Cross structured light calibration method based on planar target

Abstract

This invention proposes a method for calibrating a cross-shaped structured light based on a planar target. The method includes acquiring images of intersecting laser stripes projected onto a planar target, extracting the center points of the laser stripes, removing center points of laser stripes located at the edges of feature points on the planar target, extracting the positions of the intersection points of the intersecting laser stripes in the image using image moments, incorporating these intersection points into a line fitting process, and performing line fitting to remove laser stripe center points that deviate significantly from the fitted line. Finally, the internal and external parameters of the camera target are obtained through camera calibration, the three-dimensional coordinates of the laser stripe center points on multiple images are calculated, and light plane fitting is performed to complete the calibration of the cross-shaped structured light. This invention effectively improves the calibration accuracy of the cross-shaped structured light by removing laser stripe center points close to the edges of feature points on the planar target and incorporating the intersection points of the intersecting laser stripes into the cross-shaped structured light calibration process.

Description

Technical Field

[0001] This invention relates to the field of laser calibration, and more specifically to a cross-shaped calibration method based on a planar target. Background Technology

[0002] Line structured light inspection is one of the most common methods for measuring the geometric parameters of objects and achieving 3D reconstruction. Due to its non-contact nature, large range, high speed, high accuracy, stable algorithm, and simple structure, it has been widely used in industrial inspection.

[0003] Currently, linear structured light is commonly used as the projection source. To address the potential issue of incomplete information in 3D reconstruction using linear structured light, a cross-shaped structured light vision sensor has been developed based on structured light 3D vision measurement technology. The cross-shaped structured light vision sensor is a vision sensor based on the cross-shaped structured light method and mainly consists of an optical projector and a camera.

[0004] An optical projector projects two intersecting light planes in space. The two light planes intersect with the surface of the object being measured to form cross-shaped deformation feature lines. A camera simultaneously captures images of the cross-shaped deformation feature lines, which are then processed to obtain the three-dimensional information of the object's surface.

[0005] The cross-shaped vision sensor is an extension of single-line structured light. It can acquire surface information of an object in two directions simultaneously, avoiding the limitations of complex mechanical devices required for single-line structured light scanning. This significantly improves detection efficiency and makes it ideal for non-contact, high-efficiency measurement of object geometric parameters and form and position errors. The calibration of the measurement model and model parameters of the cross-shaped structured light sensor is crucial for its successful application.

[0006] However, when existing calibration methods use planar targets for structural calibration, on the one hand, multiple planar targets are required, resulting in a more complex structure; on the other hand, in order to improve the extraction accuracy of feature points on the planar targets, the colors of the feature points on the planar targets will have a large color difference from the background color of the planar targets. However, since different colors have different reflectivities, if the laser stripe of the laser is projected onto the edge of the feature point of the planar target with a significant color change, the laser stripe no longer conforms to a Gaussian distribution, which will cause the position of the center line extracted by the Steger algorithm to deviate. This deviation will directly affect the subsequent calibration process of system parameters, resulting in inaccurate calibration. Summary of the Invention

[0007] To address the shortcomings of the prior art, this invention provides a cross-shaped beam calibration method based on a planar target. By removing the center points of laser stripes near the edge of the feature points of the planar target and incorporating the intersection points of the intersecting laser stripes into the cross-shaped beam calibration process, the calibration accuracy of the cross-shaped beam is effectively improved.

[0008] Specifically, a cross-shaped target aiming method based on a planar target includes the following steps:

[0009] S1. Image Acquisition: Acquire P groups of planar target images in different poses using a camera and store them. Each group of images includes one planar target image with laser stripes and one planar target image without laser stripes.

[0010] S2. Extract the center points of the laser stripes from the P images of the planar target with laser stripes and perform the first screening: After processing the two images in each of the P groups of images obtained in step S1, P images are obtained. Using the method of image contour extraction and dilation, the center points of the laser stripes that are close to the edge of the feature points of the planar target are removed from the images. The extracted center points of the laser stripes are screened for the first time, and the center points of the laser stripes located on the contour of the feature points of the planar target in the P images are removed, resulting in two sets of center points of the laser stripes in the P images.

