Pavement structure depth measurement method based on binocular structured light

Abstract

The invention relates to a pavement structure depth measurement method based on binocular structured light, belongs to the technical field of pavement structure depth measurement, solves the technical problems of low measurement efficiency and poor stability, and comprises the following steps: erecting two cameras beside a pavement to be measured according to a certain angle, and determining internal and external parameters of the cameras by using a checkerboard calibration method; a projector is arranged between the two cameras, and the generated Gray codes and grid line images are sequentially projected on the surface of a measured object according to a sequence. Processing the acquired image to obtain an unwrapped phase diagram of the asphalt pavement; matching the left image and the right image by using a gray difference absolute value method AD to obtain a depth map, and calculating a real physical depth according to a camera calibration result; and calculating the construction depth of the surface of the asphalt pavement according to the definition of the construction depth. The measurement method is used for pavement structure depth measurement.

Description

Technical Field

[0001] The invention belongs to the technical field of pavement structure depth measurement, and in particular relates to a pavement structure depth measurement method based on binocular structured light. Background Art

[0002] Pavement structure depth is an indicator used to evaluate the size of the macro structure of asphalt pavement. A good macro structure can provide sufficient friction for vehicles, ensure the braking distance of vehicles, and improve driving safety. At present, the commonly used methods for measuring pavement structure depth are traditional sand spreading method, laser structure depth meter and digital image method; the traditional sand spreading method is simple in principle, but the operation is more complicated, and the measurement uncertainty is high. The laser structure depth meter is expensive, and the digital image method is easily disturbed by the external environment and cannot obtain measurement results in real time. Structured light is a triangulation method that is often used to measure the surface morphology of an object. It is usually necessary to use a sinusoidal grating to project the grating field onto its surface. Since the grating line projection field is modulated by the three-dimensional shape of the object and deformed, it is necessary to calculate the deformation of the grating field and demodulate the phase representing the height information of the object; expand the phase to obtain more accurate surface morphology information of the object. Structured light is not easily disturbed by external ambient light, and can provide more feature information for asphalt with a single surface texture feature, which is conducive to binocular matching. In the early days, binocular vision was mainly used in the fields of photogrammetry and computer vision. Two cameras were used to capture the feature information of the surface of the object being measured from different perspectives. The imaging principle and parallax calculation principle of binocular vision were used to obtain the geometric information of the object surface by matching feature points.

[0003] Specifically, there is a patent design for measuring the structural depth of asphalt pavement. Patent document CN109919856A discloses a method for measuring the structural depth of asphalt pavement based on binocular vision. However, it is still susceptible to interference from ambient light, and because the surface features of the asphalt pavement are single, multiple improved algorithms are required when performing binocular matching, which increases the calculation time and cannot obtain measurement results faster. Summary of the invention

[0004] In order to overcome the shortcomings of low efficiency and poor stability in pavement structure depth measurement, the present invention proposes a pavement structure depth measurement method based on binocular structured light.

[0005] The technical solution adopted by the present invention to solve the technical problem is:

[0006] A method for measuring road surface depth based on binocular structured light comprises the following steps:

[0007] Step 1: Based on the principle of binocular vision, two cameras are set up at a certain angle next to the road surface to be tested, and the internal and external parameters of the cameras are determined using the checkerboard calibration method.

[0008] Step 2: Place the projector between the two cameras and project the pre-generated Gray code and grid line images onto the surface of the object to be measured in sequence. Process the collected images to obtain the unwrapped phase map of the asphalt pavement.

[0009] Step 3: After the image correction, the left and right image phase maps are directly matched using the grayscale difference absolute value method AD to obtain the depth map, and the real physical depth is calculated based on the camera calibration result.

[0010] Step 4: Calculate the construction depth of the asphalt pavement surface according to the definition of construction depth.

[0011] For step 1, after arranging the two cameras at a certain angle, use the checkerboard calibration method to take about 20 calibration images. Use Zhang Zhengyou's calibration method to obtain the camera's internal and external parameters and write them into the program as fixed parameters.

