Automobile part size correction method based on image processing

Abstract

The invention relates to the technical field of image data processing, in particular to an automobile part size correction method based on image processing, and the method comprises the steps: obtaining the thickness of a rotating plate, a to-be-detected image of the rotating plate, and a standard image; an improved polynomial interpolation sub-pixel edge positioning algorithm is adopted to extract an edge line in the to-be-detected image, and if deviation exists between the edge line and the edge line of the standard image, correction is carried out; wherein the gray scale sampling direction in the algorithm is composed of a preset gray scale sampling direction and a deviation direction, the deviation direction is obtained through calculation based on a springback factor of an edge pixel point, springback constants of a plurality of preset direction templates and direction angle information, and the springback factor is the ratio of the deformation degree of the edge point to the thickness of the rotating plate; the springback constant represents the deviation angle of the edge point. The problem that an existing algorithm is not high in detection precision is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing. More specifically, the present invention relates to a method for correcting the dimensions of automotive parts based on image processing. Background Art

[0002] In the modern automotive manufacturing and quality control processes, the accurate measurement and correction of part dimensions have become a key link in ensuring assembly accuracy, improving product consistency and reliability. Traditional dimension detection methods mainly rely on manual measurement, contact measurement instruments or single-view image measurement systems. However, these methods have various deficiencies in practical applications. For example, contact measurement will cause wear on the part surface, manual measurement has low efficiency and is easily affected by subjective factors, while single-view image measurement often cannot provide complete geometric information in the face of complex curved surfaces or occlusion situations, resulting in low measurement accuracy and being unable to effectively meet the on-line measurement requirements of complex-shaped automotive parts.

[0003] In the field of automotive part detection, the polynomial interpolation sub-pixel edge localization algorithm is widely used. For example, when detecting the edge contours of high-precision structural parts such as rotating plates, gears, crankshafts, and camshafts, this algorithm can effectively extract the boundary curve and achieve micron-level error control; in complex curved surface parts such as headlight housings and rearview mirror brackets, it can accurately model and correct the dimensions of irregular edges. In addition, this algorithm can also be applied to automated quality control links such as solder joint detection, seal strip fitting position confirmation, and glue application trajectory consistency detection, significantly improving the robustness and practicality of the image measurement system in actual industrial scenarios.

[0004] To improve the edge localization accuracy, various sub-pixel edge localization algorithms have been widely studied in the academic and industrial fields. Among them, the polynomial interpolation method has attracted much attention due to its advantages such as a stable mathematical model, high fitting accuracy, and relatively high calculation efficiency. The polynomial interpolation sub-pixel edge localization algorithm constructs a gray-scale change curve in the neighborhood of pixel-level edge points and uses a polynomial function for fitting, and then solves the exact position of the gray-scale extreme point to obtain edge coordinate information with a higher resolution than the original. However, traditional algorithms usually use a fixed gradient direction for gray-scale sampling and cannot flexibly handle situations where the edge direction changes greatly or there are deformations, which leads to the problem of low detection accuracy. Summary of the Invention

[0005] To solve the problem of low detection accuracy proposed in the above background art, the present invention provides the following solutions.

[0006] The present invention provides a method for correcting the size of automotive parts based on image processing, including: obtaining the thickness of the rotating plate in the automotive parts, the image to be detected of the rotating plate, and the standard image; using an improved sub-pixel edge localization algorithm based on polynomial interpolation to obtain the edge line in the image to be detected. If there is a deviation between the edge line in the image to be detected and the edge line in the standard image, correction is performed; wherein, the improved sub-pixel edge localization algorithm based on polynomial interpolation includes a gray sampling direction, and the gray sampling direction is the sum of a preset gray sampling direction and a deviation direction; the deviation direction is: , is the rebound factor after normalization of the th edge pixel point, is the th preset direction template corresponding to the th edge pixel point in the image to be detected, the rebound constant after normalization, is the th angle direction of the preset direction template, is the total number of preset direction templates, is the th preset gray sampling direction of the th edge pixel point; the rebound factor is the ratio of the deformation degree of the th edge pixel point to the thickness of the rotating plate; the rebound constant characterizes the th offset angle of the edge pixel point.

