Image matching method and system for a grinding wheel segment

Abstract

The application relates to the technical field of image matching, in particular to an image matching method and system for a grinding wheel piece, which comprises the following steps: analyzing the similarity of feature descriptors between each sampling point and any standard point to determine a similar feature value; analyzing the difference in pixel values between all pixel points in the neighborhood of each sampling point and all pixel points in the neighborhood of any standard point, and the distinctness of all pixel points in the neighborhood of each sampling point and any standard point, to determine a relative intensity difference degree, and combining the similar feature value to determine a matching degree; comparing the difference between the matching degree between each sampling point and any standard point and the maximum matching degree between each sampling point and all standard points to optimize the matching degree and locate the position of a standard grinding wheel in a to-be-matched grinding wheel piece image. The application solves the problem of false matching under complex working conditions of a grinding wheel piece and improves the matching precision of a grinding wheel piece image.

Description

Technical Field

[0001] This application relates to the field of image matching technology, and specifically to an image matching method and system for grinding wheels. Background Technology

[0002] Image matching is a core technology in computer vision and industrial automation inspection. Its goal is to locate the target region in an image to be matched that matches a given template image. In the manufacturing process of grinding wheels, using image matching technology for wheel positioning and defect detection is of great significance for ensuring product quality and improving production efficiency.

[0003] Currently, image matching algorithms based on Best Partner Similarity (BBS) are widely used due to their robustness to non-rigid deformation, partial occlusion, and changes in illumination. However, while widely applied, BBS-based image matching algorithms have two major limitations in complex industrial scenarios such as automated inspection of grinding wheels: First, the algorithm relies on local pixel information, making it difficult to extract effective features when faced with low contrast or background texture interference, leading to decreased matching accuracy. Second, it underutilizes global and local structural information, easily resulting in mismatches or even failures when dealing with scratches, stains, or partial occlusion. Therefore, in summary, traditional image matching algorithms are prone to mismatches, reducing the accuracy of grinding wheel matching. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide an image matching method and system for grinding wheels, and the specific technical solution adopted is as follows: In a first aspect, embodiments of this application provide an image matching method for grinding wheels, the method comprising the following steps: Obtain the image of the grinding wheel to be matched and the image of the standard grinding wheel; Sampling points are selected in each of the preset candidate regions of the grinding wheel image to be matched, and standard points are selected in the region of interest of the standard grinding wheel image. By analyzing the similarity of the feature descriptors between each sampling point and any standard point, the similarity feature values ​​between each sampling point and any standard point are determined. By analyzing the differences in pixel values ​​between all pixels in the neighborhood of each sampling point and all pixels in the neighborhood of any standard point, as well as the significance of all pixels in the neighborhood of each sampling point and any standard point, the relative intensity difference between each sampling point and any standard point is determined, and the matching degree between each sampling point and any standard point is determined by combining the similarity feature values. Based on the difference between the matching degree between each sampling point and all standard points and the maximum matching degree among them, the optimization factor of each sampling point is determined to optimize the matching degree between each sampling point and any standard point. The position of the standard grinding wheel in the image of the grinding wheel to be matched is located based on the optimized matching degree between all sampling points in each candidate region and any standard point.

[0005] Preferably, the size of each candidate region is the same as the size of the region of interest.

[0006] Preferably, the similarity feature value between each sampling point and any standard point is the cosine similarity of the feature descriptors between each sampling point and any standard point.

[0007] Preferably, the expression for the relative intensity difference between each sampling point and any standard point is: In the formula, This indicates the relative intensity difference between sampling point p and standard point b; , These represent the normalized pixel value of the j-th pixel in the neighborhood of sampling point p and the normalized pixel value of the j-th pixel in the neighborhood of standard point b, respectively. This represents the average significance of the j-th pixel in the neighborhood of sampling point p and the j-th pixel in the neighborhood of standard point b. This represents the number of all pixels in the neighborhood of sampling point p.

