Three-dimensional contour defect detection method of contact instrument based on texture feature analysis

Abstract

The present application provides a contact instrument three-dimensional contour defect detection method based on texture feature analysis, which relates to the field of image processing technology, including: fitting according to the mapping relationship between the image and the three-dimensional contour coordinates to construct the image distribution coordinates of the three-dimensional contour; performing texture recognition on the distribution image of the three-dimensional contour through a filter group to obtain texture features and their distribution coordinates; identifying and separating defect areas according to texture features to obtain defect recognition areas, evaluating the defect impact of the defect recognition areas according to the distribution coordinates, and feeding back the defect detection results. This application can solve the technical problems that traditional defect detection methods have weak ability to accurately extract and analyze surface textures, and it is difficult to accurately locate defects on complex three-dimensional surfaces, resulting in insufficient accuracy, comprehensiveness and effectiveness of defect detection results. It can achieve the technical effect of accurately locating defect areas and effectively evaluating the degree of defect impact.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a three-dimensional contour defect detection method of a contact instrument based on texture feature analysis. Background Art

[0002] At present, the contact instrument three-dimensional contour defect detection method has been widely used in the field of industrial inspection. This type of method usually relies on the contact between the probe and the surface of the object to be tested to obtain the three-dimensional geometric shape and contour information of the object surface. However, although this method shows high accuracy in obtaining the geometric shape of the object, it has significant limitations in the extraction and analysis capabilities of surface texture features, and it is difficult to effectively identify tiny surface defects, especially in inspection scenarios involving high-precision manufacturing and complex surface structures. Surface texture often reflects key quality issues such as scratches, dents, cracks, etc. Traditional inspection methods only focus on geometric features and ignore subtle changes in surface texture. They cannot accurately identify and locate these surface defects. This limitation directly leads to insufficient accuracy in defect detection, which is prone to missed detection or false detection, affecting the effectiveness of the overall inspection.

[0003] In summary, traditional contact instrument 3D contour defect detection methods can often only capture the geometric shape and contour of an object, and have weak capabilities for accurate extraction and analysis of surface texture. At the same time, it is difficult to accurately locate defects on complex or irregular 3D surfaces, resulting in technical problems such as insufficient accuracy, comprehensiveness and effectiveness of defect detection results. Summary of the invention

[0004] The purpose of this application is to provide a contact instrument three-dimensional contour defect detection method based on texture feature analysis, so as to solve the technical problem that traditional contact instrument three-dimensional contour defect detection methods can often only capture the geometric shape and contour of the object, but have weak capabilities for accurate extraction and analysis of surface texture, and it is difficult to accurately locate defects on complex or irregular three-dimensional surfaces, resulting in insufficient accuracy, comprehensiveness and effectiveness of defect detection results.

[0005] In view of the above problems, the present application provides a contact instrument three-dimensional contour defect detection method based on texture feature analysis, including: extracting three-dimensional contour coordinate data of the contact instrument under test; performing image acquisition according to the three-dimensional contour coordinate data, fitting according to the mapping relationship between the image and the three-dimensional contour coordinates, and constructing the image distribution coordinates of the three-dimensional contour; constructing a filter group, and performing texture recognition on the distribution image of the three-dimensional contour through the filter group to obtain texture features and their distribution coordinates; identifying and separating defect areas according to the texture features to obtain defect recognition areas, performing defect impact assessment on the defect recognition areas according to the distribution coordinates, and feeding back defect detection results.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] By acquiring images according to the three-dimensional contour coordinate data of the contact instrument under test and fitting them according to the mapping relationship between the image and the three-dimensional contour coordinates, the image distribution coordinates of the three-dimensional contour are constructed; then, texture recognition is performed on the distribution image of the three-dimensional contour through a filter group to obtain texture features and their distribution coordinates; then, defect area recognition and separation are performed according to the texture features to obtain the defect recognition area; finally, the defect impact of the defect recognition area is evaluated according to the distribution coordinates, and the defect detection results are fed back; through the above steps, the surface texture features of the contact instrument can be effectively extracted and analyzed, and the technical goal of accurately locating the defect area can be achieved. At the same time, the degree of defect impact can be effectively evaluated, thereby improving the accuracy, comprehensiveness and reliability of the defect detection results, and providing effective support for the subsequent maintenance and management of the contact instrument.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented according to the contents of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are specifically cited below. It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easy to understand through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.