[0011] S3. A second screening of the laser stripe center points is performed through line fitting: Using the intersection criterion as the endpoint, the center points of the two intersecting laser stripes in each image after screening in step S2 are traversed to obtain the region where the intersection of the two intersecting laser stripes is located. Then, the precise position of the intersection of the two laser stripes is calculated using image moments. This intersection point and the two center point sets obtained in step S2 are used as input for line fitting. Line fitting is performed, and the two points with the best fitting effect for the center line of each laser stripe in each image are retained. The four retained points in each image are then combined into four new point sets. Specifically, this includes the following steps:

[0012] S31. The coordinates of the center points of the laser stripes in the P images after the first screening in step S2 are represented as follows:

[0013]

[0014] Where p represents the sequence number of different images, pints p,1,i This represents the center point of the i-th laser stripe projected by the first laser onto the planar target in the p-th image, where Ns is the total number of center points on that laser stripe, and u i and v i They represent points respectively p,1,i x and y coordinates in an image; points p,2,j u represents the center point of the j-th laser stripe projected by the second laser onto the planar target in the p-th image. j and v j They represent points respectively p,2,j In the image, the horizontal and vertical coordinates are represented by Ms, which is the total number of center points on the laser stripe. If the line segment point...p,1,i point p,1,i+1 and line segment point p,2, j point p,2,j+1 If they intersect, then the following intersection criterion applies:

[0015]

[0016] By points p,1,i points p,2,j The two center point sequences are iteratively searched until the intersection criterion described above is met, at which point the values ​​of i and j are determined. Then, the preliminary position (point) of the intersection point A of the two laser stripes on the planar target in each image is calculated. p,A,0 :

[0017]

[0018] Among them, u A0 and v A0 Representing Point p,A,0 The horizontal and vertical coordinates in the image.

[0019] S32. Calculate the exact coordinates of intersection point A in each image: obtain the preliminary position of intersection point A. p,A,0 Then, using point p,A,0 Region V is extracted from the image with center W and width W. The zeroth-order and first-order image moments of region V are calculated, and the exact coordinates of the gray-level center of the region, i.e., the intersection point A, are finally obtained. p,A ;

[0020] Zeroth moment:

[0021] First moment:

[0022]

[0023] point p，A =[u A v A ,1] T

[0024] Where I and J are the row and column sequence numbers in region V, respectively, V(I,J) is the pixel value in the I-th row and J-th column of region V, and M... 10 and M 01 These are the two components of the first moment of the image, u A v is the x-coordinate of the grayscale center. A The ordinate of the grayscale center;

[0025] S33. Using the Ransac algorithm, utilize the two laser stripe center point sets (points) selected in step S2 in each image. p，1，i points p，2，j The exact coordinates of the intersection point A p，A Perform linear fitting on the center lines of the two laser stripes in each image:

[0026] The three points used for straight-line fitting of the center line of the laser stripe projected by the first laser are: point p，1，i1 point p，1，i2 point p，A The three points used for straight-line fitting of the center line of the laser stripe projected by the second laser are: point p，2，j1 point p，2，j2 point p，A ;

[0027] Where point p，1，i1 point p，1，i2 From point set p，1，i Two points randomly selected from the data, point p，2，j1 point p，2，j2 From point set p，2，j Two points randomly selected from the data;

[0028] After the Ransac straight line fitting process is completed, the four points randomly extracted from each image with the best fitting effect are retained to form four new laser stripe center point sets pr for P images. p，1，1 pr p，1，2 pr p，2，1 pr p，2，2 And expressed as:

[0029]