[0008] Preferably, the matching degree between each sampling point and any standard point is positively correlated with the similarity feature value between each sampling point and any standard point, and negatively correlated with the relative intensity difference.

[0009] Preferably, the optimization factor for each sampling point is the sum and normalized value of the sum of the matching degree between each sampling point and all standard points compared to the largest matching degree difference among them.

[0010] Preferably, optimizing the matching degree between each sampling point and any standard point includes: Optimized matching degree between sampling point p and standard point b The expression is: In the formula, This indicates the degree of matching between sampling point p and standard point b; This represents the optimization factor for sampling point p.

[0011] Preferably, the method for locating the standard grinding wheel in the image of the grinding wheel to be matched is as follows: Based on the optimized matching degree between all sampling points in each candidate region and any standard point, the matching point pairs between each candidate region of the grinding wheel image to be matched and the standard grinding wheel image are obtained; For each candidate region of the grinding wheel image to be matched and the standard grinding wheel image, calculate the sum of the optimized matching degree between all matching points, and record it as the comprehensive matching score of each candidate region; The position of the standard grinding wheel in the image of the grinding wheel to be matched is the candidate region with the highest comprehensive matching score.

[0012] Preferably, the matching point pairs between each candidate region of the grinding wheel image to be matched and the standard grinding wheel image include: Among all sampling points in each candidate region of the grinding wheel image to be matched, the sampling point with the highest matching degree with the standard point b in the standard grinding wheel image is taken as the matching point pair. All standard points are traversed to obtain all matching point pairs.

[0013] Secondly, embodiments of this application also provide an image matching system for grinding wheels, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described image matching methods for grinding wheels.

[0014] This application has at least the following beneficial effects: This application first constructs a multi-dimensional feature descriptor that integrates geometric priors, local texture, and saliency to build similarity feature values, thereby quantifying the macroscopic similarity between sampling points and standard points. This effectively overcomes the interference of low contrast and complex backgrounds, laying the foundation for subsequent accurate matching. Furthermore, this application introduces relative intensity difference to compensate for the neglect of local texture details by macroscopic features, effectively suppressing illumination changes and noise interference. By combining macroscopic similarity feature values ​​and microscopic relative intensity difference, a matching degree that comprehensively evaluates feature consistency and local stability is finally constructed, thus providing a more comprehensive and reliable assessment of the matching potential between sampling points and standard points. Furthermore, this application… This application introduces an optimization factor based on the dispersion of the matching degree distribution to adaptively weight the original matching degree, thereby quantifying the uniqueness and exclusivity of the matching relationship and incorporating it into the final score, which helps to improve the accuracy of the matching decision. Finally, based on the optimized matching degree, this application constructs matching point pairs by finding the best partner for each standard point, and accumulates the optimized matching degree of all point pairs in the candidate region as the comprehensive matching score. Finally, the candidate region with the highest comprehensive matching score is located as the position of the standard grinding wheel, thus transforming the high-quality, high-confidence matching relationship constructed in the previous steps into the final accurate and robust positioning result, improving the matching accuracy of the grinding wheel image. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the steps of an image matching method for grinding wheels provided in one embodiment of this application; Figure 2 This is a schematic diagram of the optimized matching degree extraction process provided in one embodiment of this application. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by this application to achieve the intended inventive purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the image matching method and system for grinding wheels proposed in this application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the image matching method and system for grinding wheels provided in this application.

[0020] Please see Figure 1 The diagram illustrates a flowchart of an image matching method for grinding wheels according to an embodiment of this application, the method comprising the following steps: Step S1: Obtain the image of the grinding wheel to be matched and the image of the standard grinding wheel.

