阅读说明：本技术 一种基于机器视觉相机图像处理的钢坯定位方法及应用 (A kind of steel billet localization method and application based on the processing of machine vision camera image ) 是由 王喜 帅红 郑超国 于 2019-08-28 设计创作，主要内容包括：本发明提供一种基于机器视觉相机图像处理的钢坯定位方法及应用,包括以下步骤：图像采集,图像变换,图像增强,图像分割,图像识别,钢坯定位,应用；该基于机器视觉相机图像处理的钢坯定位方法及应用设计合理,使用机器视觉相机进行图像采集,优化钢坯定位算法,提升深挖钢坯定位的使用面,实现了高精度、强实时性、安装方便的新式钢坯定位系统,适用面广泛,适合大量推广。(The present invention provide it is a kind of based on machine vision camera image processing steel billet localization method and application, comprising the following steps: Image Acquisition, image transformation, image enhancement, image segmentation, image recognition, steel billet positioning, application；The steel billet localization method and application design based on the processing of machine vision camera image are reasonable, Image Acquisition is carried out using machine vision camera, optimize steel billet location algorithm, promote the using face for deep-cutting steel billet positioning, realize high-precision, hard real-time, New steel billet positioning system easy for installation, applicable surface is extensive, is suitble to large-scale popularization.)

1. a kind of steel billet localization method and application based on the processing of machine vision camera image, which is characterized in that including following step It is rapid:

Step 1: Image Acquisition；Image is acquired using machine vision camera, machine vision camera has the following characteristics that

1.: the stable and reliable for performance of machine vision camera is easily installed, camera structure it is compact it is solid be hardly damaged, running hours Between it is long, can be used continuously for a long time under poor environment；

2.: the aperture time of machine vision camera is very short, can capture swiftly passing object, common camera it is general it is per second most More 25 frames, machine vision camera can then accomplish thousands of frame/seconds, thus have very high-precision to movement steel billet positioning；

3.: the imaging sensor of machine vision camera is progressive scan, and the imaging sensor of common camera is that interlacing is swept It retouches, the imaging sensor production technology of progressive scan is more complicated, but imaging effect is more preferable, especially in image processing algorithm The information such as sensitive edge, angle point more comprehensively, more can accurately orient location information, surface information and the direction letter of steel billet Breath；

4.: the output of machine vision camera is uncorrected data, spectral region also tend to it is wider, from far infrared, near-infrared, Light, ultraviolet etc. have, and compare the image processing algorithm for being appropriate for high quality, and the picture of other cameras shooting, spectrum model It encloses and is only suitable for human eye vision, after video compress, picture quality is poor, is unfavorable for analysis processing, therefore machine vision camera It can be widely applied to that cold base (room temperature steel billet), hot base (steel billet that the continuous casting stage has just produced), hot material (roll by the steel rolling stage Steel billet during system) etc. in each stage steel billet position fixing process；

Step 2: image transformation；Since pattern matrix is very big, directly handled in the spatial domain, it is very big to be related to calculation amount, because The processing of spatial domain is converted to transform-domain processing, can not only reduce calculating by this often using the method for various images transformation Amount, and can get more effectively processing, it, can be directly in image by Hough transform in steel billet positioning image processing process The pixel data for meeting steel billet shape information is extracted in number, thus quickly position steel billet possible position, principle be by Simple cumulative statistics is carried out in parameter space, then detects straight line in the method that Hough parameter space finds accumulator peak value, The essence of Hough transform is to cluster the pixel in image space with certain relationship, these pixels can be used certain by searching The parameter space that one analytical form connects accumulates corresponding points, in the case where parameter space is no more than two-dimensional situation, this transformation There is ideal effect, calculation amount is very big in actual use for conventional Hough transform transformation, and real-time is poor, the present invention couple Hough transform localized, the improvement of randomization, is eliminated useless, duplicate data and calculating in analysis, has been reached life Produce the requirement of real-time；

Step 3: image enhancement；Image enhancement and the purpose of recovery are the quality in order to improve image, such as remove noise, are improved The clarity etc. of image enhances part interested in prominent image, strengthens image high fdrequency component, can make objects in images wheel To clean up clear, details is obvious, and noise in image influence can be reduced by strengthening low frequency component, in steel billet positioning image processing process, figure Image intensifying mainly includes removal noise, eliminates interference and steel billet imaging self-adaptive three parts, and there are a large amount of electromagnetism for industry spot The noise generated is radiated, is presented as snow random on time domain and frequency domain in the picture, since steel billet is higher than environment temperature, In Color is partially shallow, partially white in image, if white noise without filtering out, will cause steel billet and whether there is, there are the erroneous judgement of position, Common gaussian filtering handles image, removes random noise；Production scene not can avoid equipment movement, personnel's walking, steel Base such as retracts at the normal condition of production, and these situations can to cause to the analysis of the mistake of steel billet, accidentally positioning, tracking error situations such as, At this moment just state, the position that steel billet is practically in must be inferred to using existing image information combination steel billet positioning contextual information Deng, avoid mistake control signal generate；

