A method and device for extracting the center of line structured light stripes in FPGA

By implementing the Steger algorithm on the FPGA platform, combining two-dimensional Gaussian filtering and Hessian matrix solution, fast and high-precision extraction of laser stripe centers is achieved, solving the contradiction between accuracy and speed in the existing technology, and satisfying the laser stripe centers Real-time extraction requirements of the line.

CN114943761BActive Publication Date: 2024-11-15NANJING UNIV OF POSTS & TELECOMM
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210572013.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-05-25
Publication Date
2024-11-15
Estimated Expiration
2042-05-25

AI Technical Summary

Technical Problem

In the prior art, when extracting the center of laser stripes, it is difficult to find a balance between accuracy and speed, resulting in lower extraction accuracy or slower speed, which cannot meet the real-time measurement requirements.

Method used

The fast laser stripe center extraction method based on the Steger algorithm is adopted, and parallel calculation and pipeline structure are used to perform FPGA, and precise center point extraction at the subpixel level is achieved through two-dimensional Gaussian filtering, derivative calculation and Hessian matrix solution.

Benefits of technology

While ensuring high accuracy, it significantly improves processing speed, and can extract the laser stripe centerline in real time, with an error of less than 0.3 pixels, meeting the requirements of high-precision measurement and reducing calculation errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114943761B_ABST
    Figure CN114943761B_ABST
Patent Text Reader

Abstract

The present invention discloses a method and device for extracting the center of a linear structured light stripe in an FPGA. The method comprises: caching image data to obtain pixel points of adjacent rows and columns in the image; performing two-dimensional Gaussian filtering on the image using cached data and Gaussian kernel parameters to obtain a filtered laser stripe image; constructing a Hessian matrix by solving the second-order partial derivative of each pixel in the image to obtain the normal vector of each pixel; using the pixel point where the first-order derivative crosses the zero point and the second-order derivative obtains the minimum value as the judgment condition of the center point, and screening out the linear structured light stripe center point that meets the condition. The present invention can ensure the accuracy and stability of the single-pixel center point extraction and improve the speed of extracting the center point of the linear structured light stripe.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a method and a device for extracting the center of a line structured light stripe in an FPGA, and belongs to the technical field of line structured light three-dimensional vision measurement. Background Art

[0002] Among the optical non-contact measurement methods, line structured light 3D vision measurement technology is widely used in the fields of dimension measurement, 3D shape reconstruction and surface defect detection in industrial production engineering due to its advantages of fast speed, high real-time performance and easy implementation. The principle of line structured light surface shape measurement is to calculate the 3D coordinates of the object in the object space based on the position of the projected structured light imaging on the CCD pixel by inverting the parameters of the calibration system. The center line of the projected line structured light stripe contains all the spatial position information of the object being measured and is the basis of the entire measurement system. Therefore, in actual measurement, it is necessary to improve the accuracy, speed, noise resistance and stability of the line structured light stripe center extraction as much as possible.