[0012] For step 2, we place the projector between the two cameras, and project the pre-generated four-bit Gray code image and a complement code image onto the surface of the object to be tested, and project a four-step phase shift image. The camera is used to collect the image projected on the surface of the object. For the collected image, first solve its wrapped phase, and then use Gray coding to expand the phase.

[0013] For step 3, the specific operation steps are as follows: perform epipolar correction on the images to be matched; for a straight line on the left camera, its corresponding matching point must be on a straight line on the right camera. These two straight lines will limit the matching search range to a straight line, greatly reducing the search space size and improving matching efficiency. Epipolar correction is achieved by rotating the two cameras and redefining the new image plane so that the epipolar lines are collinear and parallel to a coordinate axis of the image plane. This operation also establishes a new stereo pair of images, so that the same matching point pair is located in the same row of the two views, and they only have a difference in horizontal coordinates, that is, parallax.

[0014] The main steps for performing epipolar correction are as follows:

[0015] 1. The rotation matrix of the right image relative to the left image plane calculated by camera calibration is converted into an axis angle using the Rodriguez formula. R is the rotation matrix, K is the basis vector of the defined rotation axis, and θ is the angle of rotation around axis K. The left and right cameras are rotated by θ / 2 respectively, and the rotation matrices R of the two cameras can be obtained by the inverse operation of the formula. l and R r .

[0016] R=I+sin(θ)K+(1-cos(θ))K 2

[0017] 2. Using the two rotation matrices obtained in the previous step, multiply the left and right cameras by the corresponding rotation matrices to make their axes parallel. However, at this time, the translation matrix T between the left and right cameras requires a new rotation matrix R to be constructed rect Make the baseline parallel to the imaging plane. Use the translation matrix T between the left and right cameras, normalize it to get e1, which is the translation vector between the left and right cameras; the direction of e2 is perpendicular to e1 and orthogonal to the main optical axis, then e2 can be obtained by the cross product of e1 and the main optical axis and normalized; finally, e3 can be obtained by the cross product of e1 and e2:

[0018] T=[T x T y T z ]

[0019]

[0020]

[0021] e3=e1×e2

[0022] Then we can get the matrix R that transforms the pole of the left camera to infinity rect as follows:

[0023] R rect =[e1 e2 e3] T

[0024] The two rotation matrices calculated in the first two steps are multiplied by the corresponding rotation matrices of the left and right cameras to obtain the overall rotation matrix of the left and right cameras. The corrected image can be obtained by multiplying the left and right camera coordinate systems by their corresponding overall rotation matrices, as shown in the formula:

[0025]

[0026] Through the above operations, the images of the left and right cameras after epipolar correction can be obtained.

[0027] After correction, it is necessary to redefine the image center so that the common field of view of the left and right images is maximized; that is, the intrinsic parameter matrix K needs to be reset. n ; Theoretically, K can be set arbitrarily, but in order to keep it consistent with the old camera as much as possible, the left and right images rectified by the left and right cameras are selected, and the ordinates of the matching points corresponding to the left and right images are equal.

[0028] After the image is processed, the phase images obtained by the left and right cameras can be matched using the principle of binocular vision. According to the corrected image, the corresponding point on the left camera image must be found on the corresponding row of the parallel right camera image, and the same point only has column pixels that are different in the two images. The three-dimensional morphology of the object surface can be obtained directly by using the principle of minimum absolute value of grayscale difference.

[0029]

[0030] For step 4, the specific operation steps are as follows: the total volume V of the measured area is calculated by calculating the minimum depth value of the asphalt area in the image and the interpolated sum of the depths of other pixels in the measured area. pixel .

[0031] V pixel =∑[h i -h min ]

[0032] h min Indicates the minimum value in the asphalt pavement area being tested; h i The depth values ​​of the pixels in the remaining measured areas. According to the volume method introduced above, the road surface structure depth in the measured area can be calculated by the following formula:

[0033]

[0034] Where m and n represent the number of rows and columns of pixels of the image occupied by the selected measured area, respectively.