Step 4: image segmentation；Image segmentation is to extract characteristic significant in image, significant feature There are edge, region, the angle point etc. in image, is further progress image recognition, analysis and the basis of understanding and present image One of hot spot studied in processing, although the method for having developed many edge extractings, region segmentation at present, but without one kind The effective ways of various images are generally applicable to, therefore, to the research of image segmentation also among deepening continuously, are positioned in steel billet Image processing process in, by effective image segmentation, the range that steel billet is not in can be rejected, eliminate and avoid mistake Positioning analysis can reject ineffective image data, reduce image-processing operations amount, improve system real time, and image segmentation is divided into Artificial segmentation and automatic segmentation two parts:

1.: artificial segmentation is to mark out the approximate region of steel billet appearance in camera image in advance, and region is generally quadrangle, this Kind mode is applicable in the more fixed occasion in steel billet moving region, such as slab cutting, slab conveying etc.；

2.: automatic segmentation is to be drawn by the color in region, edge etc. to image in whole image without any artificial mark Point, it is applicable in when steel billet moving region is completely filled with camera image pickup area, such as steel billet cold bed counting etc.；

Step 5: image recognition；Image recognition (classification) belongs to the scope of pattern-recognition, is that image (increases by certain pretreatments By force, restore) after, image segmentation and feature extraction are carried out, to make decisions classification, image classification is known frequently with classical mode Other method has statistical pattern classification and tactic pattern classification, in recent years Fuzzy Pattern Recognition and artificial neural network pattern classification It is also increasingly taken seriously in image recognition；

Step 6: steel billet positioning；On the basis of traditional images processing positioning slab method, finite state plane mechanism, root are introduced The characteristics of image moved in space according to steel billet takes out the Changes of each state of its spatial movement, conventional edge is known The big data quantity of other localization method calculates, and is converted to statistical analysis, substantially increases the real-time of detection method, meanwhile, having Limit state machine mechanism in, introduce self-learning function, detection method can constantly according to current steel billet image feature value learn and more New state machine characteristic value, to greatly expand the performance of steel billet localization method；

Step 7: application；Including fixed point detection application, length detection application, length+density applications, count application and fixed point+temperature Degree application, specific as follows:

1. fixed point detection application: fixed point detection application refer to by detect steel billet reach some predeterminated position after trigger specific operation Application mode, comprising:

1) pinpoint early warning: detection steel billet reaches triggering sound, light, electric signal early warning behind designated position, reaches abnormal station to steel billet Early warning is carried out, safety accident is avoided；

2) fixed point cutting: detection steel billet triggers cutting action after reaching designated position, cuts slab by designated length；

3) pinpoint roller control: control roller starts turning after detection steel billet base head reaches roller position, after base tail leaves roller position Roller stops operating, and reduces roller abrasion, avoids roller stall, mitigates labor intensity of workers；

2. length detection application: length application, which refers to synchronize the two or more predeterminated positions of steel billet when pinpointing detection, to be detected, The application mode that steel billet length derives therefrom when obtaining steel billet detection in place by predeterminated position spatial positional information, packet It includes:

1) fixed length cutting: detection steel billet length executes cutting action after reaching certain length；

2) Surface testing: billet surface oxide skin, crack equal length, area reach certain standard it is determined that current steel billet not Qualification, basic at present to use artificial sampling observation mode, low efficiency, error is high, positions these spies on steel billet by machine vision camera The indexs such as length, the area of sign, greatly promote Detection accuracy；

3. length+density applications: length+application refers to the synchronous detection of the two or more predeterminated positions of steel billet, passes through and presets position Set the application mode that steel billet length derives therefrom when spatial positional information obtains steel billet detection in place, comprising:

1) fixed to cut again: in continuous casting, when pulling rate (bloom production speed) is stablized, steel billet density tends towards stability (specific Numerical value regards steel grade difference and slightly difference), this density value is set in cut-to-length sizes machine with functions, weight=length * sectional area is (permanent It is fixed) * density, it is possible thereby to realize that the fixed of slab is cut again；

4. counting application: count application refer to steel billet quantity is counted in conjunction with steel billet location context when pinpointing detection and The application mode derived, comprising:

1) slab counts: continuous-cast blank cuts the stage, delivery phase, steel rolling are heat-fed the stage, require to the quantity of slab into Row accurately counts and generates statistical report form, and slab accurate metering core is the abnormal motion for steel billet, such as (the steel that retracts Base has left counting region completely or partially and has retracted again), in parallel (several steel billet levels are in juxtaposition), series connection (two slabs Front one is closely joined together end to end with one below) situations such as according to steel billet conveying context environmental make accurately Judgement；

2) automatic ejection: in continuous casting cold bed section, the automatic double steel when slab reaches designated position is needed (steel billet to be lifted out roller Road guarantees that next slab can be transported to same position in roller-way), after steel turning, which need to add up certain amount, (can set, software Programming count) then control billet transferring vehicle once mostly will push to cold bed with steel billet to steel billet；

5. fixed point+temperature applications: temperature applications, which refer to, utilizes machine vision camera wide spectrum on the basis of the detection of the fixed point of steel billet Especially infrared band imaging data steel billet temperature is measured and the application mode that derives, comprising:

1) temperature controls: in continuous casting and hot rolling stage, the temperature of steel billet is the important parameter in production, the absolute value of temperature, temperature The distribution of degree field all has important role to the production control of upstream and downstream, commonly uses professional temperature measuring device at this stage, and application surface is narrow, from Dynamicization degree is low, using far infrared machine vision camera, steel billet movement monitoring simultaneously, produce steel billet temperature field pattern, from Dynamic detection, analysis, early warning, greatly promote production efficiency.

