A method for three-dimensional measurement of complex object surface based on multi-line structured light encoding and decoding

Abstract

The application discloses a kind of complex object surface three-dimensional measurement method based on multi-line structured light encoding and decoding.First, the logical encoding of multi-line stripe is carried out by line shift projection scheme.In the decoding process, the stubborn noise points in the center line graph are filtered out by signal-to-noise distribution state.The horizontal distance between the center line point and the centroid of each center line is taken as the basis for sorting, and the non-ideal line stripe in the first center line graph is sorted, and the region segmentation graph is generated, and the middle region of the bilateral adjacent line is divided.The other center line graphs are respectively subjected to "and" logical operation with the region segmentation graph, and the appropriate bias is added to each line according to the encoding logic, the sorting is realized, and the decoding process is completed.The application eliminates the influence of the sorting difficulty caused by the non-ideal line stripe, and can realize the rapid and robust three-dimensional measurement of discontinuous objects in a short time.The interval gradient filtering algorithm and the mean nearest neighbor algorithm can effectively solve the problems of low signal-to-noise ratio of the center line graph and complex sorting of the non-ideal line stripe.

Description

Technical Field

[0001] This invention relates to the field of multi-line structured light encoding and decoding technology, and in particular to a method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding. Background Technology

[0002] Structured light technology, as a key fusion of modern optics and computer vision, is becoming increasingly important. This technology enables non-contact, high-precision 3D measurement by precisely projecting and capturing light information, demonstrating broad application prospects in fields such as industrial manufacturing, biomedicine, and cultural heritage preservation, providing strong technical support for numerous industries. With its unique advantages, structured light technology plays an increasingly vital role in the field of 3D measurement and reconstruction.

[0003] In industrial applications of structured light 3D measurement technology, line structured light methods play a crucial role, with single-line laser structured light and multi-line structured light being particularly common. Single-line laser structured light is favored for its high precision and simple coding; however, it measures only one line at a time, resulting in relatively low efficiency and often requiring the assistance of scanning mobile devices, thus increasing costs. In contrast, multi-line structured light technology significantly improves image processing efficiency and reduces reliance on mobile scanning devices by projecting multiple beams onto the object surface. However, due to the potential for missing or broken lines when projected onto the object surface, these lines may be irregularly segmented into multiple independent parts, greatly increasing the difficulty of line sequencing. Accurate line sequencing is crucial for ensuring measurement accuracy; therefore, improving the accuracy of line sequencing while maintaining measurement speed has become an important research direction for the development of multi-line structured light technology. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding.

[0005] To achieve the above objectives, the technical solution provided by this invention is as follows:

[0006] A method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding is disclosed. The method is implemented through a multi-line structured light system, which includes a projector, a camera, and a computer. The projector is responsible for projecting a preset pattern, the camera synchronously captures images, and the computer analyzes and processes the captured images.

[0007] The methods include:

[0008] The computer performs logical encoding of multi-line stripes, including setting a projection matrix A of size m*n. m*n ;

[0009] Projection matrix A m*n Circularly shift to the right to obtain other projected images;

[0010] The object to be tested is placed above a reference plane, and a projector sequentially projects multi-line fringe structured light, obtained based on the projected image, onto the object. A camera simultaneously captures fringe images with deformed fringes. i i is the number of projected images, and the striped image I i The data is transmitted to a computer for analysis.

[0011] Computer analysis includes:

[0012] Based on stripe image I i Find the extreme points in the horizontal direction and obtain the binary graph I of the center line. ’ i ;

[0013] An interval gradient filtering algorithm based on the signal-to-noise distribution state in the line stripe pattern is used to process the binary image I of the center line. ’ i Noise reduction processing is performed;

[0014] The mean nearest neighbor algorithm is used to refine the centerline binary image I after noise reduction. ’ Sort all the centerlines in 1, including calculating the abscissa of the centroid of each stripe;

[0015] Combining the abscissas of the centroids of each fringe, and based on the binary image of the centerlines I, which has been denoised and sorted for all centerlines. ’ 1. Implement sorting of the remaining unsorted binary centerline maps;

[0016] The center line points of the decoded multi-line stripes are stereo matched with the center line points in the projected image, and the object's depth information is reconstructed by triangulation based on calibration parameters.