[0030] Among them, pr p，1，1 and pr p，1，2 Let pr represent the two points extracted when the center line of the laser stripe projected by the first laser in the p-th image has the best fitting effect. p，2，1 and pr p，2，2 These represent the two points extracted when the center line of the laser stripe projected by the second laser in the p-th image has the best fitting effect;

[0031] S4. Perform light plane fitting and complete the light plane calibration of the cross-shaped structured light based on the light plane fitting results: calibrate the camera, and use the Ransac algorithm to perform light plane fitting of the cross-shaped structured light from the four laser stripe center point sets left after the screening in step S3 and the point set composed of the intersection points of the intersecting laser stripes in each image. This specifically includes the following sub-steps:

[0032] S41. Calibrate the camera. The camera's internal parameters are:

[0033]

[0034] The coordinates of the intersection point A of the laser stripes on the planar target on the camera imaging plane are:

[0035] camera p，A =K -1 ·point p，A =[x c y c ,1] T ;

[0036] The coordinates in the camera coordinate system are:

[0037] Camera p，A =Z c ·[x c y c ,1] T =[X c Y c Z c ] T ;

[0038] The coordinates in the plane target coordinate system are:

[0039] board p，A =[X A Y A [0, 1] T ;

[0040] The pose of the planar target in the p-th image is:

[0041]

[0042] The coordinates of the intersection point A of the laser stripes on the planar target in the camera coordinate system are:

[0043]

[0044] After simplification, we get:

[0045]

[0046] Combining the above two equations, we obtain the position of the intersection point A of the laser stripes in the p-th image in the camera coordinate system. p，A ;

[0047] Similarly, the set of four laser stripe center points pr, formed by the points extracted when the Ransec algorithm achieves the best line fitting effect in step S3, is obtained.p，1，1 pr p，1，2 pr p，2，1 pr p，2，2 Three-dimensional coordinates in the camera coordinate system:

[0048]

[0049] S42. Use the Ransac algorithm to perform optical plane fitting for the first laser, from the above point set camera. p，1，1 camera p，1，2 One point is randomly selected from each of the point sets Camera. p，A Two points are randomly selected from the above point set, where p = 1, 2, ..., P. The optical plane equation of the first laser is obtained by fitting these four points. The process of randomly selecting four points and fitting the plane is repeated, and the result with the best fitting effect is retained as the final result of the optical plane fitting. Similarly, from the above point set, the optical plane equation of the first laser is obtained. p，2，1 camera p，2，2 One point is randomly selected from each of the point sets Camera. p，A Two points are randomly selected from the sample, and the final result of the optical plane fitting of the second laser is obtained according to the above method; finally, the optical plane calibration of the cross-shaped structured light is completed, and the optical plane equations of the two lasers are obtained:

[0050]

[0051] Preferably, step S1 specifically includes the following steps:

[0052] S11. Fix the position of the camera and the two line lasers. The camera and the line lasers are fixedly mounted on the same bracket, and the relative position between them is fixed.

[0053] S12. Fix the planar target within the camera's field of view. Two lasers project lasers onto the planar target to form intersecting laser stripes. The camera takes one image with laser stripes and one image without laser stripes on the planar target when the lasers are on and off, respectively.

[0054] S13. Remove the planar target and change its pose, then take another picture. Change the pose of the planar target P times and take a total of P sets of planar target images with different poses. Each set of images corresponds to a pose of the planar target.

[0055] S14. Store the captured P-group planar target images.

[0056] Preferably, step S2 specifically includes the following steps:

[0057] S21. Extract the contours of the P planar target images without laser stripes obtained in step S1 to obtain the contour map of the planar target. Dilate the contour map. The width of the convolution kernel after dilation is greater than the width of the convolution kernel when the center of the laser stripe is extracted using the Steger algorithm.

[0058] S22. Subtract the planar target image with laser stripes from the planar target image without laser stripes in each group of images obtained in step S1 to obtain P images. Remove the influence of ambient light and use the Steger algorithm to extract the center point of the laser stripes in the image after the subtraction operation. The width of the convolution kernel in the Steger algorithm is smaller than the width of the convolution kernel during dilation.