2. a kind of steel billet localization method and application based on the processing of machine vision camera image according to claim 1, Be characterized in that: machine vision is fast-developing direction of artificial intelligence, and the purpose of machine vision is using machine generation Measurement and judgement are made for human eye, NI Vision Builder for Automated Inspection is will to be ingested mesh by image-forming component (being divided to two kinds of CMOS and CCD) Mark is converted into picture signal, sends dedicated image processing system to, the shape information of target subject is obtained, according to pixel distribution With the information such as brightness, color, it is transformed into digitized signal, picture system carries out various operations to these signals to extract target The characteristics of feature, and then control according to the result of differentiation the device action at scene, NI Vision Builder for Automated Inspection is improve production soft Property and the degree of automation, are difficult to the field met the requirements in some dangerous work environment for being not suitable for manual work or artificial vision It closes, machine in normal service vision substitutes artificial vision, while in high-volume industrial processes, manually visual inspection product matter Amount efficiency is low and precision is not high, and the degree of automation of production efficiency and production can be greatly improved with machine vision detection method, And machine vision is easily achieved information integration, is the basic technology for realizing computer integrated manufacturing system.

3. a kind of steel billet localization method and application based on the processing of machine vision camera image according to claim 1, Be characterized in that: the method that image converts in the step 2 is that Fourier transform, Walsh transformation, discrete cosine transform etc. are indirect Processing technique.

4. a kind of steel billet localization method and application based on the processing of machine vision camera image according to claim 1, Be characterized in that: the Hough transform in the step 2 is a kind of parameter estimation techniques using voting principle, and principle is to utilize Test problems in image space are transformed into parameter space by the point-line duality of image space and Hough parameter space.

Technical field

The invention belongs to metallurgy industry industry spot steel billet field of locating technology, in particular to a kind of to be based on machine vision phase The steel billet localization method of machine image procossing and application.

Background technique

Various steel billet measurements, location technology are widely applied in industry spot, especially in metallurgy industry, are positioned to steel billet Demand show the trend that range is increasingly wider, precision is higher and higher, existing steel billet object location techniques can substantially be sorted out To following three kinds: one, laser ranging positions, advantage: 1, it is non-contact, compared with contact type measurement deployment easy to install, 2, close-in measurement Shi Jingdu is very high.Disadvantage: 1, narrow range is disposed.Laser direction is too strong, it is therefore necessary to be mounted on and check object front end, other positions It sets and is unable to measure, greatly limit its application；2, optical device is very fragile, generally requires extremely complex shell and is prevented Shield, while cannot carry out close-in measurement to high temperature billet steel, can not play the high-precision of laser ranging, and 3, optical material is to temperature Very sensitive, measurement error caused by temperature drift is very big；Summarize: only part required precision is high and work for limit for laser ranging positioning The extremely excellent lower use of condition.Two, contact positions, and advantage: anti-interference is best；Disadvantage: 1, contact type measurement, high temp objects contact are led Hot serious, detection device is easy to damage, is not available substantially, 2, steel billet belong to big quality testing object, inertia is big, and detection device is easy It is subjected to displacement by steel billet impact, measurement error is big；Summarize: contact is located in have been eliminated completely at this stage.Three, image procossing is fixed Position, advantage: 1, non-contact, deployment easy to install can even be spliced with polyphaser parallel connection for big detection zone and be detected, and 2, by temperature Influence it is small, by adjust camera parameter be suitable for extensive scene；Disadvantage: 1, it is limited to the limitation of ccd sensor imaging unit size It is limited with lens imaging principle, precision is difficult to be promoted after reaching 0.05 millimeter of rank；2, it is easy dry by ambient, pedestrian etc. It disturbs, but can be by improving image processing algorithm, the measures such as specification scene solve；In conclusion in metallurgy industry steel billet millimeter Under grade or submillimeter level positioning application scenarios, image procossing positioning is optimal selection.

In image procossing positioning method, existing frequently-used following two camera apparatus；One, analogue camera, analogue camera are Conventional coaxial camera, acquisition link include analogue camera and image pick-up card two parts, and advantage: real-time is high；Disadvantage: 1, precision Low, analogue camera is limited to coaxial cable bandwidth, and highest resolution is no more than 800x600, and high-precision occasion is not available, 2, away from From close.More than 50 meters signals of general work distance will decay to unavailable, 3, easily interference to coaxial cable in industrial settings, together Shaft cable transmits analog video signal, is easy by live electromagnetic interference.

Two, video camera, advantage: 1, precision is high, with high resolution CMOS technology maturation, 2K even 4K (12,000,000 pictures Element) network video camera continues to bring out.Compared to the promotion that analogue camera image quality has matter, 2, distance it is remote.It can be protected using cable Stablize transmission in 100 meters of card, farther to add relaying, 3, strong antijamming capability, Digital Transmission avoid analog signal transmission easily quilt The problem of interference；Disadvantage: 1, real-time is poor, compared to analogue camera by the mode of PC driving image pick-up card acquisition image, video Camera just encodes image in phase generator terminal, and the image after compression (generally H.264) is transferred to figure by Ethernet As on processing PC, transmission delay has reached the second grade, this is disabled in the conjunction of real-time high field, 2, information loss, video class Camera is low as far as possible in order to guarantee bandwidth utilization rate, generally all using H.264, H.265 etc. Video codings mode compiles image Code causes the positioning key message such as image border, angle point to be lost, and these exactly guarantee number necessary to steel billet positioning accuracy According to.