[0017] Furthermore, the logical encoding of the multi-line stripes is performed by computer, including:

[0018] Take *a* equally spaced single-pixel straight lines of length *n* as the lines to be projected. The x-coordinate of each line satisfies the following formula:

[0019] u k = (k-1)*Δw+x0,

[0020] (a-1)*(Δw+1)+x0≤n,

[0021] Where k = 1, 2, ..., a, Δw represents the width between lines, x0 represents the x-coordinate of the first line, and u k This represents the x-coordinate of the k-th line.

[0022] Furthermore, the projection matrix A m*n When cyclically shifting to the right to obtain other projected images, the translation amount of the other projected images satisfies the following formula:

[0023] Δα t =t*Δw / T,

[0024] Where t = 1, 2, ..., T-1, T represents the total number of projection maps, and Δα t This indicates that the t-th projected image is relative to A. m*n The cyclic translation amount.

[0025] Furthermore, obtain the binary image I of the centerline. ’ i ,include:

[0026] To extract the center line from the original image, first extract image I... i Grayscale conversion is achieved by calculating I. i The centerline is extracted from the local maxima of the image, and then the background pixels are set to 0 while the centerline pixels are set to 1, thus obtaining the binary image I of the centerline. ’ i The formula is as follows:

[0027]

[0028] Among them, I ’ i (x, y) represents the binary graph I to the center line. ’ i The grayscale value of pixel (x,y) in the binary image I, where grad(x,y) represents the distance from the center line. i The horizontal gradient of a pixel (x,y) is given by the formula grad(x,y)=I i (x,y)-I i Find (x-1, y), I i (x, y) represents the binary graph I to the center line. i The grayscale value of the middle pixel (x,y).

[0029] Furthermore, based on the signal-to-noise distribution state in the line stripe pattern, the interval gradient filtering algorithm is applied to the binary image I of the center line. ’ i Noise reduction processing is performed, including:

[0030] Obtain the binary image of the centerline I ’ i The interval gradient matrix f i (x, y) satisfies the following formula:

[0031] f i (x,y)=[2*I i(x,y)-I i (x-2,y)-I i [x+2,y)] / 2

[0032] Among them, I i (x,y) represents Figure I i The grayscale value of the middle pixel (x,y);

[0033] Binary diagram of the center line I ’ i All non-zero points are used to determine whether they are noise points. If the gradient value of a pixel's interval is less than a set threshold, its grayscale value is set to 1, meaning the pixel is a centerline point; if it is greater than or equal to the set threshold, its grayscale value is set to 0, meaning the pixel is a noise point. The formula used is as follows:

[0034]

[0035] Furthermore, the mean nearest neighbor algorithm is used to refine the binary image I of the centerline after noise reduction. ’ All center lines in 1 are sorted, including:

[0036] A1. Binary image of the centerline after noise reduction (I) ’ Sort some points in 1: Traverse the denoised centerline binary image I by row ’ 1. Filter out rows that contain exactly a center line points. After this filtering, assuming a total of d rows that meet the conditions are found, temporarily classify the a center line points in the filtered rows into the first to the ath lines according to their order.

[0037] A2. Sum the x-coordinates of the sorted points on the same line and divide by the total number of condition rows d to estimate the x-coordinates of the centroids of the a stripes.

[0038] A3. Retain the sorting of some centerline points from step A1, and generate the binary image of the centerline after noise reduction, I. ’ In step 1, the abscissas of the unsorted centerline points (excluding those sorted in step A1) are compared one by one with the estimated abscissas of the centroids of the stripes; the distance between the abscissas of the centerline points and the abscissas of the centroids of each stripe is calculated, that is, the lateral distance between the centerline points and the centroids of each stripe, and the centerline points are classified into the stripes with the closest lateral distance to the centroids of the stripes.

[0039] Furthermore, the mean nearest neighbor algorithm is used to refine the binary image I of the centerline after noise reduction. ’ All center lines in 1 are sorted, including:

[0040] B1. Binary image of the centerline after noise reduction (I) ’Sort some points in 1: Traverse the denoised centerline binary image I by row ’ 1. Filter out rows that contain exactly a center line points. After this filtering, assuming a total of d rows that meet the conditions are found, temporarily classify the a center line points in the filtered rows into the first to the ath lines according to their order.

[0041] B2. Sum the x-coordinates of the sorted points on the same line and divide by the total number of condition rows d to estimate the x-coordinates of the centroids of the a stripes.

[0042] B3. Discard the sorting of some centerline points in step B1, and process the binary image of the centerline after noise reduction. ’ The x-coordinates of all centerline points in step 1 are compared one by one with the estimated x-coordinates of the a-stripe centroids; the distance between the x-coordinates of the centerline points and the x-coordinates of each stripe centroid is calculated, that is, the lateral distance between the centerline points and each stripe centroid, and the centerline points are classified into the stripes with the closest lateral distance to the stripe centroids.

