Die forging gear surface defect detection method and system based on machine vision

Abstract

The application belongs to the technical field of image recognition, and particularly relates to a die forging gear surface defect detection method and system based on machine vision, which comprises the following steps: collecting a co-polarization image and a cross-polarization image of a surface of a gear gasket to be recognized through an imaging device; calculating a metal surface purity index atlas based on pixel-level differences of the co-polarization image and the cross-polarization image; performing spatial consistency verification and optimization on the metal surface purity index atlas to enhance continuous structures conforming to annular geometric priori and suppress isolated noise points not conforming to the annular geometric priori; and identifying a defect area of the surface of the gear gasket based on the optimized metal surface purity index atlas through an image segmentation algorithm, so as to realize detection of the surface defects of the gear gasket. The application can effectively overcome optical interference of a high-reflective surface and realize accurate detection of the surface defects of the gear gasket.

Description

Technical Field

[0001] This invention relates to the field of image recognition technology. More specifically, this invention relates to a machine vision-based method and system for detecting surface defects in forged gears. Background Technology

[0002] In modern precision manufacturing, especially in automotive transmission systems, aerospace, and high-end equipment, forged gears are core components for transmitting power and achieving precision motion. These gears and their matching precision components, such as gear washers or spacers, typically require precision machining to ensure a high degree of surface flatness and smoothness. The surface quality of these components directly affects the assembly accuracy, operating noise, and final service life of the entire gearbox system. Therefore, rigorous and comprehensive surface defect inspection of these components before leaving the factory, and the rejection of defective products with scratches, rust spots, or oil residue, is a crucial step in ensuring product quality and safe operation.

[0003] Currently, machine vision technology, due to its high efficiency and non-contact advantages, has been widely used to replace traditional manual visual inspection. However, when applying traditional image processing methods to highly reflective metal parts such as forged gears, two problems arise. First, these precision-machined metal surfaces have a high gloss level, which easily produces strong specular reflections under industrial lighting conditions, forming large areas of image saturation highlights. These highlights completely obscure the texture features of the tiny defects beneath. Second, in actual industrial environments, the surfaces of these parts often have residual coolant from the processing or a transparent oil film used for rust prevention.

[0004] Optical interference caused by highlights and oil films lacks clear distinction from real surface defects in image features, making it difficult for traditional algorithms based on grayscale thresholding, edge detection, or template matching to operate stably. This results in high false positive and false negative rates, and the reliability and repeatability of the detection results fail to meet the stringent requirements of precision manufacturing. Therefore, how to effectively separate the optical interference signals from real defect signals on highly reflective metal surfaces and achieve accurate defect detection of high-gloss, oil-film-coated forged gear components is a pressing technical problem that needs to be solved in this field. Summary of the Invention

[0005] To address the technical problem of inaccurate detection of surface defects in forged gear components using existing technologies, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a machine vision-based method for detecting surface defects in forged gears, comprising: acquiring co-polarized and orthogonally polarized images of the surface of a gear shim to be identified using an imaging device; calculating a metal surface purity index spectrum based on the pixel-level differences between the co-polarized and orthogonally polarized images, wherein the region corresponding to the gear shim is initially enhanced in the spectrum; performing spatial consistency verification and optimization on the metal surface purity index spectrum to enhance continuous structures conforming to the ring geometry prior and suppress isolated noise points that do not conform to the ring geometry prior; and identifying defect regions on the surface of the gear shim using an image segmentation algorithm based on the optimized metal surface purity index spectrum, thereby achieving the detection of surface defects on the gear shim.

[0007] This invention suppresses interference from glossy oil film at the physical source through polarized optical imaging, and then combines the annular geometric prior of the parts to spatially verify and enhance the physical features, ensuring that the recognition results have high accuracy and high robustness in complex industrial environments.

[0008] Preferably, the purity index of the metal surface satisfies the following relationship: ;in, Let (i,k) be the purity index of the metal surface at pixel (i,k). and , i, k, represent the gray values ​​of pixel (i, k) in the same polarization image and the orthogonal polarization image, respectively. The maximum gray value of the image with the same polarization direction. To prevent regularization constants with denominators of zero, and It is a non-linear adjustment factor.

[0009] By combining the degree of linear polarization, which characterizes polarization retention, and the reflectivity, which characterizes metallic luster, metal surfaces can be accurately identified based on physical principles, thus effectively improving the identification capability.

