阅读说明：本技术 一种基于计算鬼成像的三维目标重构方法 (A kind of objective reconstructing method based on calculating ghost imaging ) 是由 张闻文 张磊 林子韬 何睿清 钱燕 何伟基 陈钱 顾国华 于 2018-05-25 设计创作，主要内容包括：本发明公开了一种基于计算鬼成像的三维目标重构的方法,首先对参考平面进行标定,根据图像中每个像素的灰度,计算每个像素所代表的参考平面上对应区域中心处光源方向和桶探测器方向之间的夹角；将三维目标放置在参考平面上,根据图像中每个像素的灰度,计算每个像素所代表的目标上对应区域中心处光源方向和桶探测器方向之间的夹角；最后计算参考平面和目标对应区域处角度的差值,得到目标距离参考平面的深度,从而重构出目标的三维形态。本发明增加一个参考平面,将桶探测器的数目减少至一个,降低了空间成本和系统成本；将三维目标重构技术与鬼成像技术相结合,有助于鬼成像在实际应用中的推广,特别是在工业零件和生物样本的检测中。(The invention discloses a kind of methods based on the objective reconstruct for calculating ghost imaging, reference planes are demarcated first, according to the gray scale of pixel each in image, the angle in reference planes representated by each pixel at the center of corresponding region between light source direction and bucket detector direction is calculated；Objective is placed on the reference plane, according to the gray scale of pixel each in image, calculates the angle in target representated by each pixel at the center of corresponding region between light source direction and bucket detector direction；The difference for finally calculating angle at reference planes and target corresponding region, obtains the depth of target range reference planes, to reconstruct the three-dimensional configuration of target.The present invention increases a reference planes, and the number of bucket detector is reduced to one, space cost and system cost are reduced；Objective reconfiguration technique is combined with terrible imaging technique, facilitates the popularization of ghost imaging in practical applications, especially in the detection of industrial part and biological sample.)

1. a kind of method based on the objective reconstruct for calculating ghost imaging, it is characterised in that: marked first to reference planes It is fixed, i.e., reference planes are placed within the areas imaging of system, reference planes are carried out to calculate terrible imaging, utilize Lambert cosine Application of the reflection law in terrible imaging system calculates each pixel institute's generation according to the gray scale Gr (x, y) of pixel each in image Angle α (x, y) in the reference planes of table at the center of corresponding region between light source direction and bucket detector direction, establishes angle value Mapping relations between gray value of image；By objective place on the reference plane, using terrible imaging system to target into Row imaging, using application of the Lambert cosine reflection law in terrible imaging system, according to the gray scale Gt of pixel each in image (x, Y), the angle β in target representated by each pixel at the center of corresponding region between light source direction and bucket detector direction is calculated (x, y) establishes the mapping relations between angle value and gray value of image；Finally calculate angle at reference planes and target corresponding region The difference θ (x, y) of degree obtains the depth delta d (x, y) of target range reference planes, to reconstruct the three-dimensional configuration of target.

2. according to the method described in claim 1, it is characterized by: demarcated to terrible imaging system is calculated, i.e., by projector, Computer, bucket detector and reference planes are placed according to following position, at the optical center of projector and the center of reference planes It can be vertically projected in reference planes in the two-value projection for going out, projector projects in same optical axis, bucket detector is placed on partially Position from projector guarantees that the central optical axis of bucket detector passes through the center of reference planes, and relative position parameter is it is known that utilize The two-value projected image for meeting hadamard matrix distribution carries out reference planes to calculate terrible imaging, and reference planes are by calculating ghost imaging Gr (x, y) is imaged to obtain in formula calculating.