In the method for existing image procossing detection slab position, mostly there are following several respects: 1, stateless Memory, error rate are high.When steel billet is subjected to displacement in space, well-regulated state change can occur for moving image feature, this A little status changes are continuously, to determine.Existing detection method fails in consideration of it, 2, algorithm is complicated, and real-time is poor.It is existing There is algorithm to be all based on edge detection greatly, and convolution algorithm of the edge detection based on image, operation times and time and image resolution Rate it is square directly proportional.Present industry spot is higher and higher to the requirement of slab position detection accuracy, constantly raises video camera resolution Rate, based entirely on limb recognition algorithm real-time worse and worse, 3, learning disabilities, scalability is poor.Traditional images processing is calculated Method core concept is exactly that slab position is analyzed and determined by preset characteristics of image, when ambient enviroment changes or slab When shape changes, corresponding characteristics of image can also change, and using preset characteristic value come when analyzing, Jiu Huifa Raw erroneous judgement, extended capability are very poor；For this purpose, the present invention proposes a kind of steel billet localization method based on the processing of machine vision camera image And application.

Summary of the invention

Of the existing technology in order to solve the problems, such as, the present invention provides a kind of based on the processing of machine vision camera image Steel billet localization method and application should be made based on the steel billet localization method and application design that machine vision camera image is handled rationally Image Acquisition is carried out with machine vision camera, optimizes steel billet location algorithm, the using face for deep-cutting steel billet positioning is promoted, realizes height Precision, hard real-time, New steel billet positioning system easy for installation, applicable surface is extensive, is suitble to large-scale popularization.

To achieve the goals above, the present invention is to realize by the following technical solutions: one kind is based on machine vision phase The steel billet localization method of machine image procossing and application, comprising the following steps:

Step 1: Image Acquisition；Image is acquired using machine vision camera, machine vision camera has the following characteristics that

1.: the stable and reliable for performance of machine vision camera is easily installed, camera structure it is compact it is solid be hardly damaged, continuous work Make that the time is long, can be used continuously for a long time under poor environment；

2.: the aperture time of machine vision camera is very short, can capture swiftly passing object, and common camera is generally every Second most 25 frames, machine vision camera can then accomplish thousands of frame/seconds, thus have to movement steel billet positioning very high-precision Degree；

3.: the imaging sensor of machine vision camera be progressive scan, and the imaging sensor of common camera be every Row scanning, the imaging sensor production technology of progressive scan is more complicated, but imaging effect is more preferable, especially calculates image procossing The information such as sensitive edge, angle point more comprehensively, more can accurately orient the location information, surface information and direction of steel billet in method Information；

4.: the output of machine vision camera is uncorrected data, spectral region also tend to it is wider, from far infrared, near-infrared, Visible light, ultraviolet etc. have, and compare the image processing algorithm for being appropriate for high quality, and the picture of other cameras shooting, light Spectral limit is only suitable for human eye vision, and after video compress, picture quality is poor, is unfavorable for analysis processing, therefore machine vision Camera can be widely applied to cold base (room temperature steel billet), hot base (steel billet that the continuous casting stage has just produced), hot material (steel rolling rank Steel billet in the section operation of rolling) etc. in each stage steel billet position fixing process；

Step 2: image transformation；It since pattern matrix is very big, is directly handled in the spatial domain, is related to calculation amount very Greatly, therefore, often the processing of spatial domain is converted into transform-domain processing, can not only be reduced using the method for various images transformation Calculation amount, and can get more effectively processing, in steel billet positioning image processing process, by Hough transform, can directly it exist The pixel data for meeting steel billet shape information is extracted in picture number, to quickly position steel billet possible position, principle is logical It crosses and carries out simple cumulative statistics in parameter space, the method detection of accumulator peak value is then found in Hough parameter space Straight line, the essence of Hough transform are to cluster the pixel in image space with certain relationship, and searching can be these pixels Corresponding points are accumulated with the parameter space that a certain analytical form connects, it is this in the case where parameter space is no more than two-dimensional situation Transformation has ideal effect, and calculation amount is very big in actual use for conventional Hough transform transformation, and real-time is poor, the present invention It is localized to Hough transform, the improvement of randomization, eliminates useless, duplicate data and calculating in analysis, reach Produce the requirement of real-time；

Step 3: image enhancement；Image enhancement and the purpose of recovery are the quality in order to improve image, such as remove noise, The clarity etc. of image is improved, enhances part interested in prominent image, strengthens image high fdrequency component, object in image can be made Body is clear-cut, and details is obvious, and noise in image influence can be reduced by strengthening low frequency component, positions image processing process in steel billet In, image enhancement mainly includes removal noise, eliminates interference and steel billet imaging self-adaptive three parts, and industry spot exists a large amount of The noise that electromagnetic radiation generates, is presented as snow random on time domain and frequency domain, since steel billet compares environment temperature in the picture Height, color is partially shallow, partially white in the picture, if white noise without filtering out, will cause, steel billet whether there is, there are positions Erroneous judgement, common gaussian filtering handle image, remove random noise；Production scene not can avoid equipment movement, Ren Yuanhang It walks, the normal condition of production such as steel billet retracts, and these situations can be to causing to the analysis of the mistake of steel billet, accidentally positioning, tracking error etc. At this moment situation just must be inferred to the shape that steel billet is practically in using existing image information combination steel billet positioning contextual information State, position etc. avoid the control signal of mistake from generating；