[0059] S23. Using the dilated planar target feature point contour, the laser stripe center points extracted by the Steger algorithm are filtered out. The laser stripe center points located on the dilated planar target feature point contour in the image set are removed, and the remaining laser stripe center points are retained to obtain two sets of laser stripe center points for P images.

[0060] Preferably, the camera is calibrated using the Zhang Zhengyou calibration method in S41.

[0061] Preferably, in step S2, the least squares method is used to perform line fitting to obtain the fitted straight line of the two intersecting laser stripes.

[0062] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0063] (1) The innovative calibration method of this invention introduces the intersection of the two intersecting laser stripes generated by the cross structure light projected onto the planar target into the calibration process of the cross structure light. It can overcome the deviation generated during the calibration process, calculate the accurate position of the center line of the laser stripe, and effectively improve the calibration accuracy of the cross structure light.

[0064] (2) Before extracting the center point of the laser stripe, the present invention uses an image dilation algorithm to remove the center point of the laser stripe located near the edge of the feature point of the planar target, which can ensure the accuracy of the extraction of the center point of the laser stripe and thus effectively improve the positioning accuracy of the cross structure cursor.

[0065] (3) The present invention calculates the precise position of the intersection of two laser stripes by image moments, introduces the intersection point into the straight line fitting and retains the four points selected when the fitting effect is best in each image, forms a new point set by the four points retained in each image, and performs light plane fitting again based on the point set, thereby ensuring the accuracy of calibration.

[0066] (4) The device of the cross-shaped cursor positioning method proposed in this invention is very simple. It only requires a planar target and has no strict requirements on the installation position of the target. It is easy to operate and not prone to errors. It will not cause problems with the accuracy of cross-shaped cursor positioning due to operational errors. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the process of the present invention;

[0068] Figure 2 This is a schematic diagram of the overall structure of the present invention;

[0069] Figure 3 This is a schematic diagram of a planar target image with laser stripes captured by the camera of this invention;

[0070] Figure 4 This is the outline of the planar target after contour extraction according to the present invention;

[0071] Figure 5 This is a contour diagram of the planar target after contour expansion according to the present invention;

[0072] Figure 6 This is a schematic diagram illustrating the method of using image moments to determine the intersection points of the laser stripe center lines in this invention.

[0073] Figure 7 This is a comparison chart of errors obtained after the control experiment in the embodiments of the present invention. Detailed Implementation

[0074] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0075] like Figure 1 As shown, the cross-shaped cursor positioning method proposed in this invention is performed according to the following steps:

[0076] S1. Fix the positions of the camera and two line lasers. The camera and line lasers are fixedly mounted on the same bracket, with their relative positions remaining unchanged. Fix the planar target within the camera's field of view. The two lasers project their lasers onto the planar target, forming intersecting laser stripes. The camera and laser positions are as follows: Figure 2 As shown, the camera takes pictures of the planar target when the laser is off and on, respectively, obtaining one image of the planar target without laser stripes and one image of the planar target with laser stripes. The planar target is removed and its pose is changed, and the pictures are taken again. The pose of the planar target is changed P times, and a total of P sets of planar target images with different poses are taken. Each set of images corresponds to one pose of the planar target. The P sets of images are then input into the computer. Figure 3 The image shows, schematically, a planar target image with laser stripes.

[0077] S2. Perform image processing on the P groups of images obtained in the first step. Use image contour extraction and dilation processing methods to remove the center points of laser stripes that are close to the edge of the planar target feature points in the image. Then, perform a screening on the center points of the extracted laser stripes.