[0010] Preferably, the spatial consistency verification and optimization of the metal surface purity index map includes: performing preliminary processing on the metal surface purity index map to estimate the global parameters of the ring geometric prior, the global parameters including the center coordinates and inner and outer radii; calculating the geometric consistency score for each pixel in the metal surface purity index map based on the global parameters; and performing weighted optimization on the metal surface purity index map based on the geometric consistency score.

[0011] Preferably, the geometric consistency score satisfies the following relation: ;in, The geometric consistency score is given for pixel (i,k). This is the set of neighboring pixels selected centered at pixel (i,k) along the tangential direction defined by the global parameters. Let (i,k) be the set of neighboring pixels selected along the radial direction defined by the global parameters, centered at pixel (i,k). This is an index of the purity of the metal surface for neighboring pixels (p, q). To prevent regularization constants with denominators of zero.

[0012] By evaluating the consistency between a pixel and its neighborhood on the desired ring structure, it is possible to distinguish between pixel clusters that constitute the real substrate and isolated noise points with similar physical properties but random spatial locations, which greatly enhances the robustness of detection.

[0013] Preferably, the weighted optimization of the metal surface purity index spectrum based on the geometric consistency score is specifically achieved through the following relationship: ;in, The optimized metal surface purity index. The purity index of the metal surface before optimization. The score is for geometric consistency.

[0014] Preferably, the acquisition of co-polarized and orthogonal polarized images of the surface of the gear shim to be identified by the imaging device includes: acquiring images using an active polarization optical imaging system, the system including a light source, a polarizer placed in front of the light source, an industrial camera, and an analyzer placed in front of the industrial camera; acquiring co-polarized images when the polarization direction of the analyzer is adjusted to be parallel to the polarization direction of the polarizer; and acquiring orthogonal polarized images when the polarization direction of the analyzer is adjusted to be perpendicular to the polarization direction of the polarizer.

[0015] Preferably, the light source illuminates the gear shim surface at an angle near Brewster's angle, and the industrial camera takes pictures perpendicular to the gear shim to maximize the difference in polarization characteristics between metallic mirror reflection and non-metallic scattered light.

[0016] Preferably, the step of identifying the defect area on the surface of the gear shim using an image segmentation algorithm specifically involves: using a threshold segmentation method to segment the optimized metal surface purity index map to generate a binary mask image characterizing the defects on the surface of the gear shim.

[0017] Secondly, the present invention provides a machine vision-based surface defect detection system for forged gears, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned machine vision-based surface defect detection method for forged gears is implemented.

[0018] By adopting the above technical solution, a computer program for detecting surface defects of forged gears based on machine vision is generated and stored in a memory so that it can be loaded and executed by a processor. This allows for the creation of a terminal device based on the memory and processor, making it convenient to use.

[0019] This invention utilizes active polarization imaging technology to separate interference signals from a physical optics perspective, calculates a preliminary purity index for the metal surface, then uses the inherent annular geometric prior knowledge of the gear shim to construct a geometric consistency score, and uses this to verify and optimize the preliminary purity index spectrum, enhance structural features and suppress isolated noise, thereby achieving accurate segmentation of defects.

[0020] Furthermore, the present invention can effectively shield surface interference such as high gloss and oil film, and directly respond to the physical and geometric characteristics of the gear shim surface itself, thereby making the test results more accurate. Attached Figure Description

[0021] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0022] Figure 1 This is a flowchart illustrating a machine vision-based method for detecting surface defects in forged gears according to the present invention.

[0023] Figure 2 This is a schematic diagram illustrating the same-direction polarization of a gear shim;

[0024] Figure 3 This is a schematic diagram illustrating the orthogonal polarization image of a gear shim;

[0025] Figure 4 This is a schematic diagram illustrating the purity index of a metal surface;

[0026] Figure 5 This is a schematic diagram illustrating the graph after calculating the geometric consistency score;

[0027] Figure 6 This is a schematic diagram illustrating the optimized purity index of the metal surface. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] This invention discloses a machine vision-based method for detecting surface defects in forged gears, referring to... Figure 1 This includes steps S1-S4:

[0031] S1. Acquire co-polarized and orthogonally polarized images of the surface of the gear pad to be identified using an imaging device.