[0007] Through the improved sub-pixel edge localization algorithm based on polynomial interpolation, the above technical solution can more accurately locate the edge line in the image to be detected and effectively correct it when there is a deviation between the edge line and the standard image. By introducing the adjustment of the gray sampling direction and the introduction of the rebound factor, the accuracy of edge localization is further improved, which can adapt to the deformation and thickness change of the rotating plate, and enhance the adaptability and robustness to complex edges. The calculation of the rebound constant can accurately characterize the edge offset, providing a more detailed analysis of morphological changes, thereby improving the accuracy and reliability of the size correction of automotive parts.

[0008] Further, the rebound constant is specifically: taking the th edge pixel point as the center, obtaining a sequence composed of the gradient amplitudes of a set of continuous pixel points along the gradient direction of the th edge pixel point, using a peak detection algorithm to obtain the peak point in the sequence, and taking the number of pixel points between the th edge pixel point and the peak point in the change sequence as the th edge pixel point's rebound constant.

[0009] The above technical solution obtains the gradient magnitude sequences of multiple consecutive pixel points along the gradient direction of the edge pixel points, and uses the peak detection algorithm to determine the peak points, so as to accurately quantify the springback constant of the edge pixel points. By calculating the number of pixel points between the edge pixel points and the peak points, the slight changes and morphological features of the edge position can be effectively reflected, which provides a reliable basis for further edge positioning and image correction.

[0010] Further, the degree of deformation is specifically: obtaining a preset number of adjacent edge pixel points before and after the th edge pixel point on its edge line, using the curve fitting algorithm to obtain the adjacent fitting curves of the preset number of edge pixel points, and taking the curvature of the adjacent fitting curves as the th edge pixel point's degree of deformation.

[0011] The above technical solution can accurately depict the geometric morphological features of the edge locally by obtaining multiple adjacent pixel points on the edge line and performing curve fitting, and taking the curvature of the fitting curve as the degree of deformation of the edge pixel points, which helps to identify small but significant local deformations in the image. It not only improves the sensitivity to edge morphological changes, but also enhances the accuracy and stability of sub-pixel edge positioning, providing a reliable deformation basis for subsequent deviation direction calculation and image precision correction.

[0012] Further, the image to be detected corresponding to the preset direction template is specifically: performing convolution operations on the image to be detected of the rotating plate respectively using a plurality of convolution kernels with the same size but different directions, where each convolution kernel corresponds to a preset direction template, so as to obtain the images to be detected corresponding to a plurality of preset direction templates.

[0013] The above technical solution can enhance the edge response ability of the image in different directions by performing convolution operations on the image to be detected of the rotating plate respectively using a plurality of convolution kernels with different directions, enabling the edge features in the image to be fully extracted in multiple directions, effectively overcoming the problem that edge extraction in a single direction may be missed or inaccurately recognized, thereby improving the comprehensiveness and robustness of edge detection, and providing a richer and more stable directional information basis for subsequent edge fitting and sub-pixel precise positioning.

[0014] Further, the total number of the preset direction templates is 8, and the corresponding angular directions are 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees.

[0015] Further, a CCD camera or a CMOS camera is used to obtain the image to be detected and the standard image of the rotating plate in the automotive part, and a laser scanning profiler is used to obtain the thickness of the rotating plate in the automotive part.

[0016] Further, perform denoising and grayscale processing on the image to be detected of the rotating plate and the standard image.

[0017] Through the above technical solution, by performing denoising and grayscale processing on the image to be detected of the rotating plate and the standard image, it helps to eliminate the noise interference in the image caused by environmental illumination, sensor interference, or texture details, making the image information more pure and unified. At the same time, it reduces the interference of color channels on edge detection and contour extraction, improves the stability of image processing and the accuracy of edge positioning, thereby providing more reliable basic data for subsequent size correction.

[0018] Further, the denoising is median filtering or Gaussian filtering.

[0019] Further, use an edge detection algorithm to obtain the edge pixel points in the image to be detected of the rotating plate.

[0020] Further, the edge detection algorithm is Sobel edge detection algorithm or Canny edge detection algorithm.