[0035] The beneficial effects of the present invention are:

[0036] A road surface structural depth measurement method based on binocular structured light uses a non-contact measurement method to quickly and efficiently obtain the structural depth of the road surface. It is not easily disturbed by external ambient light, is conducive to portable measurement, and overcomes the problem of less characteristic information and difficult matching of asphalt pavement. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 A schematic diagram of four-bit Gray code encoding in the Gray code phase shift method used in an embodiment of the present application;

[0038] Figure 2 This is a coding diagram of a complementary Gray code added to a four-bit Gray code in an embodiment of the present application;

[0039] Figure 3 This is a schematic diagram of binocular vision imaging in an embodiment of the present application;

[0040] Figure 4 This is a schematic diagram of binocular matching after epipolar correction of an image in an embodiment of the present application;

[0041] Figure 5 A depth map of the asphalt pavement measured in an embodiment of the present application;

[0042] Figure 6 This is the three-dimensional morphology of the asphalt pavement restored in the embodiment of the present application. DETAILED DESCRIPTION

[0043] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] Embodiment 1

[0045] A road surface structure depth measurement method based on binocular structured light, using the basic principles of binocular vision and grid line projection, includes the following steps:

[0046] Step 1: Based on the principle of binocular vision, two cameras are set up at a certain angle next to the road surface to be tested, and the internal and external parameters of the cameras are determined using the checkerboard calibration method.

[0047] Step 2: Place the projector between the two cameras and project the pre-generated Gray code and grid line images onto the asphalt pavement in sequence. Process the collected images to obtain an unwrapped phase map of the asphalt pavement.

[0048] Step 3: After the image correction, the left and right image phase maps are directly matched using the grayscale difference absolute value method AD to obtain the depth map, and the real physical depth is calculated based on the camera calibration result.

[0049] Step 4: Calculate the construction depth of the asphalt pavement surface according to the definition of construction depth.

[0050] For the method described in step 1, the two cameras are arranged at an angle of thirty degrees to make the area of ​​the common field of view of the two cameras as large as possible. The checkerboard calibration is used to obtain the intrinsic and extrinsic parameters of the cameras. The two cameras are fixed and the intrinsic and extrinsic parameters of the cameras are written into the cameras. That is, the position and posture of the cameras are fixed and the cameras are not moved in subsequent measurements.

[0051] The camera calibration adopts Zhang Zhengyou calibration method. The world coordinate system is established on the calibration plate, and the coordinates of the feature points of the calibration plate are (X w ,Y w ,0). By shooting the calibration plate from different angles, the pixel coordinates (u,v) and world coordinates (X w ,Y w ,0) information, linearly solve the camera internal and external parameters;

[0052]

[0053] Among them, r1, r2, r3 are three column vectors of the rotation matrix R, simplifying the above formula:

[0054]

[0055] Where s is the scale factor, H is the homography matrix, which is used to describe the mapping relationship between the feature point coordinates in the plane and the image plane, h1 = Kr1, h2 = Kr2, h3 = Kt. K is the internal parameter matrix.

[0056] When constructing the rotation matrix, each column vector is mutually orthogonal, so each homography matrix can provide two constraints: the dot product of the rotation vector is zero and the length of the rotation vector is equal, that is,

[0057]

[0058] make

[0059]

[0060] B is a symmetric matrix, which can be expressed as follows:

[0061] b=[B 11 B 12 B 22 B 13 B 23 B 33 ] T

[0062] The six unknowns in vector b require six sets of equations to be solved. Each calibration image can provide two equations. Therefore, the linear solution of the camera intrinsic parameters requires at least three calibration images. After solving vector b, the camera intrinsic parameters can be expressed as:

[0063]

[0064] Cx and Cy are the coordinates of the principal point, fx and fy are the focal lengths in the x and y axes.