[0021] On an automated grinding wheel production line, photoelectric sensors are used to position the grinding wheels on the conveyor belt. When a grinding wheel enters a preset shooting station, the photoelectric sensor triggers a signal, driving an industrial camera to perform image acquisition. The acquired image is recorded as the grinding wheel image to be matched. To ensure image quality and accuracy of subsequent matching, the camera is configured with its optical axis perpendicular to the surface of the grinding wheel. Simultaneously, a standard grinding wheel image is acquired. After acquisition, for ease of processing, both the acquired grinding wheel image to be matched and the standard grinding wheel image are converted to grayscale, transforming the RGB images into grayscale images. Therefore, the pixel values ​​involved in subsequent processes are grayscale values.

[0022] The image of the grinding wheel to be matched includes not only the grinding wheel itself, but also other information such as the production and processing environment in which it is located. Furthermore, the process of converting the RGB image to grayscale is a well-known technique and will not be described in detail here.

[0023] Step S2: Calculate the initial matching degree by combining similar feature values ​​and relative intensity difference, and use the adaptive optimization factor of the matching degree distribution to weight it to determine the best match between the grinding wheel image to be matched and the standard grinding wheel image.

[0024] Existing image matching methods, such as those based on Best Partner Similarity (BBS), primarily rely on local information. Their inherent limitations are amplified by the superposition of multiple complex working conditions, leading to a sharp decline in matching performance. When the contrast between the grinding wheel and the conveyor belt background is weak, or when the grinding texture itself is difficult to distinguish from the complex background, image matching algorithms struggle to extract stable and reliable matching primitives. Furthermore, when faced with local occlusions such as stains and scratches, a few damaged pixels can overturn the overall matching result, and the inherent circular geometric prior of the grinding wheel cannot be used to resist numerous false matching interferences in the background, severely weakening the robustness of image matching algorithms in practical deployments. Therefore, when multiple adverse factors such as low contrast, complex backgrounds, local occlusion, and environmental interference intertwine, the inherent defects of image matching algorithms reduce matching accuracy.

[0025] Therefore, to improve image matching accuracy, this embodiment calculates the initial matching degree by combining similar feature values ​​and relative intensity difference, and then uses an adaptive optimization factor of the matching degree distribution to weight it, in order to determine the best match between the grinding wheel image to be matched and the standard grinding wheel image. The specific process is as follows: S2.1: Select sampling points in each of the preset candidate regions of the grinding wheel image to be matched, and select standard points in the region of interest of the standard grinding wheel image; determine the similarity feature value between each sampling point and any standard point by analyzing the similarity of the feature descriptors between each sampling point and any standard point.

[0026] First, sampling points are selected in each preset candidate region of the grinding wheel image to be matched, and standard points are selected in the region of interest of the standard grinding wheel image. Specifically: In the standard grinding wheel image, a region of interest (ROI) is marked. Similarly, the grinding wheel image to be matched is divided into multiple regions of the same size as the ROI, denoted as candidate regions. M pixels are selected from the candidate regions using uniform sampling, denoted as sampling points. Likewise, N pixels are selected from the ROI using uniform sampling, denoted as standard points. In this embodiment, M is set to 1000 and N is set to 10. .

[0027] It should be noted that this embodiment is an optimization of the image matching algorithm based on the best partner similarity framework. Setting the value of the number of sampling points M in the candidate region to be much larger than that of the standard points is the core parameter design principle of this image matching algorithm. Specifically, a small number of standard points constitute the standard reference primitives in the high-dimensional feature space. The number of standard points is controlled at a low level to optimize computational complexity and avoid the introduction of redundant information, while the number of sampling points must be large enough to ensure that for each standard point, its potential best pair or partner has a statistically high probability of being included. Moreover, the range of the number of standard points is usually 5 to 50, and the range of the number of sampling points is usually hundreds to thousands. Therefore, in this embodiment, the number of standard points is set to 10 and the number of sampling points is set to 1000.

[0028] Furthermore, this implementation determines the similarity feature value between each sampling point and any standard point by analyzing the similarity of the feature descriptors between each sampling point and any standard point. Specifically: In this embodiment, the cosine similarity of the feature descriptors between each sampling point and any standard point is used as the similarity feature value between each sampling point and any standard point. The calculation method of cosine similarity is a well-known technique and will not be described in detail here.