Step 4: image segmentation；Image segmentation is to extract characteristic significant in image, significant Feature has edge, region, angle point in image etc., is further progress image recognition, analysis and the basis of understanding, and current One of hot spot studied in image procossing, although the method for having developed many edge extractings, region segmentation at present, but without A kind of effective ways being generally applicable to various images, therefore, to the research of image segmentation also among deepening continuously, in steel billet In the image processing process of positioning, by effective image segmentation, the range that steel billet is not in can be rejected, elimination avoids mistake Positioning analysis accidentally can reject ineffective image data, reduce image-processing operations amount, improve system real time, image segmentation It is divided into artificial segmentation and automatic segmentation two parts:

1.: artificial segmentation is to mark out the approximate region of steel billet appearance in camera image in advance, and region is generally four sides Shape, this mode are applicable in the more fixed occasion in steel billet moving region, such as slab cutting, slab conveying etc.；

2.: automatic segmentation is in whole image without any artificial mark, by the color in region, edge etc. to image into Row divides, and is applicable in when steel billet moving region is completely filled with camera image pickup area, such as steel billet cold bed counting etc.；

Step 5: image recognition；Image recognition (classification) belongs to the scope of pattern-recognition, is image by certain pretreatments After (enhancing is restored), image segmentation and feature extraction are carried out, to make decisions classification, image classification is frequently with classical mould Formula recognition methods has statistical pattern classification and tactic pattern classification, in recent years Fuzzy Pattern Recognition and artificial neural network mode It is sorted in image recognition and is also increasingly taken seriously；

Step 6: steel billet positioning；On the basis of traditional images processing positioning slab method, finite state machine machine is introduced System, the characteristics of image moved in space according to steel billet takes out the Changes of each state of its spatial movement, by traditional side The big data quantity of edge recognition positioning method calculates, and is converted to statistical analysis, substantially increases the real-time of detection method, meanwhile, In finite state plane mechanism, self-learning function is introduced, detection method can constantly learn according to current steel billet image feature value And state machine feature value is updated, to greatly expand the performance of steel billet localization method；

Step 7: application；Including fixed point detection application, length detection application, length+density applications, counts and apply and determine Point+temperature applications, specific as follows:

1. fixed point detection application: fixed point detection application refer to by detect steel billet reach trigger after some predeterminated position it is specific The application mode of operation, comprising:

1) pinpoint early warning: detection steel billet reaches triggering sound, light, electric signal early warning behind designated position, reaches to steel billet abnormal Station carries out early warning, avoids safety accident；

2) fixed point cutting: detection steel billet triggers cutting action after reaching designated position, cuts slab by designated length；

3) pinpoint roller control: control roller starts turning after detection steel billet base head reaches roller position, and base tail leaves roller position It postpones roller to stop operating, reduces roller abrasion, avoid roller stall, mitigate labor intensity of workers；

2. length detection application: length application refers to examine synchronous to the two or more predeterminated positions of steel billet when pinpointing detection It surveys, the application mode that steel billet length derives therefrom when obtaining steel billet detection in place by predeterminated position spatial positional information, Include:

1) fixed length cutting: detection steel billet length executes cutting action after reaching certain length；

2) Surface testing: billet surface oxide skin, crack equal length, area reach certain standard it is determined that current steel Base is unqualified, and basic at present using artificial sampling observation mode, low efficiency, error is high, positions on steel billet this by machine vision camera The indexs such as length, the area of a little features, greatly promote Detection accuracy；

3. length+density applications: length+application refers to the synchronous detection of the two or more predeterminated positions of steel billet, by pre- The application mode that steel billet length derives therefrom when if locational space location information obtains steel billet detection in place, comprising:

1) fixed to cut again: in continuous casting, when pulling rate (bloom production speed) is stablized, steel billet density tends towards stability (specific value regards steel grade difference and slightly difference), this density value is set in cut-to-length sizes machine with functions, weight=length * sectional area (constant) * density, it is possible thereby to realize that the fixed of slab is cut again；

4. counting application: counting application refers to combines steel billet location context to unite steel billet quantity when pinpointing detection The application mode counted and derived, comprising:

1) slab counts: cutting stage, delivery phase, steel rolling hot sending stage in continuous-cast blank, requires the number to slab Amount is accurately counted and is generated statistical report form, and slab accurate metering core is the abnormal motion for steel billet, such as returns Move back (steel billet has left counting region completely or partially and retracted again), (several steel billet levels are in juxtaposition) in parallel, series connection (two One is closely joined together end to end with one below before slab) situations such as according to steel billet conveying context environmental make standard True judgement；

2) automatic ejection: in continuous casting cold bed section, the automatic double steel when slab reaches designated position is needed (to be lifted out steel billet Roller-way, guarantee next slab same position can be transported in roller-way), after steel turning need to add up certain amount (can set, it is soft Part programming count) then control billet transferring vehicle once mostly will push to cold bed with steel billet to steel billet；

5. fixed point+temperature applications: temperature applications, which refer to, utilizes machine vision phase machine width on the basis of the detection of the fixed point of steel billet The application mode that spectrum especially infrared band imaging data derives steel billet temperature measurement, comprising:

1) temperature control: in continuous casting and hot rolling stage, the temperature of steel billet be production in important parameter, temperature it is absolute The distribution of value, temperature field all has important role to the production control of upstream and downstream, commonly uses professional temperature measuring device, application surface at this stage Narrow, the degree of automation is low, using far infrared machine vision camera, steel billet movement monitoring simultaneously, produce steel billet temperature field distribution Figure, automatic detection, analysis, early warning, greatly promotes production efficiency.