[0078] Specifically:

[0079] Contour extraction is performed on the P original planar target images without laser stripes obtained in the first step to obtain the contour maps of the planar target. The contour maps are then dilated, with the convolution kernel width being larger than the kernel width used in the Steger algorithm to extract the laser stripe centers. The P planar target images with and without laser stripes obtained in the first step are subtracted to remove the influence of ambient light. The Steger algorithm is then used to extract the laser stripe center points in the subtracted image, where the convolution kernel width in the Steger algorithm is smaller than the kernel width used in the dilation process. The laser stripe center points extracted by the Steger algorithm are then filtered using the dilated planar target feature point contours. Laser stripe center points located on the dilated planar target feature point contours are removed, and the remaining laser stripe center points are retained, resulting in two sets of laser stripe center points from a total of P images.

[0080] Figure 4 This schematically illustrates the contour map of a planar target after contour extraction. Figure 5 The schematic diagram shows the outline of a planar target after contour dilation.

[0081] If the center points of the laser stripes extracted by the Steger algorithm are used directly without screening to fit a straight line, the average deviation of the fitted line is 0.0358 pixels. However, if the center points of the laser stripes remaining after screening are fitted with dilation processing, the average deviation of the fitted line is 0.0608 pixels. This shows that screening before fitting the line significantly reduces the error.

[0082] S3. Traverse the center points of the two laser stripes obtained after the second step of screening to obtain the region where the intersection of the two laser stripes is located. Then calculate the precise position of the intersection of the two laser stripe centers using image moments. Introduce the intersection point into the line fitting. Use the Ransac algorithm to perform line fitting and retain the 4 points selected with the best fitting effect in each image. The 4 points retained in each image form a new point set.

[0083] The coordinates of the center points of the laser stripes in the P images after the second step of filtering are represented as follows:

[0084]

[0085] Where p represents the image index, points p，1，i u represents the i-th center point of the laser stripe projected by the first laser onto the planar target in the p-th image. i and v i They represent points respectively p，1，i In the image, the horizontal and vertical coordinates are represented by Ns, which is the total number of center points on the laser stripe. p，2，j u represents the center point of the j-th laser stripe projected by the second laser onto the planar target in the p-th image. j and v j They represent points respectively p，2，j In the image, the horizontal and vertical coordinates are represented by Ms, which is the total number of center points on the laser stripe. If the line segment point... p，1，i point p，1，i+1 and line segment point p，2，j point p，2，j+1 If they intersect, then the following intersection criterion applies:

[0086]

[0087] By points p，1，i points p，2，j The two point sequences are iteratively searched until the intersection criterion described above is met, at which point the values ​​of i and j are determined. Then, the preliminary position (point) of the intersection point A of the two laser stripes on the planar target in each image is calculated. p，A，0 :

[0088]

[0089] Among them, u A0 and v A0 Representing Point p，A，0 The horizontal and vertical coordinates in the image.

[0090] After obtaining the initial position of intersection point A, use point p，A，0 Region V is extracted from the image with Q as the center and Q as the width. The zeroth-order and first-order image moments of region V are calculated, and finally the gray-level center of the region, i.e., the accurate coordinates of the intersection point A, are obtained. p，A ;

[0091] Zeroth moment:

[0092] First moment:

[0093]

[0094] point p，A =[uA v A ,1] T

[0095] Where I and J are the row and column sequence numbers in the region, respectively, V(I,J) is the pixel value in the I-th row and J-th column of the region, and M... 10 M 01 Let u be the two components of the first moment of the image. A v is the x-coordinate of the grayscale center. A The ordinate of the grayscale center;

[0096] Then, the Ransac algorithm is used to utilize the two laser stripe center point sets (points) selected in the second step in each image. p，1，i points p，2，j Then, perform a straight-line fitting again on the center lines of the two laser stripes in the image at intersection point A. The three points used for the straight-line fitting of the center lines of the laser stripes projected by the first laser are: point p，1，i1 point p，1，i2 point p，A The three points used for straight-line fitting of the center line of the laser stripe projected by the second laser are: point p，2，j1 point p，2，j2 point p，A .