[0029] To further explain, the process of obtaining the feature descriptors is as follows: First, the Hough circle detection algorithm is used to obtain the center point of the grinding wheel image to be matched. If the center of multiple circles is the same point, then that point is the center point of the grinding wheel. All pixels on the line connecting each sampling point to the center point are recorded as feature points of each sampling point. These feature points effectively characterize the texture information between the center point and the sampling point. The distance from each sampling point to the center point, the LBP value, maximum gradient direction, and saliency of each sampling point, as well as the vector composed of the distance from all feature points to the center point, the LBP value of all feature points, the maximum gradient direction of all feature points, and the saliency of all feature points, are used as the feature descriptors of each sampling point. The saliency is obtained by using the grinding wheel image to be matched as input to the HC saliency detection algorithm, which outputs the saliency value of each pixel in the image to be matched, used to measure the saliency of the pixel.

[0030] The process of obtaining the center point of the grinding wheel in the image of the grinding wheel to be matched using the Hough circle detection algorithm and the process of obtaining the saliency value of each pixel using the HC saliency detection algorithm are both well-known techniques and will not be described in detail here.

[0031] Based on the similarity feature value between each sampling point and any standard point, it can be understood that the similarity feature value is used to characterize the closeness between the sampling point and the standard point on the multi-dimensional feature descriptor composed of geometric relationship, texture, gradient direction and saliency. It reflects the macroscopic consistency of the structural and texture information contained in the sampling point and the standard point. If the similarity feature value between the sampling point and the standard point is larger, it means that the sampling point in the candidate region is more similar to the standard point in the standard grinding wheel image in terms of overall features. This means that the sampling point is more likely to correspond to the position of the standard point in the grinding wheel image to be matched. Conversely, if the similarity feature value between the sampling point and the standard point is smaller, it indicates that the sampling point in the candidate area and the standard point in the standard grinding wheel image are significantly different in overall features. The structural and texture information contained in the two lacks macroscopic consistency, which means that the sampling point is very unlikely to be the correct corresponding position of the standard point and should be regarded as an unreliable candidate and excluded in the matching process.

[0032] Thus, this embodiment constructs a multi-dimensional feature descriptor that integrates geometric priors (circular structure of a grinding wheel), local texture, and saliency to build similarity feature values, thereby quantifying the macroscopic similarity between sampling points and standard points. This effectively overcomes the interference of low contrast and complex backgrounds, laying the foundation for subsequent accurate matching.

[0033] S2.2: By analyzing the differences in pixel values ​​between all pixels in the neighborhood of each sampling point and all pixels in the neighborhood of any standard point, as well as the salience of all pixels in the neighborhood of each sampling point and any standard point, the relative intensity difference between each sampling point and any standard point is determined, and combined with the similarity feature value, the matching degree between each sampling point and any standard point is determined.

[0034] Although similarity feature values ​​effectively capture the overall texture information contained in a sampling point by comparing feature descriptors, they ignore the texture information contained in its local neighborhood. Therefore, this embodiment analyzes the differences in pixel values ​​between all pixels in the neighborhood of each sampling point and all pixels in the neighborhood of any standard point, as well as the salience of all pixels in the neighborhood of each sampling point and any standard point, to determine the relative intensity difference between each sampling point and any standard point. Combined with the similarity feature values, the matching degree between each sampling point and any standard point is determined. The specific process is as follows: First, by analyzing the differences in pixel values ​​between all pixels in the neighborhood of each sampling point and all pixels in the neighborhood of any standard point, as well as the significance of all pixels in the neighborhood of each sampling point and any standard point, the relative intensity difference between each sampling point and any standard point is determined. Specifically: As one implementation method, in this embodiment, the relative intensity difference between sampling point p and standard point b is... The expression is: In the formula, , These represent the normalized pixel value of the j-th pixel in the neighborhood of sampling point p and the normalized pixel value of the j-th pixel in the neighborhood of standard point b, respectively. This represents the average significance of the j-th pixel in the neighborhood of sampling point p and the j-th pixel in the neighborhood of standard point b. This represents the number of all pixels in the neighborhood of sampling point p.