[0065] After finding the internal parameter matrix K, solve the external parameters:

[0066]

[0067] After obtaining the initial values ​​of the camera's internal and external parameters, the least squares method can be used to solve the initial values. In reality, there must be errors in the extraction of feature points and the processing of the calibration plate, so the linear solution is greatly affected by noise. At this time, the nonlinear optimization method is used to optimize all internal and external parameters, and the results are made more accurate during the iteration process.

[0068] According to the pixel coordinates of the feature points (u i ,v i) and the coordinates of the space point (X i ,Y i ,Z i ) have the following projection relationships:

[0069]

[0070] When the feature points in the world coordinate system are projected to the pixel coordinates through the above formula, due to the presence of noise and other factors, there will be an error between the projection position in the three-dimensional space and the coordinates of the feature points detected in the image. This error is called the reprojection error. The Euclidean distance (2 norm) between the theoretical and actual coordinates of all feature points is summed up and the least squares problem is constructed to perform nonlinear optimization on the parameters in the K and T matrices. The objective function is as follows

[0071]

[0072] For the calibration of the binocular camera, at a point P in space, its coordinates in the world coordinate system are (X w ,Y w ,Z w ), the coordinates in the left and right camera coordinate systems are and The external parameters of the left and right cameras are R1, t1 and R2, t2 respectively, and we can get:

[0073]

[0074] Simplifying the above formula, we get:

[0075]

[0076] All the obtained parameters can be nonlinearly optimized by minimizing the reprojection error to obtain the calibration result of the binocular camera.

[0077] For the method described in step 2, a Gray code-phase shift method is used for time encoding of the fringe image projected by the projector.

[0078] The phase shift method solves the phase information of the object surface. The grayscale relationship of the object surface is shown in the formula, where is the phase of the projection. For the N-step phase shift method, the phase difference between the two projected phase images is 2π / N, which can be expressed as the following set of equations:

[0079]

[0080] Where i = 0, 1, 2, ..., N-2, N-1. A system of equations consisting of N equations, including three unknowns, can be solved theoretically when N ≥ 3. However, due to accuracy issues, in actual operation, N ≥ 4 is generally used, so that the solution system can be transformed into a hyperstatic equation solution.

[0081] The current system of equations is a nonlinear system of equations, which seems to be impossible to solve using direct methods. However, it is noted that the equations can be transformed into the following form:

[0082]

[0083] Where i = 0, 1, 2, 3, ..., N-2, N-1. Through simple trigonometric expansion, the quantities to be determined are A(x, y), B(x, y) and The nonlinear equations of are transformed into linear equations about the quantities to be determined A(x,y), B1(x,y) and B2(x,y). That is, the set of equations to be solved can be transformed into the following matrix form:

[0084]

[0085] The above formula can be simplified as:

[0086] N (N×3) x (3×1) =I (N×1)

[0087] According to the theory of least square method for solving statically indeterminate equations, the optimal solution of the equation is:

[0088] x=(N T N -1 )N T I

[0089] According to the construction of B1(x,y) and B2(x,y), we can know that:

[0090]

[0091] Combining the least squares solution of the statically indeterminate equations, we obtain:

[0092]

[0093] The phase values ​​obtained by the above method are distributed between [-π,π]. At this time, the obtained phase diagram is a wrapped phase. In order to obtain the real phase information of the object surface, the phase needs to be unfolded. Specifically, every time a new cycle is encountered, the next cycle needs to be corrected, usually by adding or subtracting 2π cycle phases. The formula is as follows:

[0094]

[0095] In this method, we use Gray code to unwrap the phase.

[0096] Gray code is an unweighted code. Its adjacent codes only change by one digit. It can reduce the error caused by misalignment during decoding. Therefore, it is also called minimum error code. The method of generating Gray code usually adopts recursion. After observation, it can be found that n-bit Gray code can be obtained from (n-1)-bit Gray code, that is:

[0097] 1. (n-1)-bit Gray code is arranged in positive order and the leftmost (prefix) is filled with 0;

[0098] 2. Arrange the (n-1)-bit Gray code in reverse order and add 1 to the leftmost (prefix);

[0099] 3.1. Arrange the results obtained in steps 3.1 and 3.2 in sequence to obtain an n-bit Gray code.