[0043] Furthermore, combining the abscissas of the centroids of each fringe, and based on the denoised binary image of the centerlines after all centerlines have been sorted, I... ’ 1. Implement sorting of the remaining unsorted binary centerline maps, including:

[0044] Using the obtained centroid abscissas of each fringe, the missing parts of the lines are filled in and the ends of the lines are extended to the vertical region of the entire image. The centerline binary image I after noise reduction is then defined. ’ 1. Set the value of all pixels in the region between the k-th line and the (k+1)-th line to k, where k = 1, 2, ..., a-1; and set the value of all pixels in the region to the right of the a-th line to a, thus obtaining the region segmentation map M.

[0045] Unsorted centerline binary image I ’ j j = 2, 3, ..., T, where T represents the total number of projection images. Each projection image is ANDed with M, and then the corresponding bias is added to the number of each line according to the encoding logic to sort all the lines.

[0046] Compared with existing technologies, the principles and advantages of this technical solution are as follows:

[0047] 1. By processing lines by row, the overall problem is divided into local problems, eliminating the sorting difficulties caused by non-ideal lines and stripes. With a small number of projected images, fast and robust 3D measurement of discontinuous objects can be achieved. The proposed interval gradient filtering algorithm and mean nearest neighbor algorithm effectively solve the problems of low signal-to-noise ratio of centerline images and complex sorting of non-ideal lines and stripes.

[0048] 2. Compared with the traditional method of eliminating incorrectly sorted feature points by constraining the field of view, this technical solution can achieve decoding solely through image processing, avoiding the tedious parameter adjustment process and significantly improving the speed of 3D measurement.

[0049] 3. The correlation between pixels in spatial domain measurement is studied, and corresponding constraints are added. These constraints can ensure the correct solution of the positional relationship between feature points in each row, thereby further improving the measurement accuracy.

[0050] 4. Based on the needs of actual application scenarios, this technical solution provides the option of primary or secondary sorting of some centerline points, allowing users to flexibly adjust the emphasis on accuracy and speed as needed. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the services required in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0052] Figure 1 This is a flowchart illustrating the principle of a three-dimensional measurement method for the surface of complex objects based on multi-line structured light encoding and decoding according to the present invention.

[0053] Figure 2 A schematic diagram of the logical encoding of the first four lines in the projected image;

[0054] Figure 3 To capture images simultaneously with the camera;

[0055] Figure 4 This is a centerline plot without noise reduction;

[0056] Figure 5 This is a centerline diagram after noise reduction;

[0057] Figure 6 For some centerline points that meet the conditions;

[0058] Figure 7 A graph showing the estimated abscissa of the centroid;

[0059] Figure 8 A binary image of the centerline to complete the sorting;

[0060] Figure 9 This is a region segmentation map;

[0061] Figure 10 Sort all the lines and stripes in the diagram. Detailed Implementation

[0062] The present invention will be further described below with reference to specific embodiments:

[0063] This embodiment describes a method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding. It is implemented through a multi-line structured light system, which includes a projector, a camera, and a computer. The projector is responsible for projecting a preset pattern, the camera synchronously captures images, and the computer analyzes and processes the captured images.

[0064] like Figure 1 As shown, the specific work steps are as follows:

[0065] S1. Logical encoding of multi-line stripes using a computer:

[0066] First, set the projection matrix A to a size of 720*480. 720*480 Twenty equally spaced single-pixel straight lines, each 480 pixels long, are selected as projection lines. The x-coordinate of each line satisfies the following formula:

[0067] u k = (k-1)*36+1

[0068] (20-1)*(Δw+1)+x0≤480

[0069] Where k = 1, 2, ..., 20, Δw represents the width between lines, x0 represents the x-coordinate of the first line, and u k This represents the x-coordinate of the k-th line.

[0070] S2, Project the matrix A 720*480 By cyclically shifting to the right, other projected images are obtained. The translation amount of these other projected images satisfies the following formula:

[0071] Δα t =t*Δw / T,

[0072] Where t = 1, 2, ..., T-1, T represents the total number of projection maps, and Δα t This indicates that the t-th projected image is relative to A. m*n The cyclic translation amount.