[0035] The significance value is the significance value obtained by the HC significance detection algorithm in S2.1.

[0036] It should be noted that the neighborhood size of the sampling point is the same as that of the standard point. The neighborhood size of the sampling point is 8 neighborhoods centered on the sampling point. Similarly, the neighborhood size of the standard point is also 8 neighborhoods centered on the standard point.

[0037] To clarify, all pixels in the neighborhood of each sampling point are numbered sequentially from left to right and from top to bottom, starting from the top left corner of the neighborhood. All pixels in the neighborhood of the standard point are numbered using the same method.

[0038] Based on the relative intensity difference between each sampling point and any standard point, it can be understood that the relative intensity difference is used to characterize the difference between the textures within the local area of ​​the sampling point and the standard point. If the difference between the pixel values ​​of all pixels in the neighborhood of sampling point p and the pixel values ​​of the corresponding numbered pixels in the neighborhood of standard point b is greater, it indicates that the pixel intensity relationship between the neighborhood of sampling point p and the neighborhood of standard point b is more serious, and the corresponding relative intensity difference is greater. This means that potential texture differences or noise interference weaken the reliability of the matching between sampling point p and standard point b. Furthermore, if the significance of the j-th pixel in the neighborhood of sampling point p is greater than the average significance of all pixels in the neighborhood of standard point b, it indicates that the difference in pixel texture between the neighborhood of sampling point p and the neighborhood of standard point b is more significant, contributing more to the overall relative intensity difference and relatively increasing the attention to the spectrum of sampling points and standard point b. Conversely, the smaller the difference between the pixel values ​​of all pixels in the neighborhood of sampling point p and the pixel values ​​of the corresponding numbered pixels in the neighborhood of standard point b, the higher the pixel intensity relationship between the neighborhood of sampling point p and the neighborhood of standard point b, and the smaller the corresponding relative intensity difference. This means that the two have a high degree of consistency in local texture, which greatly enhances the reliability of the matching between sampling point p and standard point b. Furthermore, the closer the saliency of the j-th pixel in the neighborhood of sampling point p is to the mean saliency of all pixels in the neighborhood of standard point b, the more subtle the difference in pixel texture between the neighborhood of sampling point p and the neighborhood of standard point b, the smaller the contribution to the overall relative intensity difference, and the relatively weaker the influence of this point on the overall matching decision.

[0039] Furthermore, this embodiment determines the matching degree between each sampling point and any standard point based on the relative intensity difference between each sampling point and any standard point, combined with the similarity feature value. Specifically: In this embodiment, the matching degree between each sampling point and any standard point is positively correlated with the similarity feature value between each sampling point and any standard point, and negatively correlated with the relative intensity difference.

[0040] It should be understood that a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. The specific relationship can be additive or multiplicative, etc., and is determined by the actual application. This application does not impose any special restrictions. A negative correlation means that the dependent variable decreases as the independent variable increases, and the dependent variable increases as the independent variable decreases. The relationship can be subtractive or divisive, etc., and is determined by the actual application.

[0041] Preferably, as one implementation method, in this embodiment, the expression for the matching degree between each sampling point and any standard point is: In the formula, This indicates the degree of matching between sampling point p and standard point b; This represents the normalized value of the similarity feature values ​​between sampling point p and standard point b. The normalized value represents the relative intensity difference between sampling point p and standard point b; This indicates a preset constant greater than 0, used to prevent the denominator from being 0. The value is set manually, in this embodiment. The value of is 0.01. Provided that the denominator is not zero and does not excessively affect the calculation result, the implementer may also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0042] It should be noted that there are many commonly used normalization methods. In this embodiment, the similarity feature values ​​between all sampling points and any standard point in each candidate region of the grinding wheel image to be matched are used as the input of the max-min normalization method, and the normalized value of the similarity feature values ​​between all sampling points and any standard point is output. Similarly, the relative intensity difference between all sampling points and any standard point in each candidate region of the grinding wheel image to be matched is used as the input of the max-min normalization method, and the normalized value of the relative intensity difference between all sampling points and any standard point is output. In practical applications, as other implementation methods, implementers may also use other normalization methods such as z-score normalization according to specific circumstances. This embodiment does not impose any special restrictions on the selection of normalization methods.