As a kind of preferred embodiment of the invention, machine vision is fast-developing direction of artificial intelligence, machine The purpose of device vision is to replace human eye using machine to make measurement and judgement, and NI Vision Builder for Automated Inspection (is divided by image-forming component CMOS and two kinds of CCD) it target will be ingested is converted into picture signal, it sends dedicated image processing system to, obtains mesh shot Target shape information is transformed into digitized signal, picture system is to these letters according to the information such as pixel distribution and brightness, color Number various operations are carried out to extract clarification of objective, and then the device action at scene is controlled according to the result of differentiation, machine view The characteristics of feel system is the flexibility and the degree of automation for improving production, in some dangerous work environment for being not suitable for manual work Or artificial vision is difficult to the occasion met the requirements, machine in normal service vision substitutes artificial vision, while in high-volume industrial production In the process, manually visual inspection product quality low efficiency and precision is not high, can be greatly improved with machine vision detection method The degree of automation of production efficiency and production, and machine vision is easily achieved information integration, is to realize computer integrated manufacturing system Basic technology.

As a kind of preferred embodiment of the invention, the method for image transformation is Fourier transform, Wall in the step 2 The indirect processing techniques such as assorted transformation, discrete cosine transform.

As a kind of preferred embodiment of the invention, the Hough transform in the step 2 is a kind of using voting principle Parameter estimation techniques, principle is the point-line duality using image space and Hough parameter space, in image space Test problems are transformed into parameter space.

Beneficial effects of the present invention:

1, this based on machine vision camera image processing steel billet localization method and apply in the position fixing process to steel billet New detection device is used, new image processing algorithm greatly simplifies in-site installation, using the model of framing steel billet It encloses, significantly improves the precision and real-time of the positioning of industry spot steel billet, and on this basis, deeply excavate steel billet positioning A series of application prospect, the industrial application based on proposing and realizing positioning countings by steel billet, reduces in actual use Worker's amount of labour, reduces equipment attrition, improves detection accuracy, has created economic benefit.

2, this based on machine vision camera image processing steel billet localization method and using machine vision camera carry out Image Acquisition optimizes steel billet location algorithm, promotes the using face for deep-cutting steel billet positioning, realizes high-precision, hard real-time, installation Convenient New steel billet positioning system, applicable surface is extensive, is suitble to large-scale popularization.

3, this based on machine vision camera image processing steel billet localization method and apply traditional images processing positioning casting On the basis of blank technology, finite state plane mechanism is introduced, the characteristics of image moved in space according to steel billet takes out its space The big data quantity of conventional edge recognition positioning method is calculated, is converted to statistical analysis, greatly by the Changes for moving each state The real-time of detection method is improved greatly, meanwhile, in finite state plane mechanism, self-learning function is introduced, detection method can not Disconnected learns according to current steel billet image feature value and updates state machine feature value, to greatly expand steel billet localization method Performance.

4, this steel billet localization method and application based on the processing of machine vision camera image are acquired using machine vision camera The drawbacks of live image carries out steel billet positioning, can be to avoid existing various positioning devices, precision is higher, and real-time is stronger, deployment Convenient, wide application, machine vision camera compares network cameras, does not carry out any compression to real image, directly transports original Image data, transmission time-consuming substantially reduce, and have reached the available program of industrial real-time, and machine vision camera does not carry out image The information such as positioning edge, angle point used are not lost in compression, and positioning accuracy is high, while also having all advantages of video camera.

Detailed description of the invention

Fig. 1 is a kind of steel billet localization method based on the processing of machine vision camera image and the step process signal of application Figure；

Specific embodiment

To be easy to understand the technical means, the creative features, the aims and the efficiencies achieved by the present invention, below with reference to Specific embodiment, the present invention is further explained.

Referring to Fig. 1, the present invention provides a kind of technical solution: a kind of steel billet based on the processing of machine vision camera image is fixed Position method and application, comprising the following steps:

Step 1: Image Acquisition；Image is acquired using machine vision camera, machine vision camera has the following characteristics that

1.: the stable and reliable for performance of machine vision camera is easily installed, camera structure it is compact it is solid be hardly damaged, continuous work Make that the time is long, can be used continuously for a long time under poor environment；

2.: the aperture time of machine vision camera is very short, can capture swiftly passing object, and common camera is generally every Second most 25 frames, machine vision camera can then accomplish thousands of frame/seconds, thus have to movement steel billet positioning very high-precision Degree；

3.: the imaging sensor of machine vision camera be progressive scan, and the imaging sensor of common camera be every Row scanning, the imaging sensor production technology of progressive scan is more complicated, but imaging effect is more preferable, especially calculates image procossing The information such as sensitive edge, angle point more comprehensively, more can accurately orient the location information, surface information and direction of steel billet in method Information；