[0100] In the experiment, the number of Gray code bits should be selected according to the period of the projected grating line, that is, x ≥ 2 n , x is the number of grid line cycles, and n is the number of Gray code bits. In the experiment, the collected Gray code image is not black and white, so to binarize the image, it is necessary to project two additional images of all white and all black on the surface of the object to be measured, and take the average of the two images as the threshold for binarizing the Gray code. In order to ensure the accuracy of phase unfolding, complementary Gray code is usually used to assist in phase unfolding, that is, an additional Gray code pattern is projected on the traditional Gray code pattern, and n+1-bit Gray code encoding is used to assist in decoding the phase, avoiding spike errors caused by Gray code boundary decoding errors at the truncated phase of the unfolded continuous phase.

[0101] GC1 to GC4 are four images generated by four-bit Gray code, GC5 is a complementary Gray code image, k1 is obtained in the same way as the traditional Gray code, and k2 is solved by adding the fifth Gray code on the basis of the previous one.

[0102]

[0103] k2(x,y)=INT((i(V2(x,y))+1) / 2)

[0104] The obtained k2 has a half-cycle shift relative to k1, and then the wrapped phase is expanded:

[0105]

[0106] It can be seen that the complementary Gray code diagram can effectively avoid the level error problem caused by phase jump when the phase is unfolded.

[0107] For the method described in step 3, the specific operation steps are as follows: perform epipolar correction on the image to be matched; for a straight line on the left camera, its corresponding matching point must be on a straight line on the right camera, which limits the matching search range to a straight line, greatly reducing the search space size and improving matching efficiency. Epipolar correction is achieved by rotating the two cameras and redefining the new image plane so that the epipolar pairs are collinear and parallel to a coordinate axis of the image plane. This operation also establishes a new stereo pair, so that the same matching point pair is located in the same row of the two views, and they only have a difference in horizontal coordinates, that is, parallax.

[0108] The main steps for performing epipolar correction are as follows:

[0109] 1. The rotation matrix of the right image relative to the left image plane calculated by camera calibration is converted into an axis angle using the Rodriguez formula. R is the rotation matrix, K is the basis vector of the defined rotation axis, and θ is the angle of rotation around axis K. The left and right cameras are rotated by θ / 2 respectively, and the rotation matrices R of the two cameras can be obtained by the inverse operation of the formula. l and R r .

[0110] R=I+sin(θ)K+(1-cos(θ))K 2

[0111] 2. Using the two rotation matrices obtained in the previous step, multiply the left and right cameras by the corresponding rotation matrices to make their axes parallel. However, at this time, the translation matrix T between the left and right cameras requires a new rotation matrix R to be constructed rect Make the baseline parallel to the imaging plane. Use the translation matrix T between the left and right cameras, normalize it to e1, which is the translation vector between the left and right cameras; the direction of e2 is perpendicular to e1 and orthogonal to the main optical axis, so e2 can be obtained by the cross product of e1 and the main optical axis and normalized; T x ,T y ,T z are the three column vectors in the translation matrix; finally, e3 can be obtained by the cross product of e1 and e2:

[0112] T=[T x T y T z ]

[0113]

[0114]

[0115] e3=e1×e2

[0116] Then we can get the matrix R that transforms the pole of the left camera to infinity rect as follows:

[0117] R rect =[e1 e2 e3] T

[0118] The two rotation matrices calculated in the first two steps are multiplied by the corresponding rotation matrices of the left and right cameras to obtain the overall rotation matrix of the left and right cameras. The corrected image can be obtained by multiplying the left and right camera coordinate systems by their respective corresponding overall rotation matrices, as shown in the formula:

[0119]

[0120] Through the above operations, the images of the left and right cameras after epipolar correction can be obtained.