[0043] The process of normalizing data using the maximum-minimum normalization method is a well-known technique and will not be elaborated further.

[0044] Based on the matching degree between each sampling point and any standard point, it can be understood that the matching degree reflects the comprehensive potential of the sampling point and the standard point to become the best partner. The larger the matching degree, the more similar the sampling point and the standard point are not only in overall features but also in their local texture patterns, making them candidates for high-confidence matching. If the similarity feature value between sampling point p and standard point b is larger, and the relative intensity difference between sampling point p and standard point b is smaller, it indicates that the macroscopic feature similarity between sampling point p and standard point b is higher, contributing more to the formation of a match, and the corresponding matching degree is also larger. Conversely, if the similarity feature value between sampling point p and standard point b is smaller, and the relative intensity difference between sampling point p and standard point b is larger, it indicates that the macroscopic feature similarity between sampling point p and standard point b is lower, and there is also a significant conflict in the local texture patterns. The comprehensive potential for the two to become the best partner is extremely low, making them candidates for low-confidence matching, and the corresponding matching degree is also smaller.

[0045] Thus, this embodiment compensates for the neglect of local texture details by introducing relative intensity difference, effectively suppressing illumination changes and noise interference. By combining macroscopic similarity feature values ​​and microscopic relative intensity difference, a matching degree that can comprehensively evaluate feature consistency and local stability is finally constructed, thereby more comprehensively and reliably assessing the matching potential between sampling points and standard points.

[0046] S2.3: Based on the difference between the matching degree between each sampling point and all standard points and the maximum matching degree among them, determine the optimization factor of each sampling point to optimize the matching degree between each sampling point and any standard point.

[0047] The matching degree calculated in step S2.2 only quantifies the similarity between the sampled point and the standard point, but does not reflect the relative advantage of this similarity among all potential matches. Therefore, in order to improve the certainty of the matching decision, this embodiment determines the optimization factor of each sampled point by comparing the difference between the matching degree between each sampled point and any standard point and the maximum matching degree between each sampled point and all standard points. This optimizes the matching degree between each sampled point and any standard point, thereby suppressing mismatches caused by unclear matching relationships. The specific process is as follows: This embodiment first determines the optimization factor for each sampling point by comparing the matching degree between each sampling point and any standard point with the difference between the maximum matching degree between each sampling point and all standard points. Specifically: In this embodiment, the sum of the matching degree between each sampling point and all standard points and the sum of the differences in the maximum matching degree among them, and the result of normalization, is used as the optimization factor for each sampling point.

[0048] It should be noted that there are many methods to measure the differences between data. In this embodiment, the absolute difference between the matching degree between each sampling point and any standard point and the maximum matching degree between each sampling point and all standard points is taken as the difference between the matching degree between each sampling point and any standard point and the maximum matching degree between each sampling point and all standard points. In practical applications, as other implementation methods, implementers may also use other methods such as the square or ratio of the difference to measure the differences between data, depending on the specific circumstances. This embodiment does not impose any special restrictions on the selection of methods for measuring the differences between data.

[0049] It should be noted that the normalization function of the optimization factor here adopts the maximum-minimum value normalization method, that is, the sum of all sampling points in each candidate region is used as the input of the maximum-minimum value normalization method, and the normalized value of the sum is output. In practical applications, as other implementation methods, implementers may also adopt other normalization processing methods according to specific circumstances. This embodiment does not impose any special restrictions.