[0003] The center extraction of line structured light stripes is divided into two methods: extracting the geometric center and the energy center. The extraction of the geometric center mainly includes the threshold method, the edge method, the skeleton refinement method, etc. Among them, the threshold method and the edge method are simple to implement and have a fast extraction speed. However, the threshold selection in the threshold method is easily disturbed and can only be used for rough center extraction. The center line extracted by the edge method is unstable and has low precision. It can only be used when the image quality is good. The skeleton refinement rule mainly processes binary images. The results obtained usually have many burrs and need to be processed with other morphological operations, resulting in a slow extraction speed. The algorithms for extracting the energy center mainly include the extreme value method, the grayscale centroid method, and the Steger algorithm. Among them, the extreme value method selects the point with the largest grayscale value of the laser stripe cross section as the stripe center point. This method has a fast operation speed, but the extraction accuracy is low and it is not suitable for images with low signal-to-noise ratio. The grayscale centroid method obtains the center point of the laser stripe by weighting the grayscale value of the column coordinates of the pixel point of the cross section of the laser stripe image and then performing a division operation. This method is fast and accurate, but has poor noise resistance and is only suitable for occasions with uniform light bands and consistent widths. According to the Gaussian distribution characteristics of laser stripes, the Steger algorithm takes the point where the first-order derivative passes through zero and the second-order derivative reaches the minimum value as the center point of the laser stripe. The Steger algorithm has high extraction accuracy and good noise resistance, but the calculation is large and cannot extract the center of the laser stripe in real time. Reference [Yang Zhenhao, Yang Liu, et al. Structured light centerline extraction algorithm under complex environment [J]. Chinese Laser, 28(9):7-10, 2018] uses the extreme value method to quickly obtain the rough position and region of interest (ROI) of the laser stripe, and then uses the Hessian matrix method in the ROI area to extract the sub-pixel position of the laser stripe in the image. Although creating an ROI area reduces the amount of convolution calculation, a fixed area size ROI still cannot avoid the non-striped area from participating in the calculation. Zhou Fuqiang, Zhang Guangjun, Jiang Jie, and Hu Kun invented the patent number CN 200510123724.6, "A method for fast and high-precision extraction of the center of structured light stripes", which uses the Hessian matrix to determine the normal direction of the light stripe, and uses Taylor expansion and recursive methods to solve the sub-pixel center position of the light stripe, reducing the amount of calculation while ensuring accuracy and robustness. However, due to the limitations of the operating platform and the execution structure of the algorithm itself, it is still impossible to achieve real-time extraction of the center line of the laser stripe. Summary of the invention

[0004] The problem to be solved by the present invention is the contradiction between the accuracy and speed of extracting the center line of laser stripes in the process of measuring an object using the triangulation principle and extracting the center of laser stripes. A method for quickly extracting the center line of laser stripes based on the Steger algorithm is invented, which improves the processing speed while ensuring the extraction accuracy.

[0005] The technical solution adopted by the present invention is a fast laser stripe center extraction method based on Steger. Using FPGA as an algorithm implementation platform, multiple pixel points can be processed in parallel to improve the algorithm operation speed. The extraction method first performs preprocessing operations on the data collected by the CMOS image sensor. Then the preprocessed data is convolved with the 0th, 1st, and 2nd order derivatives of the Gaussian kernel to obtain the first-order derivative and second-order partial derivative of the pixel. Then the Hessian matrix is ​​generated based on the partial derivatives to obtain the maximum eigenvalue and corresponding eigenvector of each pixel. Finally, each pixel point is judged according to the specific constraints of the algorithm to find the center point of the laser stripe.

[0006] In a first aspect, a method for extracting the center of a line structured light stripe is provided, comprising:

[0007] Acquire line structured light stripe image;

[0008] Preprocessing the line structured light stripe image;

[0009] By using the decomposable characteristics of the two-dimensional Gaussian function, the preprocessed line structured light stripe image is solved to obtain the first-order derivative and second-order partial derivative of the image;

[0010] The Hessian matrix is ​​established based on the second-order partial derivatives to solve the eigenvector of the image pixels;

[0011] Based on the eigenvector, the sub-pixel center point is solved according to the judgment condition that the pixel point where the first-order derivative crosses zero and the second-order derivative obtains the minimum value is the center point, which is the center of the line structured light stripe.

[0012] Preprocess the line structured light stripe image, including:

[0013] By caching the image rows, a data cache window with the same size as the two-dimensional Gaussian kernel is formed;

[0014] Two-dimensional Gaussian filtering is used to remove Gaussian white noise and random noise contained in the line structured light stripe image, thereby obtaining a high-quality line structured light stripe image.

[0015] In some embodiments, two-dimensional Gaussian filtering includes:

[0016] h(x,y)=f(x,y)*g(x,y)=f(x,y)*g(x)*g(y)

[0017] Among them, h(x,y) is the grayscale value of the image after Gaussian filtering, f(x,y) is the grayscale value at the coordinate (x,y) in the image, g(x,y) is the two-dimensional Gaussian function at (x,y), and g(x) and g(y) are two one-dimensional Gaussian functions decomposed from the two-dimensional Gaussian function.