[0121] After correction, it is necessary to redefine the image center so that the common field of view of the left and right images is maximized; that is, the intrinsic parameter matrix K needs to be reset. n Theoretically, K n It can be set arbitrarily, but in order to keep it consistent with the old camera, the left and right images corrected by the left and right cameras are selected, and the vertical coordinates of the matching points corresponding to the left and right images are equal.

[0122] After the image is processed, the phase images obtained by the left and right cameras can be matched using the principle of binocular vision. According to the corrected image, the corresponding point on the left camera image must be found on the corresponding row of the parallel right camera image, and the same point only has different columns of pixels in the two images. The absolute value of the grayscale difference C is directly used. AD The (p, d) minimum principle can be used to obtain the three-dimensional morphology of the object surface.

[0123]

[0124] For the method described in step 4, the specific operation steps are as follows: the total volume V in the measured area is calculated by calculating the minimum value of the depth value of the asphalt area in the image and the sum of the interpolated values ​​of the depths of other pixels in the measured area. pixel .

[0125] V pixel =∑[h i -h min ]

[0126] h min Indicates the minimum value in the asphalt pavement area being tested; h i is the depth value of the pixel points in the rest of the measured area. According to the volume method introduced above, the road surface structure depth in the measured area can be calculated by the following formula:

[0127]

[0128] Where m and n represent the number of rows and columns of pixels of the image occupied by the selected measured area, respectively.

Claims

1. A road surface structure depth measurement method based on binocular structured light, characterized in that: The following steps are involved: Step 1: Based on the principle of binocular vision, two cameras are set up at a certain angle next to the road surface to be tested. Use the chessboard calibration method to determine the internal and external parameters of the camera; Step 2: Arrange the projector between the two cameras, and project the generated Gray code and grid line images in sequence on the surface of the road to be tested; process the collected images to obtain the unfolded phase image of the road to be tested; Step 3: Use the grayscale difference absolute value method AD to match the phase images of the left and right images after image correction to obtain a depth map, and calculate the real physical depth based on the camera calibration result; Step 4: Calculate the construction depth of the asphalt pavement surface according to the definition of construction depth.

2. The method for measuring the depth of a road surface structure according to claim 1, characterized in that: The step 1 further comprises: Arrange the two cameras at a certain angle to make the common field of view of the two cameras as large as possible. Use chessboard calibration to obtain the internal and external parameters of the camera. The two cameras are fixed and the internal and external parameters of the camera are written into the camera. Use Zhang Zhengyou calibration method to implement camera calibration. The camera calibration process is: establish the world coordinate system on the calibration board, and the coordinates of the feature points of the calibration board are (X w ,Y w ,0); by shooting the calibration plate from different angles, the pixel coordinates (u,v) and world coordinates (X w ,Y w ,0) information, linearly solve the camera internal and external parameters; The intrinsic parameters of the camera are expressed as: After finding the internal parameter matrix K, solve the external parameters: After obtaining the initial values ​​of the camera's internal and external parameters, the least squares method is used to solve the initial values; The following formula is obtained: By minimizing the reprojection error, all the obtained parameters are nonlinearly optimized to obtain the calibration result of the binocular camera.

3. The method for measuring the depth of a road surface structure according to claim 1, characterized in that: The specific process of step 2 is: The phase shift method solves the phase information of the object surface as follows: The phase value obtained by the above formula is distributed between [-π,π]. Whenever a new cycle is encountered, the next cycle needs to be corrected by adding or subtracting 2π cycle phases. The specific calculation formula is as follows: Gray code is used to unwrap the phase.