[0050] Based on the optimization factor of each sampling point, it can be understood that the optimization factor is used to quantify the dispersion of the matching degree compared with other competitive matching degrees, reflecting the uniqueness and exclusivity of the matching relationship. The optimization factor is affected by the concentration of the matching degree distribution. If the difference between the matching degree of the current sampling point and the standard point corresponding to the maximum matching degree among other standard points is greater than the maximum matching degree, it means that the current sampling point is more likely to be the best matching point of the standard point, and therefore, the corresponding optimization factor is larger. Conversely, if the difference between the matching degree of the current sampling point and the standard point corresponding to the maximum matching degree among other standard points is smaller than the maximum matching degree, it means that the current sampling point is difficult to distinguish from other competitors in terms of matching degree, and the probability of it being the unique best matching point of the standard point is lower. The matching relationship is fuzzy and the exclusivity is weak. Therefore, the corresponding optimization factor is smaller.

[0051] Furthermore, this embodiment utilizes an optimization factor to optimize the matching degree between each sampling point and any standard point. Specifically: As one implementation method, in this embodiment, the optimized matching degree between sampling point p and standard point b is... The expression is: In the formula, This indicates the degree of matching between sampling point p and standard point b; This represents the optimization factor for sampling point p.

[0052] Preferably, the schematic diagram of the optimized matching degree extraction process provided in this embodiment is as follows: Figure 2 As shown.

[0053] Based on the optimized matching degree between each sampling point and any standard point, it can be understood that the optimized matching degree is a combination of the original matching degree and the optimization factor. If the optimization factor between sampling point p and standard point b is larger, and the original matching degree is also larger, the contribution to the optimized matching degree is greater, indicating that the probability of sampling point p and standard point b being the best matching pair is greater. Conversely, if the optimization factor between sampling point p and standard point b is smaller, and the original matching degree is also smaller, the contribution to the optimized matching degree is smaller, indicating that sampling point p and standard point b not only have insufficient original matching potential, but also low certainty of their matching relationship, and the probability of them being the best matching pair is extremely small.

[0054] Thus, this embodiment introduces an optimization factor based on the dispersion of the matching degree distribution to adaptively weight the original matching degree, thereby quantifying the uniqueness and exclusivity of the matching relationship and incorporating it into the final score, which helps to improve the accuracy of matching decisions.

[0055] Step S3: Based on the optimized matching degree between all sampling points in each candidate region and any standard point, locate the position of the standard grinding wheel in the image of the grinding wheel to be matched.

[0056] Based on the optimized matching degree obtained in step S2, the position of the standard grinding wheel in the image of the grinding wheel to be matched is further determined based on the optimized matching degree. Specifically: First, among all sampling points in each candidate region of the grinding wheel image to be matched, the sampling point with the highest matching degree with the standard point b in the standard grinding wheel image is taken as the matching point pair. All standard points are traversed to obtain all matching point pairs.

[0057] Furthermore, between each candidate region of the grinding wheel image to be matched and the standard grinding wheel image, the sum of the optimized matching degree between all matching points is calculated and recorded as the comprehensive matching score of each candidate region. The position of the standard grinding wheel in the image of the grinding wheel to be matched is the candidate region with the highest comprehensive matching score.

[0058] Thus, this embodiment constructs matching point pairs by finding the best partner for each standard point based on the optimized matching degree, and accumulates the optimized matching degree of all point pairs in the candidate region as the comprehensive matching score. Finally, the candidate region with the highest comprehensive matching score is located as the position of the standard grinding wheel, thereby transforming the high-quality, high-confidence matching relationship constructed in the previous steps into the final accurate and robust positioning result, improving the matching accuracy of the grinding wheel image.

[0059] Based on the same inventive concept as the above methods, embodiments of this application also provide an image matching system for grinding wheels, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described image matching methods for grinding wheels.