[0018] In some embodiments, the decomposable characteristics of the two-dimensional Gaussian function are used to solve the preprocessed line structured light stripe image to obtain the first-order derivative and the second-order partial derivative of the image, including:

[0019]

[0020] Among them, h′ x (x,y),h′ y (x,y) is the first-order derivative of the image, h″ xx (x,y),h″ xy (x,y),h′ yy (x, y) is the second-order partial derivative of the image, g′(x), g″(x), g′(y), and g″(y) are the first and second-order derivatives of g(x) and g(y), respectively.

[0021] In some embodiments, establishing a Hessian matrix based on the second-order partial derivatives and solving the eigenvector of the image pixel includes:

[0022] Use the second-order partial derivatives to build the Hessian matrix;

[0023]

[0024] Expand the Hessian matrix to find the eigenvalue λ and eigenvector (n x ,n y ):

[0025]

[0026]

[0027] Among them, (n x ,n y ) is the feature vector at (x,y) in the image.

[0028] In some embodiments, according to the judgment condition that the pixel point where the first-order derivative crosses zero and the second-order derivative obtains a minimum value is the center point, solving the sub-pixel center point includes:

[0029] Define the eigenvector corresponding to the Hessian matrix of each pixel as the normal vector of the pixel, perform a two-dimensional Taylor expansion on h(x,y) along the normal vector direction to find the coordinate point closest to the first-order zero-crossing point, h(x,y) along the eigenvector (n x ,n y ) direction is:

[0030]

[0031] Among them, t is the characteristic vector (nx .n y )’s weight coefficient;

[0032] Solve the specific constraint parameters in the Steger algorithm and let the first-order derivative have to:

[0033]

[0034] The corresponding second-order derivative h″ t (x,y) is solved as follows;

[0035]

[0036] Sub-pixel center point determination is based on the preset second-order derivative threshold h″ thres , filter out the second-order derivative h″ t (x,y) is greater than the threshold h″ thres The non-candidate line structured light stripe center point is found; from each discrete pixel point, find the pixel point closest to the first-order zero-crossing point, that is, satisfy The sub-pixel center coordinates of the two-dimensional image at (x, y) in the image are: (x0+tn x ,y0+tn y ), where (x0, y0) is the coordinate of the current pixel point that meets the conditions.

[0037] In a second aspect, the present invention provides a device for extracting the center of a line structured light stripe, including a processor and a storage medium;

[0038] The storage medium is used to store instructions;

[0039] The processor is used to operate according to the instructions to execute the steps of the method according to the first aspect.

[0040] In a third aspect, the present invention provides a storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect.

[0041] In a fourth aspect, a line structured light stripe center extraction system includes the line structured light stripe center extraction device described in the second aspect.

[0042] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0043] The present invention aims at the problem of accuracy and speed of extracting line structure laser stripes, adopts the Steger algorithm and FPGA computing platform, and takes advantage of FPGA parallel computing and pipeline structure to maximize the algorithm operation speed. In addition, the use of multi-bit fixed-point numbers in FPGA ensures the accuracy of the algorithm. Compared with the Steger algorithm implemented by software, the calculation error accuracy of the present invention and software is less than 0.3 pixels, and the extraction speed is greatly improved compared with the software, which can meet the requirements of real-time extraction of the center line of laser stripes. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is: Line structured light three-dimensional measurement system model;

[0045] Figure 2 For: Steger algorithm solution process;

[0046] Figure 3 For: Comparison before and after Gaussian filtering;

[0047] Figure 4 For: FPGA differential convolution structure transformation;

[0048] Figure 5 For: The center line of the laser stripe extracted by FPGA;

[0049] Figure 6 Figure 2: Error distribution of centerline extraction comparison between FPGA hardware and software. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific methods.

