runs multiple threshold values, then is multiple images), wherein the composition connected provides image patch.It can be by this group of image patch Be divided into two groups-with bounding box annotate all image patch (steps (e) in such as Figure 14 that are no any Chong Die and constituting background It is shown).Another group includes the steps that really being annotated with ground with some overlapping patches (in such as Figure 14 shown in (d)).Have This group of patch of overlapping can be trimmed by using scale cun heuristic, the noise patch left using eliminating the artifact in the form of Piece.This group of component parts may include in whole object duplicate pattern (as shown in the example in the step (a) in Figure 14) or Different components.These can be found by that can determine the clustering algorithm of the quantity of the component of object.The quantity of component parts It can also be provided by human supervision.Finally, training detector model (in such as Figure 14 for the component parts of found object The step of (f) shown in).The model is trained for detecting the particular elements of interested object in scene.

In some embodiments, characterization has the potential method packet of the presently disclosed optical goods of multiple retro-reflection elements Include distribution function.(setted wavelength with certain strength and/or partially is reflected for example, it can be characterized by retro-reflection element or feature Vibration), certain distribution of spacing and relative position with the distribution of certain size and constituent element.The characterization of the type is available In enabling additional capabilities, such as object classification is (for example, to the associated feature of a type objects and related with the second type objects Another feature of connection) or enabling product certification.Its feature can also be point generated from the dimensionless ratio determined by constellation Cloth.For example, the size of node is divided by the distance for arriving next nearest node.

In the disclosure, only the part for being enough accurately to be sampled distribution of optical goods is only required for classification 's.For example, if optical goods include many element X of a part for constellation, for significant in the statistical significance of group Sample (that is, n < < X), can only need a small amount of visible element.When the view of product is at least partially obscured or distorts, this will change The robustness of kind classification.

The disclosure also provides the system and method for carrying out the detection based on component using retroeflection.The system in combination is especially Two kinds of characteristics of the optical goods with retroreflection characteristics: under certain morphological operations on the intensity threshold image of optical goods, The composition of resulting connection may include following whole object: certain optical goods are made of component parts, or can be modified For the composition of duplicate components, and when optical goods in its posture partially visible or when being blocked by other objects, these components In some components will be visible.

Both characteristics can search for interested complete object (such as presently disclosed light by operation monolithic detector Infrastructure, clothes, attachment or the other objects that product is disposed thereon) and by it and find one or more constituting portion The detector combination of part come using.The step sequence for executing this method is depicted in Figure 15 and is described here:

Input to system is the image of scene, wherein (such as presently disclosed optical goods setting of interested object Infrastructure, clothes, attachment or other objects on it) can exist together with detector model, which is instructed Practice for finding the entire optical goods being arranged on interested object and individually finding its component parts.Due to posture Or block, the optical goods on object can be fully visible or partially visible.May include the image patch of optical goods with Two ways generates: taking intensity threshold (shown in the step of such as Figure 15 (b)) or and form by help to divide component parts Operative combination takes threshold value (shown in the step (c) in such as Figure 15).

One or more element detectors are run in the first candidate pool, because they are trained for finding optical goods Smaller group at component (step (d) shown in) in such as Figure 15, and what whole object detector extracted after morphological operation Operation on image patch (shown in the step (e) in such as Figure 15).Finally, the output of two different detectors frames of combined running (shown in the step (f) in such as Figure 15).Even if entire optical goods can not detected by monolithic detector, the detection based on component Device will be seen that some optical goods in optical goods, thus the presence of indicating article in the scene.

Figure 16 also provides for training and tests the process program of the system for observing disclosed product.Figure 16's is upper Half part describes the training of system.In some embodiments, general+(correction data) and-(negative data) is fed in training.Instruction White silk can carry out some iteration based on training problem planning.The lower half portion of Figure 16, which describes, is fed to test set by training Algorithm and the test phase assessed.If test phase determines that identification not yet meets graphical indicator, can be added Training (for example, via discribed method in the top half of Figure 16).

Inputting (training image or test image) can be multiple images, wherein each image includes real world or synthesis ring Livery configuration in border.Obtain correction data collection may include interested object (for example, different gestures, viewpoint, away from From the safety vest under, illumination and obstruction conditions) on collect the real world and/or composograph/video of design.Obtain negative It may include the image/video for collecting the other objects found in application environment according to collection.(+/-) input data includes wearing It is present in the image of the individual of certain design (such as standard H design) in some environment (for example, close to road).Including clothes The component of image indicate that (+ve) data and all other region indicate (- ve) data.The computer of the mankind or pre-training Region of (such as exploratory guiding) the labeller label in these images including clothes (+ve data) is (human expert's Under help).Based on these labels, computer is from image zooming-out correction data and negative data.

Then using all data (training set) come training algorithm, target is that positive number strong point and negative data are distinguished by association Point (image-region).Then training algorithm is assessed for independent test set.

Lower-left block in Figure 16 shows positive image/video and the independent collection of negative image/video of application environment Test data set (for example, test image).Test set can be collected in an identical manner as training set.Computational algorithm can be The classifier algorithm that interested object is trained is directed to using training dataset.Every kind of clothes are configured, it can be based on use Computational algorithm is developed in one group of training image of clothes configuration.It later can be the one group of test chart configured for the clothes Picture, this group of test image are run by the computational algorithm of the input exploitation of the training image configured by the clothes.Existed by the mankind The physical tags marked on clothes keep hiding (BLIND STUDY) from algorithm and can be used as assessing the benchmark of its performance.Curve Under area (AUC) or it is other measurement can be used for being quantitatively evaluated.

(such as it can be divided by the way that classifier to be applied in one or more test data sets to and calculated the measurement for being used to assess Class accuracy false positive rate, precision, recalls, AUC) obtain the output for carrying out system since then.Final output includes for assessing The assessment measurement for the effect of clothes configure.Output may include classification accuracy, false positive rate, specificity and sensibility.

Although this document describes a particular implementations of computing system, upon reading this disclosure, with the disclosure Range is consistent and the other configurations and embodiment of computing system within the scope of this disclosure are to those skilled in the art It will be apparent for member.In the case where without departing from the scope and spirit of the invention, the various modifications of the disclosure and repair Change will become obvious to those skilled in the art.