3. according to the method described in claim 2, it is characterized in that imaging process is as follows: computer pre-generates probability distribution symbol It, is projected at reference planes, hadamard matrix by the two-value projected image for closing hadamard matrix distribution frame by frame using projector Order be set as M², successively choose every a line and lay equal stress on and be made into the matrix of M*M, form M altogether²A orderly matrix；

For each of these matrix, therein -1 is set to become 0 first, using the matrix after change to the number in projector Micro mirror array, which is modulated, generates two-value projection, projects in target, bucket detector receives total intensity value；

1 in original matrix is become 0, -1 again and becomes 1, using this matrix after changing to the digital micromirror array in projector It is modulated and generates two-value projection, project in target, bucket detector receives total intensity value, and intensity value twice subtracts each other, knot Fruit is exactly the total intensity value received on a frame two-value projection to target, completes the one-shot measurement to target；

Identical operation is all carried out to the two-value projection that each frame hadamard matrix is formed, to the last frame two-value projection is completed Operation；Reference planes, which are received, using bucket detector projects light intensity；

Reference planes calculate imaging by calculating terrible imaging formula, that is,

Gr (x, y)=< I_r ⁽ⁱ⁾·P⁽ⁱ) (x, y)>-<I_r ⁽ⁱ⁾><P⁽ⁱ) (x, y) > (1)

X, y are plane coordinates, and Gr (x, y) is to calculate ghost imaging to reference planes calculating imaging as a result, I_r ⁽ⁱ⁾To calculate ghost imaging The value that bucket detector i-th detects in system, P⁽ⁱ⁾(x, y) is the Hadamard generated in i-th computer in terrible imaging system Matrix two-value projected image,<...>symbol is to average, with<I_r ⁽ⁱ⁾·P⁽ⁱ⁾(x, y)>for,<I_r ⁽ⁱ⁾>and<P⁽ⁱ⁾(x, y) > meter Calculation method is identical；

N is that ghost imaging is associated with calculation times in formula (2).

4. according to the method described in claim 1, it is characterized by: calculating each picture according to the gray scale of pixel each in image Angle α (x, y) in reference planes representated by element at the center of corresponding region between light source direction and bucket detector direction, that is,

α (x, y)=arccos (Gr (x, y)) (3)

Gr (x, y) is that the calculating ghost imaging of reference planes calculates the result of imaging in formula (3).

5. according to the method described in claim 1, it is characterized by: the objective for carrying out three-dimensionalreconstruction will be needed to be placed on ginseng It examines in plane, the plane of x-axis and y-axis composition is reference planes, utilizes the same group of Kazakhstan that reference planes calculate with ghost imaging Up to Ma matrix two-value projected image, to needing the objective for carrying out three-dimensionalreconstruction to carry out calculating terrible imaging, by calculating ghost imaging Formula calculates imaging results Gt (x, y), i.e.,

Gt (x, y)=< I_t ⁽ⁱ⁾·P⁽ⁱ⁾(x, y)>-<I_t ⁽ⁱ⁾><P⁽ⁱ⁾(x, y) > (4)

Gt (x, y) is to calculate ghost imaging to objective calculating imaging as a result, I in formula (4)_t ⁽ⁱ⁾To calculate in terrible imaging system The value that bucket detector i-th detects.

6. according to the method described in claim 1, it is characterized by: calculating each picture according to the gray scale of pixel each in image Angle β (x, y) on objective representated by element at the center of corresponding region between light source direction and bucket detector direction, i.e.,

β (x, y)=arccos (Gt (x, y)) (5).

7. according to the method described in claim 1, obtaining it is characterized in that will be made by angle value α (x, y) and angle value β (x, y) poor To difference θ (x, y), i.e.,

θ (x, y)=α (x, y)-β (x, y) (6)

θ (x, y) is the result that two angles make difference in formula (5).

8. according to the method described in claim 1, it is characterized in that solving each regional center of objective to reference planes pair The depth in region is answered, distance d (x, y) and difference θ (x, y) is accumulated, i.e.,

Δ d (x, y)=d (x, y) * θ (x, y) (7)

Δ d (x, y) is the depth that reference planes respective cell domain is arrived in objective each zonule, i.e., required by objective Form.

Technical field

It is especially a kind of based on the objective reconstruct side for calculating ghost imaging the invention belongs to technical field of image processing Method.

Background technique

In traditional ghost imaging, light beam is divided into two beams by beam splitter, a branch of to be known as after signal light, with interacting goals by one The bucket detector of a no spatial resolving power receives；Another beam is known as reference light, phase does not occur with target in communication process Interaction, but directly received by a detector array with spatial resolving power.The letter that two detectors are received Number it is associated operation, it will be able to clearly recover the image of target.