[0032] In an optional embodiment, images of forged gear components on an automated production line can be acquired using an active polarization optical imaging system. This system includes a high-resolution industrial camera for capturing images, an LED surface light source for providing uniform illumination, a polarizer positioned in front of the light source, and an analyzer positioned in front of the camera lens. The polarizer generates linearly polarized light in a specific direction to illuminate the gear shim surface, while the analyzer analyzes the polarization state of the light reflected from the gear shim surface. The analyzer is rotatable around the optical axis of the camera lens to change its polarization detection direction.

[0033] In this optional embodiment, to enhance the target signal, a light source can be irradiated onto the substrate of the gear shim under test at a preset tilt angle. This angle is preferably close to the Brewster angle of common dielectric materials such as oil films on the surface, for example, 50 degrees. Simultaneously, the optical axis of the industrial camera lens is adjusted to be perpendicular to the gear shim for shooting. This combination of non-perpendicular illumination and perpendicular shooting can, in principle, maximize the highlighting of the difference in polarization response between the specular reflection light from a pure metal surface and the scattered light from non-metallic materials such as oil films and stains.

[0034] Specifically, when identifying each gear shim, the system automatically completes the following acquisition operations according to a preset program: controlling the analyzer to rotate so that its transmission direction is parallel to that of the polarizer, and acquiring a co-polarized image. This maximizes the reception of specularly reflected light that retains its original polarization state, resulting in the highest brightness in the metal substrate area; controlling the analyzer to rotate 90 degrees so that its transmission direction is perpendicular to that of the polarizer, and acquiring an orthogonally polarized image. In this way, the light reflected from the ideal specular surface is effectively blocked by the analyzer, while some of the light whose polarization state has changed after being scattered by surface oil film, micro-scratches, or dust can pass through the analyzer and be recorded by the camera, making the areas with interference relatively brighter and the metal areas relatively darker.

[0035] like Figure 2 and Figure 3 The images shown are schematic diagrams of the obtained same-direction polarization image and orthogonal polarization image of the gear shim, respectively. Figure 2 The mid-highlight areas are relatively bright, preserving most of the information from the original image, while Figure 3 The mid-highlight areas were effectively suppressed, and the overall image darkened, but due to the depolarization effect of the oil stains, its contours and textures became more obvious.

[0036] In this way, two polarization images with clear physical meanings, one in the same direction and the other orthogonal, can be acquired, providing raw image data containing rich physical information for the subsequent fundamental differentiation between the target substrate and optical interference.

[0037] S2. Based on the pixel-level difference between the same polarization image and the orthogonal polarization image, calculate the purity index spectrum of the metal surface, in which the region corresponding to the gear shim is initially enhanced in the spectrum.

[0038] In an optional embodiment, after obtaining the same-polarization image and the orthogonal polarization image, each pixel can be distinguished by calculating the metal surface purity index spectrum, assigning a high score to the metal surface and a low score to interfering areas such as oil film, highlights, and stains. The metal surface purity index satisfies the following relationship:

[0039]

[0040] in, Let (i,k) be the purity index of the metal surface at pixel (i,k). and , i, k, represent the gray values ​​of pixel (i, k) in the same polarization image and the orthogonal polarization image, respectively. The maximum gray value of the image with the same polarization direction. To prevent regularization constants with denominators of zero, for example, a value of is taken. ; and For example, a non-linear adjustment factor. The value is 2. The value is 3.

[0041] Specifically, the formula for calculating the purity index of a metal surface consists of two multiplicative terms, forming a double-gated logic. The first term is the polarization purity gate, used to calculate the degree of linear polarization, which reflects the degree of preservation of the polarization state of light. For the specular reflection of pure metals, the polarization state is well preserved. much smaller For linear polarization, the degree of polarization is close to 1; however, for scatterers such as oil films, depolarization is significant. and The difference decreases, and the degree of linear polarization decreases significantly; at the same time The linear polarization degree was nonlinearly enhanced, so that even a small decrease in the linear polarization degree would lead to a significant attenuation of the result, thus enhancing the sensitivity to minor oil contamination.

[0042] Furthermore, the second term is the metal reflection gate, which uses the hyperbolic tangent function to determine the normalized same-direction polarization intensity. Only when it is sufficiently bright, exhibiting the proper luster of metal, does this term approach 1. This effectively filters out non-substrate areas in the image that, although highly polarized, are inherently dark. The sensitivity of the door's opening and closing was controlled.

[0043] In an optional embodiment, the metal surface purity index of all pixels can be calculated to generate a metal surface purity index map, in which the gear pad area appears as a high-brightness connected region, while all interfering areas are effectively suppressed. At this time, the area corresponding to the gear pad is initially enhanced in the map.