[0021] The beneficial effects of the present invention are as follows: By combining the improved sub-pixel edge localization algorithm technology based on polynomial interpolation, the present invention can accurately correct the size of automotive parts. Especially when dealing with complex components such as rotating plates, it can effectively make up for the deviation of edge positioning in the image. By introducing the calculation of the springback factor and springback constant, considering the part deformation and thickness difference, the accuracy and robustness of the correction are further improved. In addition, through denoising, grayscale processing, edge detection, and the application of a preset direction template, the image quality and edge recognition effect are optimized, making the entire correction process more accurate and reliable, thus significantly improving the manufacturing accuracy and quality control level of automotive parts. Description of the Drawings

[0022] Figure 1 It is a flowchart schematically showing a method for correcting the size of automotive parts based on image processing according to an embodiment of the present invention. Detailed Embodiments

[0023] An embodiment of a method for correcting the size of automotive parts based on image processing.

[0024] As Figure 1 shown, a flowchart of a method for correcting the size of automotive parts based on image processing according to an embodiment of the present invention includes the following steps: S1: Obtain the thickness, the image to be detected, and the corresponding standard image of the rotating plate in the automotive part.

[0025] In a preferred embodiment, to achieve precise quality inspection of the rotating plate at the critical part of automotive parts, first, a high-resolution CCD camera or CMOS camera is used to collect images of the rotating plate in the state to be detected, obtaining the images to be detected. Meanwhile, to ensure the comparability and accuracy of the detection process, a corresponding standard image needs to be obtained in advance. This standard image represents the state of the rotating plate under the conditions of no defects, intact structure, and meeting the factory standards, serving as the reference benchmark for subsequent image comparison and analysis.

[0026] In addition to obtaining image information, to further improve the comprehensiveness and accuracy of the detection, a laser scanning profiler is introduced to synchronously collect the thickness information of the rotating plate. The laser scanning profiler adopts a non-contact measurement principle and can scan and construct the thickness contour maps of the rotating plate along different cross-sections in real time. This thickness information can not only be used to detect whether there are problems such as deformation, wear, depression, or uneven structure in the rotating plate during manufacturing or long-term use, but also be used for multi-modal fusion analysis with the image information, thereby realizing the dual quality assessment of the structure and appearance.

[0027] In another preferred embodiment, to ensure the accuracy and stability of image feature extraction, preprocessing operations need to be performed on both the collected images to be detected and the standard images, specifically including image denoising and grayscale processing. The image denoising process mainly uses median filtering or Gaussian filtering algorithms. Among them, median filtering can effectively suppress the salt-and-pepper noise caused by external interference and maintain the clarity of the image edges, especially suitable for the detection image scenario with high-frequency isolated points; while Gaussian filtering realizes the overall smoothing of the image by constructing a Gaussian kernel function and is suitable for reducing the local fluctuations in the image caused by illumination changes or sensor noise.

[0028] After the denoising process, the image is grayscale processed. Grayscale processing is the process of converting a color image into a single-channel grayscale image, which simplifies the complexity of subsequent image processing, and at the same time highlights the texture features and edge contours of the rotating plate, thereby improving the stability and accuracy of algorithms such as edge detection and region segmentation.

[0029] S2: Use an improved sub-pixel edge localization algorithm based on polynomial interpolation to extract the edge lines in the image to be detected.

[0030] In one embodiment, the improved sub-pixel edge localization algorithm based on polynomial interpolation includes a grayscale sampling direction, and the grayscale sampling direction is the sum of a preset grayscale sampling direction and a deviation direction; The deviation direction of the th edge pixel point in the image to be detected is: , is the normalized springback factor of the th edge pixel point, For the th edge pixel point in the image to be detected corresponding to the th preset direction template, the normalized springback constant, For the th angle direction of the preset direction template, For the total number of preset direction templates, For the th preset gray sampling direction of the edge pixel point; By introducing a gray sampling direction including the deviation direction and comprehensively considering the image response features in multiple directions, the true offset trend of the edge pixel points can be more accurately reflected. Combining the image responses of each preset direction template with the corresponding angle information effectively enhances the sensitivity to local edge changes and improves the accuracy of sub-pixel edge localization. Through the normalization processing of the springback factor and the springback constant, it can also adapt to rotating plates with different thicknesses and deformation degrees, enhancing the adaptability of the algorithm to complex edge structures and further improving the robustness and accuracy of image size correction.

[0031] The total number of the preset direction templates can be 8, and the corresponding angle directions are 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees. Of course, it can also be set according to the actual situation.