[0051] In the description of the present invention, "several" means more than one, "many" means more than two, "greater than", "less than", "exceed" etc. are understood to exclude the number itself, and "above", "below", "within" etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0052] In the description of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0053] Example 1

[0054] A method for extracting the center of a line structured light stripe, comprising:

[0055] Acquire line structured light stripe image;

[0056] Preprocessing the line structured light stripe image;

[0057] By using the decomposable characteristics of the two-dimensional Gaussian function, the preprocessed line structured light stripe image is solved to obtain the first-order derivative and second-order partial derivative of the image;

[0058] The Hessian matrix is ​​established based on the second-order partial derivatives to solve the eigenvector of the image pixels;

[0059] Based on the eigenvector, the sub-pixel center point is solved according to the judgment condition that the pixel point where the first-order derivative crosses zero and the second-order derivative obtains the minimum value is the center point, which is the center of the line structured light stripe.

[0060] In some embodiments, the line structured light is exemplified by laser.

[0061] In some embodiments, Figure 2 The specific steps are as follows:

[0062] (1) Cache image data and Gaussian kernel parameters.

[0063] Before filtering, it is necessary to generate Gaussian kernel parameters according to the size of the Gaussian kernel, cache the corresponding image rows, and cache the image rows and Gaussian kernel parameters into the corresponding RAM or ROM.

[0064] (2) Perform a two-dimensional Gaussian filtering operation based on the image rows and Gaussian kernel parameters cached in step (1).

[0065] For the structure of two-dimensional Gaussian filtering, if the two-dimensional Gaussian parameter kernel is directly convolved with the image, for a Gaussian kernel of size N×N, it will consume N 2 Due to the limited DSP resources inside the FPGA, it is necessary to minimize the use of DSP when performing algorithm mapping. Therefore, the decomposability of the two-dimensional Gaussian function is used to decompose the two-dimensional Gaussian function into the form of multiplication of two one-dimensional Gaussian functions. The definition of one-dimensional Gaussian convolution is as follows:

[0066]

[0067]

[0068] Combining formula (1) and formula (2), when the image f(x, y) is convolved with a two-dimensional Gaussian kernel, the convolution can be rewritten as:

[0069] h(x,y)=f(x,y)*g(x,y)=f(x,y)*g(x)*g(y) (3)

[0070] Among them, h(x,y) is the grayscale value of the image after Gaussian filtering, f(x,y) is the grayscale value at the coordinate (x,y) in the image, g(x,y) is the two-dimensional Gaussian function at (x,y), and g(x) and g(y) are two one-dimensional Gaussian functions decomposed from the two-dimensional Gaussian function.

[0071] According to formula (3), the convolution of an image with a two-dimensional Gaussian kernel is decomposed into the image first undergoing row convolution with the same dimensional Gaussian kernel, and then undergoing column convolution with the same dimensional Gaussian kernel. Compared with two-dimensional Gaussian convolution, when two one-dimensional Gaussian kernels of size N are used to convolve an image, only 2N multipliers and adders are consumed, which reduces DSP resources by 0.5N times compared with two-dimensional Gaussian convolution. The contrast image of light streaks before and after filtering is shown in Figure 3 As shown in the figure, it can be clearly observed that the light stripe part of the image is more blurred than before filtering, and the image brightness is reduced. After Gaussian filtering, the maximum grayscale value is reduced, but the column-wise pixel distribution is more uniform, and the Gaussian curve formed is smoother.

[0072] (3) After image filtering, solve the first-order derivative h′ x (x,y),h′ y (x,y) and the second-order partial derivative h″ xx (x,y), h″ xy (x,y),h″ yy (x,y).