[0073] The specific steps are as follows:

[0074] S2-1, Place A 720*480 All pixels in the image are shifted 6 units to the right in a cyclic manner to obtain projection image B. 720*480 ;

[0075] S2-2, Place A 720*480 All pixels in the image are shifted 12 units to the right in a cyclic manner to obtain the projected image C. 720*480 ;

[0076] S2-3, A 720*480 All pixels in the image are shifted 18 units to the right in a cyclic manner to obtain the projected image D. 720*480 ;

[0077] Following this pattern, a total of 6 projected images A through F are obtained. The first four lines of each projected image are used as a logical encoding diagram, as follows: Figure 2 As shown. Since the last five projection images are all obtained by cyclically shifting the first projection image to the right, and the offset is less than the line width, the subsequent line sorting of the last five images can be performed by calculating the area between each pair of lines in the first image.

[0078] S3. Place the object to be measured above the reference plane. The projector sequentially projects multi-line fringe structured light obtained based on the projected image onto the object. Simultaneously, a camera captures fringe images with deformed fringes. i i =

[0079] 1,2,...,6, and the stripe image I i The data is transferred to a computer for analysis; for ease of presentation, I... i Cropping out parts of the background area, such as Figure 3 As shown;

[0080] Computer analysis includes:

[0081] S3-1, Based on stripe image I i Find the extreme points in the horizontal direction and obtain the binary graph I of the center line. ’ i The specific process is as follows:

[0082] To extract the center line from the original image, first extract image I... i Grayscale conversion is achieved by calculating I. i The centerline is extracted from the local maxima of the image, and then the background pixels are set to 0 while the centerline pixels are set to 1, thus obtaining the binary image I of the centerline. ’ i The formula is as follows:

[0083]

[0084] Among them, I ’ i (x, y) represents the binary graph I to the center line. ’ i The grayscale value of pixel (x,y) in the binary image I, where grad(x,y) represents the distance from the center line. i The horizontal gradient of a pixel (x,y) is given by the formula grad(x,y)=I i (x,y)-I i Find (x-1, y), I i(x, y) represents the binary graph I to the center line. i The grayscale value of the (x, y) pixel. The first binary image I1 representing the center line is used as an example, such as... Figure 4 As shown;

[0085] S3-2. Due to the influence of ambient light and object reflectivity, the centerline map obtained through local maxima will inevitably contain a large amount of noise, which is difficult to eliminate using traditional filtering methods such as Gaussian filtering or mean filtering. Therefore, this step uses an interval gradient filtering algorithm based on the signal-to-noise distribution state in the line stripe map to filter the binary centerline map I. ’ i Noise reduction processing is performed;

[0086] The process for this step is as follows:

[0087] The grayscale value difference is large near the center line, while the grayscale value change is generally smaller near noise points, resulting in lower contrast. Utilizing this characteristic, an interval gradient filter is designed:

[0088] Obtain the binary image of the centerline I ’ i The interval gradient matrix f i (x, y) satisfies the following formula:

[0089] f i (x,y)=[2*I i (x,y)-I i (x-2,y)-I i [x+2,y)] / 2

[0090] Among them, I i (x,y) represents Figure I i The grayscale value of the middle pixel (x,y);

[0091] Binary diagram of the center line I ’ i All non-zero points are used to determine whether they are noise points. If the gradient value of a pixel's interval is less than 120, its grayscale value is set to 1, meaning that the pixel is a center line point; if it is greater than or equal to 120, its grayscale value is set to 0, meaning that the point is a noise point. The formula used is as follows:

[0092]

[0093] Binary image of the center line after noise reduction I ’ i like Figure 5 As shown.

[0094] S3-3, Implement the mean nearest neighbor algorithm to refine the centerline binary image I after noise reduction. ’Sort all the center lines in 1 as follows:

[0095] A1. Binary image of the centerline after noise reduction (I) ’ Sort some points in 1: Traverse the denoised centerline binary image I by row ’ 1. Filter out rows that contain exactly 20 centerline points, such as Figure 6 As shown. After such filtering, assuming a total of d rows satisfying the conditions are found, the 20 centerline points in the filtered rows are temporarily classified into lines 1 to 20 according to their order;

[0096] A2. Sum the x-coordinates of the sorted points on the same line and divide by the total number of condition rows, d, to estimate the x-coordinates of the centroids of the 20 stripes. To facilitate the demonstration of the principle, round the estimated x-coordinates of the stripe centroids and use these integer values ​​to look up matrix I. ’ In line 1, select the pixels with the same column coordinates and light them up, such as... Figure 7 As shown;

[0097] A3. Retain the sorting of some centerline points from step A1, and generate the binary image of the centerline after noise reduction, I. ’ In step 1, the abscissas of the unsorted centerline points (excluding those sorted in step A1) are compared one by one with the estimated abscissas of the centroids of the 20 stripes. The distance between the abscissa of the centerline point and the abscissa of each stripe centroid is calculated, which is the lateral distance between the centerline point and each stripe centroid. The centerline point is then classified into the stripe with the closest lateral distance to the stripe centroid.