[0032] Use an edge detection algorithm to obtain the edge pixel points in the image to be detected of the rotating plate. The edge detection algorithm is the Sobel edge detection algorithm or the Canny edge detection algorithm.

[0033] By introducing a classic edge detection algorithm to extract the edge pixel points in the image to be detected, the contour information of the rotating plate in the image can be effectively highlighted, background interference can be reduced, and it helps to improve the accuracy and stability of subsequent edge localization and deformation analysis operations. Especially in the case of high image complexity or unclear edge transitions, it can improve the clarity and continuity of edge extraction, providing a reliable image basis for achieving high-precision size correction.

[0034] The springback constant is specifically: taking the th edge pixel point as the center, obtaining a sequence composed of the gradient amplitudes of a set of continuous pixel points along the gradient direction of the th edge pixel point, using a peak detection algorithm to obtain the peak point in the sequence, and taking the number of pixel points between the th edge pixel point and the peak point in the change sequence as the springback constant of the th edge pixel point.

[0035] By extracting the gradient magnitudes of a series of consecutive pixel points in the gradient direction of the edge pixel points and using a peak detection algorithm to determine the significant change positions, the distance between the edge pixel points and the strong change region is quantified and used as the springback constant of the pixel points. This processing method can effectively reflect the change intensity and distance relationship of the local edge structure of the image, and helps to capture the edge deviation trend caused by blurring, deformation or interference. Introducing the springback constant as a characterization factor of the offset angle makes the sub-pixel positioning process more sensitive to local detail changes in the image, thereby improving the accuracy and stability of edge correction.

[0036] The degree of deformation is specifically: obtaining a preset number of adjacent edge pixel points before and after the th edge pixel point on its edge line, using a curve fitting algorithm to obtain the adjacent fitting curve of the preset number of edge pixel points, and taking the curvature of the adjacent fitting curve as the th edge pixel point's degree of deformation.

[0037] By selecting edge pixel points and multiple adjacent pixel points before and after them on the edge line and using a curve fitting algorithm to calculate the curvature of the local fitting curve, the local change degree of the edge is effectively quantified, thus accurately characterizing the deformation characteristics of the edge pixel points. The above scheme can more sensitively reflect the edge fluctuations caused by deformation, manufacturing errors or shooting angles, making the subsequent deviation direction calculation more real and reliable. By introducing local geometric features into the edge positioning process, the edge adjustment effect under sub-pixel accuracy is improved.

[0038] The detected image corresponding to the preset direction template is specifically: performing convolution operations on the detected image of the rotating plate using multiple convolution kernels of the same size but different directions, where each convolution kernel corresponds to a preset direction template, so as to obtain detected images corresponding to multiple preset direction templates.

[0039] By introducing multiple convolution kernels with different directions to perform convolution operations on the detected image, the edge response information of the image can be extracted in different directions, thus more comprehensively capturing the local feature changes of the edge of the rotating plate. Compared with the traditional method of only using fixed-direction edge detection, the direction sensitivity and robustness of edge extraction are improved, making the subsequent sub-pixel edge positioning more accurate. Especially in the case of complex textures or illumination changes in the image, the accuracy of positioning and the reliability of comparing the edge line with the standard image can be effectively improved.

[0040] S3: Compare the edge lines in the detected image and the standard image, and perform correction if there is a deviation.

[0041] In one embodiment, first, the to-be-detected image and the standard image of the rotating plate are obtained, and the edge line information is extracted respectively by using an improved sub-pixel edge localization algorithm based on polynomial interpolation. Subsequently, the edge lines in the to-be-detected image are compared with the corresponding edge lines in the standard image point by point. If it is found that there are deviations in the edge contour position or shape between the two, the edge lines in the to-be-detected image are adaptively corrected according to the offset direction and amplitude to make them more conform to the edge contour of the standard image.

[0042] The above solution effectively reduces the recognition errors caused by the image shooting angle, lighting differences or minor deformations of the workpiece, thereby improving the accuracy and reliability of the automotive part dimension detection and ensuring the accuracy requirements of subsequent assembly and quality control processes.