In recent years, because of its unique imaging mode, potential application value is more and more widely paid close attention to for ghost imaging. Terrible imaging technique is more applied at present in fields such as remotely sensed images, and in imaging detection sensitivity and cloud and mist, misty rain etc. is disliked Under the conditions of bad, the superiority compared with existing remotely sensed image technology is shown.The three-dimensional configuration for restoring target is imaged by ghost It is a popular research contents.

Calculate terrible imaging technique and use unique monochromatic light road imaging technique, imaging optical path be changed to by spatial light modulator or Digital micromirror array is modulated light field, and the light field record in reference path is changed to be calculated by computer.Calculate ghost imaging letter The optical path of terrible imaging technique is changed, so that the functionization of terrible imaging technique becomes possibility.

Shape-from-shading (SFS) is a kind of side that the 3D shape widely paid close attention to by computer vision community is restored Method.This method to the effect that, radiation direction, body surface reflection model known under the conditions of, utilize single image The light and shade of gray scale changes the shape information implied, the surface three dimension form of reconstruction of objects in the middle.For real image, table Face brightness of image receives many factors, such as light source, body surface material properties and shape and camera position and parameter etc. Influence, in order to simplify problem, traditional SFS method has carried out following hypothesis: (1) light source be far point light source；(2) mould is reflected Type is lambert's camber reflection model；(3) imaging geometry is rectangular projection.In 2013, B.Q Sun propose based on SFS Objective reconfiguration system (3D Computational Imaging with Single-Pixel Detectors) in, be System at least needs the form using four bucket detectors in different position capture targets, to derive the surface of target from shade Gradient simultaneously reconstructs objective.But, entire imaging system more demanding for the synchronism of system using four bucket detectors Complex, space cost is larger；The calibrated error of bucket detector spatial position can influence final Three-dimensional Gravity to a certain extent It is configured the precision of state, the quantity of bucket detector is more, and bring error undoubtedly also can be bigger.

Summary of the invention

The purpose of the present invention is to provide a kind of methods based on the objective reconstruct for calculating ghost imaging, avoid by more A barrel of detector bring device error, while the algorithm of objective reconstruct is simplified, reduce the time in restructuring procedure Cost.

Realize the technical solution of the object of the invention: a method of it is reconstructed based on the objective for calculating ghost imaging, Reference planes are demarcated first, i.e., reference planes are placed within the areas imaging of system, reference planes are counted Ghost imaging is calculated, using application of the Lambert cosine reflection law in terrible imaging system, according to the gray scale Gr of pixel each in image (x, y) is calculated in reference planes representated by each pixel at the center of corresponding region between light source direction and bucket detector direction Angle α (x, y), establish the mapping relations between angle value and gray value of image；Objective is placed on the reference plane, Target is imaged using terrible imaging system, using application of the Lambert cosine reflection law in terrible imaging system, according to figure The gray scale Gt (x, y) of each pixel as in, calculate in target representated by each pixel at the center of corresponding region light source direction and Angle β (x, y) between bucket detector direction, establishes the mapping relations between angle value and gray value of image；Finally calculate ginseng The difference θ (x, y) for examining angle at plane and target corresponding region obtains the depth delta d (x, y) of target range reference planes, from And reconstruct the three-dimensional configuration of target.

The present invention compared with prior art, remarkable advantage are as follows: (1) increase a reference planes, by bucket detector Number is reduced to one, reduces space cost and system cost；(2) in imaging process, a bucket detection is now only needed Device is avoided by multiple barrels of detector bring device errors；(3) algorithm for simplifying objective reconstruct, reduces reconstruct Time cost in the process；(4) objective reconfiguration technique is combined with terrible imaging technique, facilitates ghost and be imaged on actually to answer Popularization in, especially in the detection of industrial part and biological sample.

The present invention is described in further detail below in conjunction with the accompanying drawings.

Detailed description of the invention

Fig. 1 is to calculate terrible imaging optical path figure.

Fig. 2 is that the present invention is based on the mathematical model organigrams that the objective of SFS reconstructs.

Fig. 3 is the objective image that the present invention uses.

When Fig. 4 is that bucket detector of the invention is located at different location, (a) objective is located at different location in bucket detector When visual angle under image；(b) under correspondence visual angle when being located at different location for bucket detector in (a), terrible imaging three-dimensional is calculated The image of the three-dimensional configuration for the objective that object reconstruction technology reengineering goes out.