[0098] Binary image of centerlines with all centerlines sorted. ’ 1. Figure 8 As shown.

[0099] S3-4. Combining the abscissas of the centroids of each stripe, and based on the binary image I of the centerlines that has been denoised and sorted, ... ’ 1. Implement sorting of the remaining unsorted binary centerline maps. The process includes:

[0100] Using the obtained centroid abscissas of each fringe, the missing parts of the lines are filled in and the ends of the lines are extended to the vertical region of the entire image. The centerline binary image I after noise reduction is then defined. ’ The value of all pixels in the region between the k-th line and the (k+1)-th line is set to k, where k = 1, 2, ..., 19; and the value of all pixels in the region to the right of the 20th line is set to 20, resulting in the region segmentation map M, as shown below. Figure 9 As shown.

[0101] Unsorted centerline binary image I ’ jj = 2, 3, ..., T, where T represents the total number of projection images. Each projection image is then ANDed with M. Finally, according to the encoding logic, the corresponding bias is added to the number of each line to achieve the sorting of all lines. Figure 10 As shown.

[0102] S3-5. The center line points of the decoded multi-line stripes are stereo matched with the center line points in the projected image. Combined with the calibration parameters, the depth information of the object is reconstructed by triangulation.

[0103] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, any changes made in accordance with the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding, which is implemented through a multi-line structured light system, the multi-line structured light system including a projector, a camera, and a computer; the projector is responsible for projecting a preset pattern, the camera synchronously captures images, and the computer analyzes and processes the captured images; characterized in that, include: The computer performs logical encoding of multi-line stripes, including setting a projection matrix A of size m*n. m*n ; Projection matrix A m*n Circularly shift to the right to obtain other projected images; The object to be tested is placed above a reference plane, and a projector sequentially projects multi-line fringe structured light, obtained based on the projected image, onto the object. A camera simultaneously captures fringe images with deformed fringes. i i is the number of projected images, and the striped image I i The data is transmitted to a computer for analysis. Computer analysis includes: Based on stripe image I i Find the extreme points in the horizontal direction and obtain the binary graph I of the center line. ’ i ; An interval gradient filtering algorithm based on the signal-to-noise distribution state in the line stripe pattern is used to process the binary image I of the center line. ’ i Noise reduction processing is performed; The mean nearest neighbor algorithm is used to refine the centerline binary image I after noise reduction. ’ Sort all the centerlines in 1, including calculating the abscissa of the centroid of each stripe; Combining the abscissas of the centroids of each fringe, and based on the binary image of the centerlines I, which has been denoised and sorted for all centerlines. ’ 1. Implement sorting of the remaining unsorted binary centerline maps; The center line points of the decoded multi-line stripes are stereo matched with the center line points in the projected image, and the object depth information is reconstructed by triangulation based on the calibration parameters. Logical encoding of multi-line stripes using a computer includes: Take *a* equally spaced single-pixel straight lines of length *n* as the lines to be projected. The x-coordinate of each line satisfies the following formula: u k =(k-1)*Δw+x0, (a-1)*(Δw+1)+x0≤n, Where k = 1, 2, ..., a, a is the number of lines, Δw represents the width between lines, x0 represents the x-coordinate of the first line, and u k This represents the x-coordinate of the k-th line; Projection matrix A m*n When cyclically shifting to the right to obtain other projected images, the translation amount of the other projected images satisfies the following formula: Da t =t*Δw / T, Where t = 1, 2, ..., T-1, T represents the total number of projection maps, and Δα t This indicates that the t-th projected image is relative to A. m*n The cyclic translation amount.

2. The method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding according to claim 1, characterized in that, Obtain the binary image of the centerline I ’ i ,include: To extract the center line from the original image, first extract image I... i Grayscale conversion is achieved by calculating I. i The centerline is extracted from the local maxima of the image, and then the background pixels are set to 0 while the centerline pixels are set to 1, thus obtaining the binary image I of the centerline. ’ i The formula is as follows: Among them, I ’ i (x, y) represents the binary graph I to the center line. ’ i The grayscale value of pixel (x,y) in the binary image I, where grad(x,y) represents the distance from the center line. i The horizontal gradient of a pixel (x,y) is given by the formula grad(x,y)=I i (x,y)-I i Find (x-1, y), I i (x, y) represents the binary graph I to the center line. i The grayscale value of the pixel (x,y).