[0044] like Figure 4 The diagram shown is a schematic representation of the purity index of a metal surface. It can be seen that by calculating the purity index of the metal surface, the substrate ring of the gear gasket can be clearly identified from the complex background and interference. At this time, the highlights and oil stains are effectively suppressed to low values.

[0045] In this way, complex physical optical phenomena can be transformed into feature maps that can be used for subsequent processing, thus achieving the initial filtering of optical interference.

[0046] S3. Spatial consistency verification and optimization of the purity index spectrum of metal surface are performed to enhance the continuous structure that conforms to the ring geometry prior and suppress isolated noise points that do not conform to the ring geometry prior.

[0047] In an optional embodiment, since the calculation of the metal surface purity index is performed in isolation for each pixel, there may be some isolated small noise points with physical properties similar to the substrate, or blurring and breakage at the edge of the gear pad. Therefore, it is necessary to correct the metal surface purity index map.

[0048] Specifically, an adaptive thresholding method, such as the Otsu method, can be used to binarize the purity index map of the metal surface to obtain a coarse substrate mask. Then, a geometric fitting algorithm, such as the stochastic Hough transform or least-squares annular fitting, can be applied to this potentially incomplete mask to estimate the global parameters of the entire ring gear shim. These global parameters include the coordinates of the center and the inner and outer radii.

[0049] Next, the geometric consistency score is calculated for each pixel in the metal surface purity index map using the obtained global parameters. The geometric consistency score satisfies the following relationship:

[0050]

[0051] in, The geometric consistency score is given for pixel (i,k). Let (i,k) be the set of neighboring pixels selected along the tangential direction defined by global parameters, centered at pixel (i,k). Let (i,k) be the set of neighboring pixels selected along a radial direction defined by global parameters, centered at pixel (i,k). This is an index of the purity of the metal surface for neighboring pixels (p, q). To prevent regularization constants with denominators of zero.

[0052] Specifically, for pixel (i,k), its tangential neighborhood is the set of neighboring pixels along the fitted circular direction. Its radial neighborhood is the set of neighboring pixels perpendicular to the direction of the annulus. The geometric consistency score of pixel (i,k) is calculated as the ratio of the average purity of all pixels in the tangential neighborhood of pixel (i,k) to the average purity of all pixels in the radial neighborhood.

[0053] If a pixel is on a circular substrate, its tangential neighbor pixels should also be on the ring, corresponding to high purity, while its radial neighbor pixels are outside the ring, corresponding to low purity, so the ratio will be very large; while for an isolated noise point, the purity of its neighbor pixels in all directions is very low, and the ratio is close to 1.

[0054] like Figure 5The diagram shown is a schematic representation of the graph after calculating the geometric consistency score. It can be seen that the base material ring of the gear shim presents a bright yellow high-scoring area, and the continuous structure of the ring geometric prior is effectively enhanced.

[0055] Furthermore, the original metal surface purity index spectrum can be weighted and optimized using the calculated geometric consistency score, specifically through the following relationship:

[0056]

[0057] in, The optimized metal surface purity index. The purity index of the metal surface before optimization. Geometric consistency score. Weighted optimization can significantly enhance the values ​​of pixels that have both high physical purity and conform to macroscopic geometry, while maintaining or suppressing pixels that do not conform to the geometry.

[0058] like Figure 6 The diagram shown is a schematic representation of the optimized metal surface purity index. It can be seen that the substrate strip of the gear shim is significantly enhanced, forming a uniform and clearly defined substrate ring, and effectively suppressing other interference signals.

[0059] Thus, by introducing macroscopic geometric priors to verify pixel-level physical features, the true substrate structure can be purified, resulting in a more complete and continuous result with stronger resistance to noise.

[0060] S4. Based on the optimized metal surface purity index map, the defect area on the surface of the gear shim is identified by the image segmentation algorithm, thereby realizing the detection of defects on the surface of the gear shim.

[0061] In an optional embodiment, since the signal-to-noise ratio of the optimized metal surface purity index map is already very high, the boundary between the substrate and defects is clear. Therefore, a reasonable global defect threshold can be set to binarize the optimized metal surface purity index map, and pixels with index values ​​lower than the global defect threshold can be identified as defect pixels, thereby generating a binary mask image characterizing the surface defects of the gear shim.