Fig. 5 is the absolute error comparison for objective terrible image reconstruction image under three kinds of different perspectivess that the present invention uses Figure.

Specific embodiment

In conjunction with Fig. 1, projector, bucket detector, computer and reference planes are placed on corresponding position according to Fig. 1, are joined It examines plane to be placed within the field range that can be imaged for calculating terrible imaging system, the optical center of projector and reference planes Center be in same optical axis, make projector projects go out two-value projection can be vertically projected in reference planes, bucket detector It is placed on the position for deviateing projector, guarantees that the central optical axis of bucket detector passes through the center of reference planes, relative position parameter It is known.The reference planes calculated in terrible imaging system are demarcated, generate one group of orderly hadamard matrix using computer, Using terrible imaging system is calculated, by the two-value projection generated by orderly hadamard matrix modulation according to the sequence of hadamard matrix It projects in reference planes frame by frame, reference planes is carried out to calculate terrible imaging, using Lambert cosine reflection law in ghost imaging system Application in system calculates image grayscale Gr (x, y) and is calculated representated by each pixel according to the gray scale of pixel each in image Angle α (x, y) in reference planes at the center of corresponding region between light source direction and bucket detector direction establishes angle value and figure As the mapping relations between gray value；The objective for needing to carry out three-dimensionalreconstruction is placed on the reference plane, it will be with reference When plane progress ghost imaging on identical one group of hadamard matrix two-value projection to objective, objective is counted Ghost imaging is calculated, image grayscale Gt (x, y) is calculated according to the gray scale of pixel each in image and calculates mesh representated by each pixel The angle β (x, y) at the center of corresponding region between light source direction and bucket detector direction is put on, angle value and image grayscale are established Mapping relations between value；Calculate the difference θ (x, y) of angle at reference planes and target corresponding region；Calculate target range ginseng The depth of plane is examined to Δ d (x, y), to reconstruct the three-dimensional configuration of objective.

In conjunction with the mathematical model organigram that Fig. 2 is reconstructed based on the objective of SFS, wherein α indicates each pixel institute Angle in the reference planes of representative at the center of corresponding region between light source direction and bucket detector direction, β indicate each pixel Angle in representative original object at the center of corresponding region between light source direction and bucket detector direction, θ indicate angle value α Make the difference of difference with angle value β, d indicates the center in each region in reference planes to the distance of bucket detector, Δ d expression three-dimensional Depth of each regional center of target to reference planes corresponding region.The present invention is based on the objective reconstruct sides for calculating ghost imaging Method, specific implementation step it is as follows:

Step 1, the terrible imaging system of calculating is demarcated, by projector, bucket detector and reference planes are placed according to Fig. 1 In corresponding position, the optical center of projector and the center of reference planes are in same optical axis, go out projector projects Two-value projection can be vertically projected in reference planes, and bucket detector is placed on the position for deviateing projector, guarantee bucket detector Central optical axis passes through the center of reference planes, and relative position parameter using the two-value for meeting hadamard matrix distribution it is known that projected Image carries out reference planes to calculate terrible imaging, and reference planes are imaged to obtain Gr (x, y) by calculating terrible imaging formula calculating, were imaged Journey is as follows:

Computer pre-generate probability distribution meet hadamard matrix distribution two-value projected image, using projector by its It is projected at reference planes frame by frame, the order of hadamard matrix is set as M², successively choose every a line and lay equal stress on and be made into the square of M*M Battle array, forms M altogether²A orderly matrix.For each of these matrix, makes therein -1 to become 0 first, utilize the square after change Battle array, which is modulated the digital micromirror array in projector, generates two-value projection, projects in target, bucket detector receives always Intensity value；1 in original matrix is become 0, -1 again and becomes 1, using this matrix after changing to the digital micro-mirror in projector Array, which is modulated, generates two-value projection, projects in target, bucket detector receives total intensity value, intensity value phase twice Subtract, the total intensity value as a result exactly received on a frame two-value projection to target is completed at last to the primary of target in this way Measurement；Identical operation is all carried out to the two-value projection that each frame hadamard matrix is formed, to the last a frame two-value has projected At operation；Reference planes, which are received, using bucket detector projects light intensity.