[0073] The calculation method of Gaussian differential convolution module is similar to that of two-dimensional Gaussian filtering. Both methods first use the decomposable characteristics of two-dimensional Gaussian function to differentiate the convolution function. According to the property that the order of convolution differentiation can be exchanged, the solution structure of the first-order derivative and the second-order partial derivative is adjusted. The adjusted structure is as follows: Figure 4 As shown in the figure, after the adjustment, the number of first-order Gaussian function modules consumed by the convolution is reduced from 10 to 8. For the 5 parameters that need to be solved, the structure of each parameter solution is the same, and the structure adopts parallel design.

[0074] (4) Using the second-order partial derivatives obtained in step (3), construct the Hessian matrix and solve for the maximum eigenvalue λ and the corresponding eigenvector (n x ,n y ).

[0075]

[0076]

[0077] According to formula (4) and formula (5), the circuit characteristics of FPGA are used to make certain adjustments to the integer operation structure and order. For example, in formula (4), h xx +h yy 、h xx -h yy , The three parts can be calculated in parallel at the same time, but the adder and multiplier consume different clock cycles during calculation, and the calculation results need to be synchronized after the operation. The structure uses a serial pipeline structure, and the calculation delay only occurs in the first pixel. After that, each clock cycle can complete the solution of the eigenvalue and eigenvector of the Hessian matrix corresponding to a pixel.

[0078] (5) Sub-pixel center point determination. There are two principles for determining the center point of the laser stripe. First, Then the second-order derivative h″ t (x, y) must be less than a certain threshold. The t value is used as the weight coefficient of the feature vector to determine whether the pixel point falls within the 0.5 pixel size range around the current pixel point (x0, y0). The solution method of the t value is shown in formula (6). When calculating, h′ x n x , h′ y n y , n x n y , The calculation is performed simultaneously. Some multiplication operations in the formula directly use shift operations to reduce the consumption of internal DSP multiplier resources.

[0079]

[0080] (6) Sub-pixel center point output. Step (5) determines whether the corresponding pixel point is the center point. The final sub-pixel coordinates of the center of the laser stripe are (x0+tn x ,y0+tn y ), in order to facilitate other devices to obtain the center point for processing, the center line data bit width output in the X-axis and Y-axis directions is 16 bits, of which the upper 10 bits are integer bits and the lower 6 bits are decimal bits.

[0081] Figure 5 This is the actual effect diagram of the laser stripe extraction actually measured.

[0082] Figure 6According to the difference between the software and hardware results, the error distribution diagram of the software and hardware comparison was drawn. Compared with the software, the error distribution of the Steger algorithm hardware extraction results is 0.2 to 0.5 pixels.

[0083] The software operating environment of the present invention is that the computer configuration is Intel Core i5-9300H CPU @ 2.4GHz, the memory is 16G, the operating system is Window 10, the compilation tool is Microsoft Visual Studio, and the line laser extraction method of the present invention is adopted, and the time consumed is 2.363 seconds. In addition, on the FPGA hardware platform, the main control used is the ZYNQ7000 processor, and the execution time is 0.0025 seconds.

[0084] According to the above experimental results, it can be seen that the present invention implements the more complex centerline extraction algorithm through FPGA, so that the overall extraction speed of the algorithm is greatly improved. In terms of extraction accuracy, the standard deviation of the coordinates of the center point of the laser stripes finally extracted is less than 0.2 pixels, which can well meet the requirements of high-precision measurement. From the perspective of versatility, the Steger algorithm selected by the present invention has good anti-noise and anti-interference capabilities, and the extraction accuracy is less affected by the shape of the object. Therefore, the present invention can have a faster extraction speed while ensuring higher measurement accuracy, and has good versatility and strong anti-interference ability.

[0085] Example 2

[0086] In a second aspect, this embodiment provides a device for extracting the center of a line structured light stripe, including a processor and a storage medium;

[0087] The storage medium is used to store instructions;

[0088] The processor is used to operate according to the instructions to execute the steps of the method according to embodiment 1.

[0089] Example 3

[0090] In a third aspect, this embodiment provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in Example 1 are implemented.