4.: the output of machine vision camera is uncorrected data, spectral region also tend to it is wider, from far infrared, near-infrared, Visible light, ultraviolet etc. have, and compare the image processing algorithm for being appropriate for high quality, and the picture of other cameras shooting, light Spectral limit is only suitable for human eye vision, and after video compress, picture quality is poor, is unfavorable for analysis processing, therefore machine vision Camera can be widely applied to cold base (room temperature steel billet), hot base (steel billet that the continuous casting stage has just produced), hot material (steel rolling rank Steel billet in the section operation of rolling) etc. in each stage steel billet position fixing process；

Step 2: image transformation；It since pattern matrix is very big, is directly handled in the spatial domain, is related to calculation amount very Greatly, therefore, often the processing of spatial domain is converted into transform-domain processing, can not only be reduced using the method for various images transformation Calculation amount, and can get more effectively processing, in steel billet positioning image processing process, by Hough transform, can directly it exist The pixel data for meeting steel billet shape information is extracted in picture number, to quickly position steel billet possible position, principle is logical It crosses and carries out simple cumulative statistics in parameter space, the method detection of accumulator peak value is then found in Hough parameter space Straight line, the essence of Hough transform are to cluster the pixel in image space with certain relationship, and searching can be these pixels Corresponding points are accumulated with the parameter space that a certain analytical form connects, it is this in the case where parameter space is no more than two-dimensional situation Transformation has ideal effect, and calculation amount is very big in actual use for conventional Hough transform transformation, and real-time is poor, the present invention It is localized to Hough transform, the improvement of randomization, eliminates useless, duplicate data and calculating in analysis, reach Produce the requirement of real-time；

Step 3: image enhancement；Image enhancement and the purpose of recovery are the quality in order to improve image, such as remove noise, The clarity etc. of image is improved, enhances part interested in prominent image, strengthens image high fdrequency component, object in image can be made Body is clear-cut, and details is obvious, and noise in image influence can be reduced by strengthening low frequency component, positions image processing process in steel billet In, image enhancement mainly includes removal noise, eliminates interference and steel billet imaging self-adaptive three parts, and industry spot exists a large amount of The noise that electromagnetic radiation generates, is presented as snow random on time domain and frequency domain, since steel billet compares environment temperature in the picture Height, color is partially shallow, partially white in the picture, if white noise without filtering out, will cause, steel billet whether there is, there are positions Erroneous judgement, common gaussian filtering handle image, remove random noise；Production scene not can avoid equipment movement, Ren Yuanhang It walks, the normal condition of production such as steel billet retracts, and these situations can be to causing to the analysis of the mistake of steel billet, accidentally positioning, tracking error etc. At this moment situation just must be inferred to the shape that steel billet is practically in using existing image information combination steel billet positioning contextual information State, position etc. avoid the control signal of mistake from generating；

Step 4: image segmentation；Image segmentation is to extract characteristic significant in image, significant Feature has edge, region, angle point in image etc., is further progress image recognition, analysis and the basis of understanding, and current One of hot spot studied in image procossing, although the method for having developed many edge extractings, region segmentation at present, but without A kind of effective ways being generally applicable to various images, therefore, to the research of image segmentation also among deepening continuously, in steel billet In the image processing process of positioning, by effective image segmentation, the range that steel billet is not in can be rejected, elimination avoids mistake Positioning analysis accidentally can reject ineffective image data, reduce image-processing operations amount, improve system real time, image segmentation It is divided into artificial segmentation and automatic segmentation two parts:

1.: artificial segmentation is to mark out the approximate region of steel billet appearance in camera image in advance, and region is generally four sides Shape, this mode are applicable in the more fixed occasion in steel billet moving region, such as slab cutting, slab conveying etc.；

2.: automatic segmentation is in whole image without any artificial mark, by the color in region, edge etc. to image into Row divides, and is applicable in when steel billet moving region is completely filled with camera image pickup area, such as steel billet cold bed counting etc.；

Step 5: image recognition；Image recognition (classification) belongs to the scope of pattern-recognition, is image by certain pretreatments After (enhancing is restored), image segmentation and feature extraction are carried out, to make decisions classification, image classification is frequently with classical mould Formula recognition methods has statistical pattern classification and tactic pattern classification, in recent years Fuzzy Pattern Recognition and artificial neural network mode It is sorted in image recognition and is also increasingly taken seriously；

Step 6: steel billet positioning；On the basis of traditional images processing positioning slab method, finite state machine machine is introduced System, the characteristics of image moved in space according to steel billet takes out the Changes of each state of its spatial movement, by traditional side The big data quantity of edge recognition positioning method calculates, and is converted to statistical analysis, substantially increases the real-time of detection method, meanwhile, In finite state plane mechanism, self-learning function is introduced, detection method can constantly learn according to current steel billet image feature value And state machine feature value is updated, to greatly expand the performance of steel billet localization method；

Step 7: application；Including fixed point detection application, length detection application, length+density applications, counts and apply and determine Point+temperature applications, specific as follows:

1. fixed point detection application: fixed point detection application refer to by detect steel billet reach trigger after some predeterminated position it is specific The application mode of operation, comprising:

1) pinpoint early warning: detection steel billet reaches triggering sound, light, electric signal early warning behind designated position, reaches to steel billet abnormal Station carries out early warning, avoids safety accident；

2) fixed point cutting: detection steel billet triggers cutting action after reaching designated position, cuts slab by designated length；

3) pinpoint roller control: control roller starts turning after detection steel billet base head reaches roller position, and base tail leaves roller position It postpones roller to stop operating, reduces roller abrasion, avoid roller stall, mitigate labor intensity of workers；

2. length detection application: length application refers to examine synchronous to the two or more predeterminated positions of steel billet when pinpointing detection It surveys, the application mode that steel billet length derives therefrom when obtaining steel billet detection in place by predeterminated position spatial positional information, Include:

1) fixed length cutting: detection steel billet length executes cutting action after reaching certain length；

2) Surface testing: billet surface oxide skin, crack equal length, area reach certain standard it is determined that current steel Base is unqualified, and basic at present using artificial sampling observation mode, low efficiency, error is high, positions on steel billet this by machine vision camera The indexs such as length, the area of a little features, greatly promote Detection accuracy；

3. length+density applications: length+application refers to the synchronous detection of the two or more predeterminated positions of steel billet, by pre- The application mode that steel billet length derives therefrom when if locational space location information obtains steel billet detection in place, comprising:

1) fixed to cut again: in continuous casting, when pulling rate (bloom production speed) is stablized, steel billet density tends towards stability (specific value regards steel grade difference and slightly difference), this density value is set in cut-to-length sizes machine with functions, weight=length * sectional area (constant) * density, it is possible thereby to realize that the fixed of slab is cut again；

4. counting application: counting application refers to combines steel billet location context to unite steel billet quantity when pinpointing detection The application mode counted and derived, comprising:

1) slab counts: cutting stage, delivery phase, steel rolling hot sending stage in continuous-cast blank, requires the number to slab Amount is accurately counted and is generated statistical report form, and slab accurate metering core is the abnormal motion for steel billet, such as returns Move back (steel billet has left counting region completely or partially and retracted again), (several steel billet levels are in juxtaposition) in parallel, series connection (two One is closely joined together end to end with one below before slab) situations such as according to steel billet conveying context environmental make standard True judgement；

2) automatic ejection: in continuous casting cold bed section, the automatic double steel when slab reaches designated position is needed (to be lifted out steel billet Roller-way, guarantee next slab same position can be transported in roller-way), after steel turning need to add up certain amount (can set, it is soft Part programming count) then control billet transferring vehicle once mostly will push to cold bed with steel billet to steel billet；

5. fixed point+temperature applications: temperature applications, which refer to, utilizes machine vision phase machine width on the basis of the detection of the fixed point of steel billet The application mode that spectrum especially infrared band imaging data derives steel billet temperature measurement, comprising:

1) temperature control: in continuous casting and hot rolling stage, the temperature of steel billet be production in important parameter, temperature it is absolute The distribution of value, temperature field all has important role to the production control of upstream and downstream, commonly uses professional temperature measuring device, application surface at this stage Narrow, the degree of automation is low, using far infrared machine vision camera, steel billet movement monitoring simultaneously, produce steel billet temperature field distribution Figure, automatic detection, analysis, early warning, greatly promotes production efficiency.

As a kind of preferred embodiment of the invention, machine vision is fast-developing direction of artificial intelligence, machine The purpose of device vision is to replace human eye using machine to make measurement and judgement, and NI Vision Builder for Automated Inspection (is divided by image-forming component CMOS and two kinds of CCD) it target will be ingested is converted into picture signal, it sends dedicated image processing system to, obtains mesh shot Target shape information is transformed into digitized signal, picture system is to these letters according to the information such as pixel distribution and brightness, color Number various operations are carried out to extract clarification of objective, and then the device action at scene is controlled according to the result of differentiation, machine view The characteristics of feel system is the flexibility and the degree of automation for improving production, in some dangerous work environment for being not suitable for manual work Or artificial vision is difficult to the occasion met the requirements, machine in normal service vision substitutes artificial vision, while in high-volume industrial production In the process, manually visual inspection product quality low efficiency and precision is not high, can be greatly improved with machine vision detection method The degree of automation of production efficiency and production, and machine vision is easily achieved information integration, is to realize computer integrated manufacturing system Basic technology.

As a kind of preferred embodiment of the invention, the method for image transformation is Fourier transform, Wall in the step 2 The indirect processing techniques such as assorted transformation, discrete cosine transform.

As a kind of preferred embodiment of the invention, the Hough transform in the step 2 is a kind of using voting principle Parameter estimation techniques, principle is the point-line duality using image space and Hough parameter space, in image space Test problems are transformed into parameter space.

The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention, for this field skill For art personnel, it is clear that invention is not limited to the details of the above exemplary embodiments, and without departing substantially from spirit of the invention or In the case where essential characteristic, the present invention can be realized in other specific forms.Therefore, in all respects, should all incite somebody to action Embodiment regards exemplary as, and is non-limiting, the scope of the present invention by appended claims rather than on state Bright restriction, it is intended that including all changes that fall within the meaning and scope of the equivalent elements of the claims in the present invention It is interior.Any reference signs in the claims should not be construed as limiting the involved claims.

In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art The other embodiments being understood that.