[0060] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments of this specification have been described above. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0061] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0062] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. An image matching method for grinding wheels, characterized in that, The method includes the following steps: Obtain the image of the grinding wheel to be matched and the image of the standard grinding wheel; Sampling points are selected in each of the preset candidate regions of the grinding wheel image to be matched, and standard points are selected in the region of interest of the standard grinding wheel image. By analyzing the similarity of the feature descriptors between each sampling point and any standard point, the similarity feature values ​​between each sampling point and any standard point are determined. By analyzing the differences in pixel values ​​between all pixels in the neighborhood of each sampling point and all pixels in the neighborhood of any standard point, as well as the significance of all pixels in the neighborhood of each sampling point and any standard point, the relative intensity difference between each sampling point and any standard point is determined, and the matching degree between each sampling point and any standard point is determined by combining the similarity feature values. Based on the difference between the matching degree between each sampling point and all standard points and the maximum matching degree among them, the optimization factor of each sampling point is determined to optimize the matching degree between each sampling point and any standard point. The position of the standard grinding wheel in the image of the grinding wheel to be matched is located based on the optimized matching degree between all sampling points in each candidate region and any standard point.

2. The image matching method for grinding wheels as described in claim 1, characterized in that, The size of each candidate region is preset to be the same as the size of the region of interest.

3. The image matching method for grinding wheels as described in claim 1, characterized in that, The similarity feature value between each sampling point and any standard point is the cosine similarity of the feature descriptors between each sampling point and any standard point.

4. The image matching method for grinding wheels as described in claim 1, characterized in that, The expression for the relative intensity difference between each sampling point and any standard point is: In the formula, This indicates the relative intensity difference between sampling point p and standard point b; , These represent the normalized pixel value of the j-th pixel in the neighborhood of sampling point p and the normalized pixel value of the j-th pixel in the neighborhood of standard point b, respectively. This represents the average significance of the j-th pixel in the neighborhood of sampling point p and the j-th pixel in the neighborhood of standard point b. This represents the number of all pixels in the neighborhood of sampling point p.

5. The image matching method for grinding wheels as described in claim 1, characterized in that, The matching degree between each sampling point and any standard point is positively correlated with the similarity feature value between each sampling point and any standard point, and negatively correlated with the relative intensity difference.

6. The image matching method for grinding wheels as described in claim 1, characterized in that, The optimization factor for each sampling point is the sum and normalized result of the sum of the differences between the matching degree of each sampling point and all standard points and the maximum matching degree among them.

7. The image matching method for grinding wheels as described in claim 1, characterized in that, The optimization of the matching degree between each sampling point and any standard point includes: Optimized matching degree between sampling point p and standard point b The expression is: In the formula, This indicates the degree of matching between sampling point p and standard point b; This represents the optimization factor for sampling point p.

8. The image matching method for grinding wheels as described in claim 1, characterized in that, The method for locating the standard grinding wheel in the image of the grinding wheel to be matched is as follows: Based on the optimized matching degree between all sampling points in each candidate region and any standard point, the matching point pairs between each candidate region of the grinding wheel image to be matched and the standard grinding wheel image are selected. For each candidate region of the grinding wheel image to be matched and the standard grinding wheel image, calculate the sum of the optimized matching degree between all matching points, and record it as the comprehensive matching score of each candidate region; The position of the standard grinding wheel in the image of the grinding wheel to be matched is the candidate region with the highest comprehensive matching score.

9. The image matching method for grinding wheels as described in claim 8, characterized in that, The matching point pairs between each candidate region of the grinding wheel image to be matched and the standard grinding wheel image include: Among all sampling points in each candidate region of the grinding wheel image to be matched, the sampling point with the highest matching degree with the standard point b in the standard grinding wheel image is taken as the matching point pair. All standard points are traversed to obtain all matching point pairs.

10. An image matching system for grinding wheels, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the image matching method for grinding wheels as described in any one of claims 1-9.

Images