[0091] Example 4

[0092] A line structured light stripe center extraction system, comprising the line structured light stripe center extraction device, and further comprising:

[0093] The image acquisition module is used to acquire the line structured light stripe image and upload it to the line structured light stripe center extraction device; in this embodiment, the image acquisition module adopts a 3D camera.

[0094] The line structured light providing module, in this embodiment, uses a laser.

[0095] In some specific embodiments, Figure 1 This is a schematic diagram of the measurement system. The entire system is designed based on laser triangulation. It uses an ON Semiconductor Python300 CMOS industrial sensor with a resolution of 600 (H) × 480 (V). The lens selected is a million-level camera with a focal length of 8mm. The main control of the camera is a XILINX ZYNQ7000 processor. The laser is a straight-line laser with a wavelength of 520nm and a power of 100mW. In order to achieve dynamic scanning of objects, a movable platform is added.

[0096] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0097] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0098] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for extracting the center of a line structured light stripe, characterized in that: include: Acquire line structured light stripe image; Preprocess the line structured light stripe image, including: By caching the image rows, a data cache window with the same size as the two-dimensional Gaussian kernel is formed; The Gaussian white noise and random noise contained in the line structured light stripe image are removed by using a two-dimensional Gaussian filter, wherein the two-dimensional Gaussian filter comprises: h(x,y)=f(x,y)*g(x,y)=f(x,y)*g(x)*g(y) Among them, h(x,y) is the gray value of the image after Gaussian filtering, f(x,y) is the gray value at the coordinate (x,y) in the image, g(x,y) is the two-dimensional Gaussian function at (x,y), g(x) and g(y) are two one-dimensional Gaussian functions decomposed from the two-dimensional Gaussian function; By using the decomposable characteristics of the two-dimensional Gaussian function, the preprocessed line structured light stripe image is solved to obtain the first-order derivative and second-order partial derivative of the image, including: Among them, h ′ x (x,y),h ′ y (x,y) is the first-order derivative of the image, h ′ x ′ x (x,y),h ′ x ′ y (x,y),h ′ yy (x,y) is the second-order partial derivative of the image, g ′ (x), g″(x), g ′ g(y) and g″(y) are the first and second order derivatives of g(x) and g(y), respectively; The Hessian matrix is ​​established based on the second-order partial derivatives to solve the eigenvector of the image pixels, including: Use the second-order partial derivatives to build the Hessian matrix; Expand the Hessian matrix to find the eigenvalue λ and eigenvector (n x ,n y ): Among them, (n x ,n y ) is the feature vector at (x,y) in the image; Based on the eigenvector, according to the judgment condition that the pixel point where the first-order derivative crosses zero and the second-order derivative obtains the minimum value is the center point, the sub-pixel center point is solved, that is, the center of the line structured light stripe, specifically including: Define the eigenvector corresponding to the Hessian matrix of each pixel as the normal vector of the pixel, perform a two-dimensional Taylor expansion on h(x,y) along the normal vector direction to find the coordinate point closest to the first-order zero-crossing point, and h(x,y) along the eigenvector (n x ,n y ) direction is: Among them, t is the characteristic vector (n x .n y )’s weight coefficient; Solve the specific constraint parameters in the Steger algorithm and let the first-order derivative have to: The corresponding second-order derivative h ′ t ′ (x,y) is solved as follows; Sub-pixel center point determination is based on the preset second-order derivative threshold h ′ t ′ hres , filter out the second-order derivative h ′ t ′ (x,y) is greater than the threshold h ′ t ′ hres The non-candidate line structured light stripe center point is found; from each discrete pixel point, find the pixel point closest to the first-order zero-crossing point, that is, satisfy The sub-pixel center coordinates of the two-dimensional image at (x, y) in the image are: (x0+tn x ,y0+tn y ), where (x0, y0) is the coordinate of the current pixel point that meets the conditions.

2. A device for extracting the center of a line structured light stripe, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to claim 1 .

3. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to claim 1 are implemented.

4. A line structured light stripe center extraction system, characterized in that: It includes the line structured light stripe center extraction device as described in claim 2.