[0062] Furthermore, through standard image processing operations such as contour extraction and geometric moment analysis, various parameters of each defect region can be accurately calculated, such as the centroid coordinates, area, and length of the defect. These high-precision parameters can be directly output to robots, actuators, or quality control systems on the production line.

[0063] Thus, by segmenting the optimized spectrum with a high signal-to-noise ratio, the final identification and location of defects on the surface of gear shims can be completed simply and accurately.

[0064] This invention also discloses a machine vision-based surface defect detection system for forged gears, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a machine vision-based surface defect detection method for forged gears according to the present invention.

[0065] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0066] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

[0067] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A machine vision based surface defect detection method for a die forged gear, the method comprising: The method comprises: acquiring a co-polarization image and a cross-polarization image of the surface of the gear gasket to be identified by an imaging device; calculating a metal surface purity index map based on the pixel-level difference between the co-polarization image and the cross-polarization image, wherein the region corresponding to the gear gasket is preliminarily enhanced in the map; the metal surface purity index satisfies the relationship: wherein, is a metal surface purity index for pixel point (i, k), and are the gray scale values of pixel point (i, k) in the co-polarization image and the cross-polarization image, respectively, is the maximum gray scale value of the co-polarization image, is a regularization constant to prevent the denominator from being zero, and is a non-linear adjustment factor; The relationship for calculating the metal surface purity index is composed of two multiplied terms, forming a double-gated logic; the first term is a polarization purity gate, which is used to calculate the linear polarization degree reflecting the degree of preservation of the polarization state of light; the second term is a metal reflection gate, which judges the normalized co-polarization intensity by using the hyperbolic tangent function; performing spatial consistency verification and optimization on the metal surface purity index map, comprising: preliminarily processing the metal surface purity index map to estimate the global parameters of the annular geometric prior, the global parameters including the center coordinates and the inner and outer radii; calculating a geometric consistency score for each pixel point in the metal surface purity index map based on the global parameters; performing weighted optimization on the metal surface purity index map according to the geometric consistency score to enhance the continuous structure conforming to the annular geometric prior and suppress isolated noise points not conforming to the annular geometric prior; the geometric consistency score satisfies the relationship: wherein, is a geometric consistency score for the pixel point (i, k), is a set of neighborhood pixels centered at the pixel point (i, k) selected along a tangential direction defined by the global parameters, is a set of neighborhood pixels centered at the pixel point (i, k) selected along a radial direction defined by the global parameters, is a metal surface purity indicator for the neighborhood pixel point (p, q), is a regularization constant to prevent the denominator from being zero; Based on the optimized metal surface purity index map, the defect region of the gear gasket surface is identified by an image segmentation algorithm, thereby realizing the detection of the surface defects of the gear gasket.

2. The machine vision based surface defect detection method of drop forged gear as claimed in claim 1 wherein, The weighted optimization of the metal surface purity index map according to the geometric consistency score is realized by the following relationship: wherein, is the metal surface cleanliness index after optimization, is the metal surface cleanliness index before optimization, is the geometric consistency score.

3. The machine vision based surface defect detection method of die-forged gears as claimed in claim 1 wherein, The co-polarization image and the cross-polarization image of the surface of the gear gasket to be identified are acquired by an imaging device, comprising: An active polarized optical imaging system is used for image acquisition, which includes a light source, a polarizer placed in front of the light source, an industrial camera, and an analyzer placed in front of the industrial camera; When the polarization direction of the analyzer is adjusted to be parallel to the polarization direction of the polarizer, the co-polarization image is acquired; When the polarization direction of the analyzer is adjusted to be perpendicular to the polarization direction of the polarizer, the cross-polarization image is acquired.

4. The machine vision-based surface defect detection method for a swage gear according to claim 3, wherein, The light source irradiates the surface of the gear gasket at an angle near the Brewster angle, and the industrial camera is perpendicular to the gear gasket for shooting, so as to maximize the difference in polarization characteristics between the metal mirror reflection and the non-metal scattering light.

5. The machine vision based surface defect detection method of die-forged gears as claimed in claim 1, wherein, The defect region of the gear gasket surface is identified by an image segmentation algorithm, specifically: setting a defect threshold, thresholding the optimized metal surface purity index map, and identifying the pixel region with an index lower than the defect threshold as a defect region.

6. A machine vision based die forged gear surface defect detection system, characterized by, The method comprises: a processor and a memory, the memory storing computer program instructions which, when executed by the processor, implement a machine vision-based die forging gear surface defect detection method according to any one of claims 1-5.

Images