[0097] Where point p，1，i1 point p，1，i2 From point set p，1，i Two points randomly selected from the data, point p，2，j1 point p，2，j2 From point set p，2，j Two points were randomly selected from the data.

[0098] After the Ransac line fitting process is completed, the four points randomly extracted from each image that show the best fit are retained, forming four new point sets:

[0099]

[0100] This step allows for further screening of the center points of the laser stripes, preparing for the next step of optical plane calibration.

[0101] S4. Calibrate the camera and use the Ransac algorithm to perform cross-shaped structured light plane fitting from the set of laser stripe center points remaining after the third step of screening and the set of points composed of the intersection points of the intersecting laser stripes in each image.

[0102] The camera was calibrated using Zhang Zhengyou's calibration method. The camera's internal parameters are as follows:

[0103]

[0104] The coordinates of the intersection point A of the laser stripes on the planar target on the camera imaging plane are:

[0105] camera p，A =K -1 ·point p，A =[x c y c ,1] T .

[0106] The coordinates in the camera coordinate system are:

[0107] Camera p，A =Z c ·[x c y c ,1] T =[X c Y c Z c ] T .

[0108] The coordinates in the plane target coordinate system are:

[0109] board p，A =[X A Y A [0, 1] T .

[0110] The pose of the planar target in the p-th image is:

[0111]

[0112] The coordinates of the intersection point A of the laser stripes on the planar target in the camera coordinate system are:

[0113]

[0114] After simplification, we get:

[0115]

[0116] Combining the above two equations, we obtain the position of the intersection point A of the laser stripes on the p-th image in the camera coordinate system. p，A .

[0117] Similarly, we can obtain the point set pr formed by the points extracted when the Ransec algorithm achieves the best line fitting effect in the third step. p，1，1 pr p，1，2 pr p，2，1 prp，2，2 Three-dimensional coordinates in the camera coordinate system:

[0118]

[0119] Then, the Ransac algorithm is used to fit the optical plane of the first laser, from the above point set camera. p，1，1 camera p，1，2 One point is randomly selected from each of the point sets Camera. p，A Two points are randomly selected from (p = 1, 2, ..., P). The optical plane equation of the first laser is obtained by fitting the above four points. The process of randomly selecting four points for plane fitting is repeated, and the result with the best fitting effect is retained as the final result of optical plane fitting. The optical plane of the second laser is obtained in the same way. Finally, the optical plane calibration of the cross-shaped structured light is completed, and the optical plane equations of the two lasers are obtained:

[0120]

[0121] Figure 6 A schematic diagram of the intersection of the center lines of the laser stripes, obtained using image moments, is shown.

[0122] To verify the superiority of the cross-shaped structured light calibration method proposed in this invention, Method 1, which simultaneously uses contour extraction and dilation processing to screen the center points of laser stripes and incorporates the intersection point A of the laser stripes into the line fitting and plane fitting processes, was used as the experimental group. Method 2, which only uses contour extraction and dilation processing to screen the center points of laser stripes, Method 3, which only incorporates the intersection point A of the laser stripes into the line fitting and plane fitting processes, and Method 4, which neither uses contour extraction and dilation processing to screen the center points of laser stripes nor incorporates the intersection point A into the line fitting and plane fitting processes, were used as the control group. A comparative experiment was conducted, using the above four methods to calibrate the cross-shaped structured light, and the calibration errors were compared. Figure 7 The error comparison chart obtained after the control experiment shows that the four bars in each group are Method 1, Method 2, Method 3 and Method 4 from left to right.

[0123]

[0124] The experimental results are shown in the table below:

[0125]

[0126] The experimental results show that the two key steps proposed in this invention—using contour extraction and dilation processing to screen the center points of laser stripes and incorporating the intersection point A of the laser stripes into the straight line fitting and plane fitting processes—can both reduce the timing error of the crosshair cursor. The crosshair cursor timing method based on a planar target provided by this invention can effectively improve the accuracy of crosshair cursor timing.