3. The method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding according to claim 1, characterized in that, An interval gradient filtering algorithm based on the signal-to-noise distribution state in the line stripe pattern is used to process the binary image I of the center line. ’ i Noise reduction processing is performed, including: Obtain the binary image of the centerline I ’ i The interval gradient matrix f i (x, y) satisfies the following formula: f i (x,y)=[2*I i (x,y)-I i (x-2,y)-I i (x+2,y)] / 2, Among them, I i (x,y) represents Figure I i The grayscale value of the middle pixel (x,y); Binary diagram of the center line I ’ i All non-zero points are used to determine whether they are noise points. If the gradient value of a pixel's interval is less than a set threshold, its grayscale value is set to 1, meaning the pixel is a centerline point; if it is greater than or equal to the set threshold, its grayscale value is set to 0, meaning the pixel is a noise point. The formula used is as follows:

4. The method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding according to claim 1, characterized in that, The mean nearest neighbor algorithm is used to refine the centerline binary image I after noise reduction. ’ All center lines in 1 are sorted, including: A1. Binary image of the centerline after noise reduction (I) ’ Sort some points in 1: Traverse the denoised centerline binary image I by row ’ 1. Filter out rows that contain exactly a center line points. After this filtering, assuming a total of d rows that meet the conditions are found, temporarily classify the a center line points in the filtered rows into the first to the ath lines according to their order. A2. Sum the x-coordinates of the sorted points on the same line and divide by the total number of condition rows d to estimate the x-coordinates of the centroids of the a stripes. A3. Retain the sorting of some centerline points from step A1, and generate the binary image of the centerline after noise reduction, I. ’ In step 1, the abscissas of the unsorted centerline points (excluding those sorted in step A1) are compared one by one with the estimated abscissas of the centroids of the stripes; the distance between the abscissas of the centerline points and the abscissas of the centroids of each stripe is calculated, that is, the lateral distance between the centerline points and the centroids of each stripe, and the centerline points are classified into the stripes with the closest lateral distance to the centroids of the stripes.

5. The method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding according to claim 1, characterized in that, The mean nearest neighbor algorithm is used to refine the centerline binary image I after noise reduction. ’ All center lines in 1 are sorted, including: B1. Binary image of the centerline after noise reduction (I) ’ Sort some points in 1: Traverse the denoised centerline binary image I by row ’ 1. Filter out rows that contain exactly a center line points. After this filtering, assuming a total of d rows that meet the conditions are found, temporarily classify the a center line points in the filtered rows into lines 1 to a in order of priority; a is the number of lines. B2. Sum the x-coordinates of the sorted points on the same line and divide by the total number of condition rows d to estimate the x-coordinates of the centroids of the a stripes. B3. Discard the sorting of some centerline points in step B1, and process the binary image of the centerline after noise reduction. ’ The x-coordinates of all centerline points in step 1 are compared one by one with the estimated x-coordinates of the a-stripe centroids; the distance between the x-coordinates of the centerline points and the x-coordinates of each stripe centroid is calculated, that is, the lateral distance between the centerline points and each stripe centroid, and the centerline points are classified into the stripes with the closest lateral distance to the stripe centroids.

6. The method for three-dimensional measurement of complex object surfaces based on multi-line structured light encoding and decoding according to claim 1, characterized in that, Combining the abscissas of the centroids of each fringe, and based on the binary image of the centerlines after noise reduction, where all centerlines have been sorted. ’ 1. Implement sorting of the remaining unsorted binary centerline maps, including: Using the obtained centroid abscissas of each fringe, the missing parts of the lines are filled in and the ends of the lines are extended to the vertical region of the entire image. The centerline binary image I after noise reduction is then defined. ’ 1. Set the value of all pixels in the region between the k-th line and the (k+1)-th line to k, where k = 1, 2, ..., a-1; and set the value of all pixels in the region to the right of the a-th line to a, thus obtaining the region segmentation map M. Unsorted centerline binary image I ’ j j = 2, 3, ..., T, where T represents the total number of projection images. Each projection image is then subjected to an AND operation with M. Finally, according to the encoding logic, the number of each line is added with the appropriate offset to achieve the sorting of all lines.

Images