[0043] The solution of the present invention accurately locates the edge of the rotating plate of the automotive part through an improved sub-pixel edge localization algorithm based on polynomial interpolation, and corrects it when there are deviations in the edge, thereby improving the detection accuracy. At the same time, the springback factor and springback constant are used to reflect the deformation and offset, further enhancing the adaptability to the deformation of the rotating plate. The use of different image processing techniques, such as denoising, grayscale processing and edge detection, ensures high-quality image input and effectively improves the reliability of image matching. The overall solution improves the accuracy, robustness and automation level of the detection process, providing strong technical support for the high-precision dimension correction of automotive parts.

[0044] In the description of this specification, the meanings of "a plurality of" and "several" are at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0045] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided only by way of example. Those skilled in the art will think of many changes, alterations and alternative ways without departing from the spirit and idea of the present invention. It should be understood that various alternative solutions to the embodiments of the present invention described herein can be adopted in the process of practicing the present invention.

Claims

1. An automotive part size correction method based on image processing, characterized in that, Including: Obtain the thickness of the rotating plate in the automotive part, the image to be detected of the rotating plate, and the standard image; Use an improved sub-pixel edge localization algorithm based on polynomial interpolation to obtain the edge line in the image to be detected. If there is a deviation between the edge line in the image to be detected and the edge line in the standard image, make a correction; Among them, the improved sub-pixel edge localization algorithm based on polynomial interpolation includes a gray sampling direction, and the gray sampling direction is the sum of a preset gray sampling direction and a deviation direction; The deviation direction is: , is the rebound factor after normalization of the th edge pixel point, is the rebound constant after normalization of the th edge pixel point in the image to be detected corresponding to the th preset direction template, is the angular direction of the th preset direction template, is the total number of preset direction templates, is the preset gray sampling direction of the th edge pixel point; The rebound factor is the ratio of the deformation degree of the th edge pixel point to the thickness of the rotating plate; the rebound constant characterizes the th edge pixel point's offset angle.

2. The method for calibrating the size of automotive parts based on image processing according to claim 1, characterized in that, The springback constant is specifically: Taking the th edge pixel as the center, obtaining a sequence composed of the gradient magnitudes of a set of multiple consecutive pixels along the gradient direction of the th edge pixel, using a peak detection algorithm to obtain the peak points in the sequence, and taking the number of pixels between the th edge pixel and the peak points in the change sequence as the th edge pixel's rebound constant.

3. A method for correcting the size of automotive parts based on image processing according to claim 1, characterized in that, The degree of deformation is specifically: Obtain the preset multiple edge pixels adjacent before and after the edge pixel on its edge line, and use a curve fitting algorithm to obtain the adjacent fitting curve of the preset multiple edge pixels, and use the curvature of the adjacent fitting curve as the deformation degree of the edge pixel.

4. A method for correcting the size of automotive parts based on image processing according to claim 1, characterized in that, The image to be detected corresponding to the preset direction template is specifically: Perform convolution operations on the image to be detected of the rotating plate using multiple convolution kernels of the same size but different directions. Among them, each convolution kernel corresponds to a preset direction template, so as to obtain the images to be detected corresponding to multiple preset direction templates.

5. A method for correcting the size of automotive parts based on image processing according to claim 1, characterized in that, The total number of the preset direction templates is 8, and the corresponding angular directions are 0 degrees, 45 degrees, 90 degrees, 135 degrees, 180 degrees, 225 degrees, 270 degrees, and 315 degrees.

6. The method for correcting the size of automotive parts based on image processing according to claim 1, characterized in that, Use a CCD camera or a CMOS camera to obtain the image to be detected of the rotating plate in the automotive part and the standard image, and use a laser scanning profiler to obtain the thickness of the rotating plate in the automotive part.

7. A method for calibrating the size of automotive parts based on image processing according to claim 1, characterized in that, Perform denoising and grayscale processing on the image to be detected of the rotating plate and the standard image.

8. A method for correcting the size of automotive parts based on image processing according to claim 7, characterized in that, The denoising is median filtering or Gaussian filtering.

9. A method for correcting the size of automotive parts based on image processing according to claim 1, characterized in that, Use an edge detection algorithm to obtain the edge pixel points in the image to be detected of the rotating plate.

10. A method for calibrating the size of automotive parts based on image processing according to claim 9, characterized in that, The edge detection algorithm is the Sobel edge detection algorithm or the Canny edge detection algorithm.

Images