[0127] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for locating a cross-shaped target based on a planar target, characterized in that, The specific steps are as follows: S1. Image Acquisition: Acquire P groups of planar target images in different poses using a camera and store them. Each group of images includes one planar target image with laser stripes and one planar target image without laser stripes. S2. Extract the center points of the laser stripes from the P images of the planar target with laser stripes and perform the first screening: After processing the two images in each of the P groups of images obtained in step S1, P images are obtained. Using the method of image contour extraction and dilation, the center points of the laser stripes that are close to the edge of the feature points of the planar target are removed from the images. The extracted center points of the laser stripes are screened for the first time, and the center points of the laser stripes located on the contour of the feature points of the planar target in the P images are removed, resulting in two sets of center points of the laser stripes in the P images. S3. Perform a second screening of the center points of the laser stripes by straight line fitting: Use the intersection criterion as the traversal endpoint to traverse the center points of the two intersecting laser stripes in each image after screening in step S2 to obtain the region where the intersection of the two intersecting laser stripes is located. Then, calculate the precise position of the intersection of the two laser stripes by image moments. Use the intersection point and the two center point sets obtained in step S2 as input for straight line fitting. Perform straight line fitting and retain the two points selected when the center line fitting effect of each laser stripe in each image is the best. Then, form four new point sets from the four points retained in each image. Specifically, it includes the following steps: S31. The coordinates of the center points of the laser stripes in the P images after the first screening in step S2 are represented as follows: ； in, Indicates the sequence number of different images. Indicates the first In the image, the first laser projects onto the center point of the i-th laser stripe on the planar target. and They represent The horizontal and vertical coordinates in the image, The total number of center points on the i-th laser stripe; Indicates the first In the image, the second laser projects onto the center point of the j-th laser stripe on the planar target. and They represent The horizontal and vertical coordinates in the image, Let be the total number of center points on the j-th laser stripe, if the line segment and line segments If they intersect, then the following intersection criterion applies: ； Through the , The two center point sequences are searched cyclically until the intersection criterion described above is met, at which point the values ​​of i and j are determined. Then, the preliminary position of the intersection point A of the two laser stripes on the planar target in each image is calculated. : ； in, and They represent The horizontal and vertical coordinates in the image; S32. Calculate the exact coordinates of intersection point A in each image: obtain the preliminary position of intersection point A. After that, with Using a point as the center and W as the width, a region V is extracted from the image. The zeroth-order and first-order image moments of region V are calculated, ultimately yielding the accurate coordinates of the gray-level center of this region, i.e., the intersection point A. ; Zeroth moment: ; First moment: ; ； ； in, , These represent the number of row sequences and the number of column sequences within region V, respectively. It is the first in region V. Line number Column pixel values, and These are the two components of the first-order moment of the image. The x-coordinate of the grayscale center. The ordinate of the grayscale center; S33. Using the Ransac algorithm, utilize the two laser stripe center point sets selected in step S2 in each image. , The exact coordinates of the intersection point A Perform linear fitting on the center lines of the two laser stripes in each image: The three points used for straight-line fitting of the center line of the laser stripe projected by the first laser are: The straight line of the center line of the laser stripe projected by the second laser is simulated. The three points that can be used together are: ; in , From point set Two points randomly selected from the data. , From point set Two points randomly selected from the data; After the Ransac straight line fitting process is completed, the four points randomly extracted from each image with the best fitting effect are retained to form four new laser stripe center point sets for P images. And expressed as: ； in, and Let represent the two points extracted when the center line of the laser stripe projected by the first laser in the p-th image has the best fitting effect. and These represent the two points extracted when the center line of the laser stripe projected by the second laser in the p-th image has the best fitting effect; S4. Perform light plane fitting and complete the light plane calibration of the cross-shaped structured light based on the light plane fitting results: calibrate the camera, and use the Ransac algorithm to perform light plane fitting of the cross-shaped structured light from the four laser stripe center point sets left after the screening in step S3 and the point set composed of the intersection points of the intersecting laser stripes in each image. This specifically includes the following sub-steps: S41. Calibrate the camera. The camera's internal parameters are: ； The coordinates of the intersection point A of the laser stripes on the planar target on the camera imaging plane are: ； The coordinates in the camera coordinate system are: ； The coordinates in the plane target coordinate system are: ； No. The pose of the planar target in the image is: ； The coordinates of the intersection point A of the laser stripes on the planar target in the camera coordinate system are: ； After simplification, we get: ； Combining the above two equations, we obtain the position of the intersection point A of the laser stripes in the p-th image in the camera coordinate system. ; Similarly, the set of four laser stripe center points formed by the points extracted when the Ransec algorithm achieves the best line fitting effect in step S3 is obtained. , , , Three-dimensional coordinates in the camera coordinate system: ； S42. Use the Ransac algorithm to fit the optical plane of the first laser, from the point set... , One point is randomly selected from each of the point sets, and then... Two points are randomly selected from the data. For the point set , One point randomly selected from the set of points and another point randomly selected from the set of points. The optical plane equation of the first laser is obtained by fitting two randomly selected points from the point set. This process is repeated. , One point randomly selected from the set of points and another point randomly selected from the set of points. The process involves randomly selecting two points from the set and performing plane fitting, retaining the result with the best fitting effect as the final result of the light plane fitting; similarly, from the point set... , One point is randomly selected from each of the point sets, and then... Two points are randomly selected from the sample, and the final result of the optical plane fitting of the second laser is obtained according to the above method; finally, the optical plane calibration of the cross-shaped structured light is completed, and the optical plane equations of the two lasers are obtained: 。 2. The cross-shaped target aiming method based on a planar target according to claim 1, characterized in that: Step S1 specifically includes the following steps: S11. Fix the position of the camera and the two line lasers. The camera and the line lasers are fixedly mounted on the same bracket, and the relative position between them is fixed. S12. Fix the planar target within the camera's field of view. Two lasers project lasers onto the planar target to form intersecting laser stripes. The camera takes one image with laser stripes and one image without laser stripes on the planar target when the lasers are on and off, respectively. S13. Remove the planar target and change its pose, then take another picture. Change the pose of the planar target P times and take a total of P sets of planar target images with different poses. Each set of images corresponds to a pose of the planar target. S14. Store the captured P-group planar target images.