4. The method for measuring the depth of a road surface structure according to claim 3, characterized in that: The Gray code phase expansion further comprises: Gray code phase unwrapping, using complementary Gray code to assist in phase unwrapping; GC1 to GC4 are four images generated by four-bit Gray code, GC5 is a complementary Gray code image, k1 is obtained in the same way as the traditional Gray code, and k2 is solved by adding the fifth Gray code on the basis of the previous one. The solution process is as follows: k2(x,y)=INT((i(V2(x,y))+1) / 2) The obtained k2 has a half-cycle shift relative to k1, and the wrapped phase is expanded as follows:

5. The method for measuring the depth of a road surface structure according to claim 1, characterized in that: The step 3 further comprises: Perform epipolar correction on the images that need to be matched; epipolar correction rotates the two cameras and redefines the new image plane so that the epipolar lines are collinear and parallel to a coordinate axis of the image plane. This operation also creates a new stereo pair of images; the same pair of matching points is located in the same row of the two views and has only parallax.

6. The method for measuring the depth of a road surface structure according to claim 5, characterized in that: The specific process of the polar line correction is as follows: The first step is to convert the rotation matrix of the right image relative to the left image plane calculated by camera calibration into an axis angle using the Rodriguez formula. R is the rotation matrix, K is the basis vector of the defined rotation axis, and θ is the angle of rotation around axis K. The left and right cameras are rotated by θ / 2 respectively, and then the rotation matrices R of the two cameras are obtained using the inverse operation of the formula. l and R r ; R=I+sin(θ)K+(1-cos(θ))K 2 In the second step, the left and right cameras are multiplied by the corresponding rotation matrix to make their axes parallel; the translation matrix T between the left and right cameras is used to normalize it, that is, e1, where e1 is the translation vector between the left and right cameras; the direction of e2 is perpendicular to e1 and orthogonal to the direction of the main optical axis; then e2 is obtained by the cross product of e1 and the direction of the main optical axis and normalized; finally, e3 is obtained by the cross product of e1 and e2: T=[T x T y T z ] e3=e1×e2 Get the matrix R that transforms the pole of the left camera to infinity rect , the expression is as follows: <h2 style=";text-align:left;direction:ltr">R<h2 style=";text-align:left;direction:ltr"> rect <h2 style=";text-align:left;direction:ltr"> (e1 e2 e3)<h2 style=";text-align:left;direction:ltr"> T The two obtained rotation matrices are multiplied with the corresponding rotation matrices of the left and right cameras to obtain the overall rotation matrix of the left and right cameras; the left and right camera coordinate systems are multiplied by their respective corresponding overall rotation matrices to obtain the corrected image, which is expressed as follows: Obtain the images of the left and right cameras after epipolar correction; After correction, it is necessary to redefine the image center so that the common field of view of the left and right images is maximized; that is, the intrinsic parameter matrix K needs to be reset. n ; Theoretically, K can be set arbitrarily. In order to keep it consistent with the old camera as much as possible, the left and right images corrected by the left and right cameras are selected, and the ordinates of the matching points corresponding to the left and right images are equal; After the image is processed, the phase images obtained by the left and right cameras are matched using the principle of binocular vision. According to the corrected image, the corresponding point on the left camera image must be found on the corresponding row of the parallel right camera image, and the same point only has column pixels different in the two images. The three-dimensional morphology of the object surface is obtained directly using the principle of the minimum absolute value of the grayscale difference. The expression is as follows; in, Indicates the gray value of a point on the left image; Indicates the grayscale value of the pixel d away from the corresponding row in the left image on the right image.

7. The method for measuring the depth of a road surface structure according to claim 1, characterized in that: The step 4 further comprises: The total volume V in the measured area is calculated by calculating the minimum depth value of the asphalt area in the image and the interpolated sum of the depths of other pixels in the measured area. pixel , the calculation formula is as follows: V pixel =∑[h i -h min ] Among them, h min Indicates the minimum value in the asphalt pavement area being tested; h i Indicates the depth values ​​of the pixels in the remaining measured areas; The road surface structure depth MTD in the measured area is calculated by the following formula: Wherein, m and n represent the number of rows and columns of pixels of the image occupied by the selected measured area, respectively.

8. The method for measuring the depth of a road surface structure according to claim 2, characterized in that: The two cameras are arranged at a certain angle, and the angle is 30°.

Images