Reference planes calculate imaging by calculating terrible imaging formula, calculation formula such as formula (1):

Gr (x, y)=< I_r ⁽ⁱ⁾·P⁽ⁱ⁾(x,y)>-<I_r ⁽ⁱ⁾><P⁽ⁱ⁾(x,y)> (1)

X, y are plane coordinates.Gr (x, y) is the result for calculating ghost imaging and reference planes being calculated with imaging.I_r ⁽ⁱ⁾To calculate The value that bucket detector i-th detects in terrible imaging system.P⁽ⁱ⁾It is generated in i-th computer in (x, y) ghost imaging system Hadamard matrix two-value projected image.<...>symbol is to average, with<I_r ⁽ⁱ⁾·P⁽ⁱ⁾(x, y)>for,<I_r ⁽ⁱ⁾>and<P⁽ⁱ⁾(x, Y) > calculation method is identical.

N is that ghost imaging is associated with calculation times in formula (2).

Step 2, imaging results Gr (x, y) is calculated by calculating terrible imaging formula (1), according to the ash of pixel each in image Degree, calculates the folder in reference planes representated by each pixel at the center of corresponding region between light source direction and bucket detector direction Angle α (x, y), calculation formula such as formula (3):

α (x, y)=arccos (Gr (x, y)) (3)

Gr (x, y) is that the calculating ghost imaging of reference planes calculates the result of imaging.

Step 3, using common dimensional measuring instrument such as steel tape, the center for measuring each region in reference planes is visited to bucket Survey the distance d (x, y) of device.

Step 4, such as Fig. 3, the objective for needing to carry out three-dimensionalreconstruction is placed on the reference plane, x-axis and y-axis composition Plane be reference planes, z-axis indicate objective on region to reference planes on corresponding region depth.Utilize step 1 In reference planes calculate with same group of hadamard matrix two-value projected image of ghost imaging, to needing to carry out three-dimensionalreconstruction Objective carries out calculating terrible imaging, calculates imaging results Gt (x, y) by calculating terrible imaging formula, calculation formula such as formula (4):

Gt (x, y)=< I_t ⁽ⁱ⁾·P⁽ⁱ⁾(x,y)>-<I_t ⁽ⁱ⁾><P⁽ⁱ⁾(x,y)> (4)

Gt (x, y) is the result for calculating ghost imaging and calculating objective imaging in formula (4).I_t ⁽ⁱ⁾To calculate ghost imaging system The value that bucket detector i-th detects in system.

Step 5, according to the gray scale of pixel each in image, corresponding region on objective representated by each pixel is calculated Angle β (x, y) at center between light source direction and bucket detector direction, calculation formula such as formula (5):

β (x, y)=arccos (Gt (x, y)) (5)

Step 6, it is poor that the angle β (x, the y) angle α (x, y) obtained by step 2 and step 5 obtained makees, and obtains difference θ (x, y), formula (6) are as follows:

θ (x, y)=α (x, y)-β (x, y) (6)

Step 7, depth of the solution each regional center of objective to reference planes corresponding region.Step 3 is measured The θ (x, y) that the distance d (x, y) and step 6 arrived is acquired is accumulated, formula such as (7):

Δ d (x, y)=d (x, y) * θ (x, y) (7)

Δ d (x, y) is depth of each regional center of objective to reference planes corresponding region in formula (7), i.e., required Objective form.

Step 8, the position for changing bucket detector makes barrel detector be in projector three different positions nearby, but wants Meeting every bit on objective has illumination to project the precondition in barrel detector, reuses step 1- step 7 Method will obtain the three-dimensional configuration of the objective reconstructed from different bucket detector visual angles.As shown in figure 4, three width in (a) Figure is the image under visual angle of the objective when bucket detector is located at different location；(b) three width figures in are visited with bucket in (a) It surveys under correspondence visual angle when device is located at different location, calculates the three of the objective of terrible imaging three-dimensional object reconstruction technology reengineering out Tie up the image of form.In the case of three width figures in figure (5) respectively indicate three kinds of corresponding diagram (4), terrible imaging three-dimensional target is calculated Depth under visual angle of the depth and objective for the objective that reconfiguration technique reconstructs when bucket detector is located at different location The absolute error of degree, z-axis indicate difference.