3. The cross-shaped target aiming method based on a planar target according to claim 1, characterized in that: Step S2 specifically includes the following steps: S21. Extract the contours of the P planar target images without laser stripes obtained in step S1 to obtain the contour map of the planar target. Dilate the contour map. The width of the convolution kernel after dilation is greater than the width of the convolution kernel when the center of the laser stripe is extracted using the Steger algorithm. S22. Subtract the planar target image with laser stripes from the planar target image without laser stripes in each group of images obtained in step S1 to obtain P images. Remove the influence of ambient light and use the Steger algorithm to extract the center point of the laser stripes in the image after the subtraction operation. The width of the convolution kernel in the Steger algorithm is smaller than the width of the convolution kernel during dilation. S23. Using the dilated planar target feature point contour, the laser stripe center points extracted by the Steger algorithm are filtered out. The laser stripe center points located on the dilated planar target feature point contour in the image set are removed, and the remaining laser stripe center points are retained to obtain two sets of laser stripe center points for P images.

4. The cross-shaped target aiming method based on a planar target according to claim 1, characterized in that: The S41 uses Zhang Zhengyou's calibration method to calibrate the camera.

5. The cross-shaped target aiming method based on a planar target according to claim 1, characterized in that: In step S2, the least squares method is used to perform line fitting to obtain the fitted straight line of the two intersecting laser stripes.

Images