[0010] Figure 1 A schematic diagram of a process flow of a contact instrument three-dimensional contour defect detection method based on texture feature analysis in this application;

[0011] Figure 2 This is a schematic diagram of the process of obtaining a defect recognition area in a three-dimensional contour defect detection method of a contact instrument based on texture feature analysis in the present application. DETAILED DESCRIPTION

[0012] This application provides a contact instrument three-dimensional contour defect detection method based on texture feature analysis, which solves the technical problem that the traditional contact instrument three-dimensional contour defect detection method can only capture the geometric shape and contour of the object, and has weak ability to accurately extract and analyze surface texture, and it is difficult to accurately locate defects on complex or irregular three-dimensional surfaces, resulting in insufficient accuracy, comprehensiveness and effectiveness of defect detection results. The surface texture features of the contact instrument can be effectively extracted and analyzed to achieve the technical goal of accurately locating the defect area, and the degree of defect impact can be effectively evaluated, thereby improving the accuracy, comprehensiveness and reliability of the defect detection results, and providing effective support for the subsequent maintenance and management of the contact instrument.

[0013] Below, the technical solutions in the present application will be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments of the present application. It should be understood that the present application is not limited to the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application. It should also be noted that, for the convenience of description, only the parts related to the present application are shown in the accompanying drawings, rather than all of them.

[0014] For examples, please see the attached Figure 1 The present application provides a contact instrument three-dimensional contour defect detection method based on texture feature analysis, which specifically includes the following steps:

[0015] Step 1: Extract the three-dimensional contour coordinate data of the contact instrument under test.

[0016] Specifically, by using contact measurement equipment, such as a coordinate measuring machine (CMM), moving along a predetermined path, the coordinate points on the surface of the object are collected point by point to form accurate three-dimensional coordinate data. During the detection process, the contact equipment will record the three-dimensional coordinates of each contact point, usually including the coordinate values ​​in the three directions of X, Y, and Z. By measuring multiple points, the three-dimensional contour of the object being measured is formed, and the three-dimensional contour coordinate data of the contact instrument being measured is obtained. The three-dimensional contour coordinate data can accurately reflect the overall geometric structure of the object, providing a basis for subsequent image fitting and analysis.

[0017] Step 2: According to the three-dimensional contour coordinate data, image acquisition is performed, and fitting is performed according to the mapping relationship between the image and the three-dimensional contour coordinates to construct the image distribution coordinates of the three-dimensional contour.

[0018] Specifically, based on the three-dimensional contour coordinate data, select a suitable angle and position to capture the image of the contact instrument under test, such as using a high-resolution camera or imaging device to ensure that the details of the object surface are captured, and the image acquisition viewing angle should cover various key areas on the object surface to ensure that each three-dimensional coordinate point can be reflected in the image. After the image acquisition is completed, the pixel points in each image are associated with their corresponding three-dimensional coordinates through projection mapping technology, that is, the pixel points in the two-dimensional image plane are projected into the three-dimensional space to form a mapping relationship between the image and the three-dimensional coordinate points, so as to achieve a one-to-one correspondence between the texture features in the image and the three-dimensional contour data of the object.

[0019] Then, according to the established mapping relationship, the collected image is fitted with the three-dimensional contour coordinate data, such as fitting through a geometric transformation algorithm. Commonly used algorithms include affine transformation, perspective transformation, etc. Technical personnel in this field can select an appropriate fitting algorithm according to actual needs. Fitting refers to mapping the pixel coordinates (u, v) in the two-dimensional image to the corresponding coordinates (X, Y, Z) in the three-dimensional space through geometric transformation. During the fitting process, it is necessary to ensure that the texture details in the image can be accurately corresponded to the three-dimensional coordinate points, that is, each three-dimensional coordinate point corresponds to a pixel point in the image, so as to ensure the close connection between the surface texture of the object and the three-dimensional geometric shape. The fitting process ensures that the texture details in each image are closely related to the three-dimensional coordinate points, thereby mapping the two-dimensional image features to the three-dimensional space to form a complete three-dimensional contour image model.

[0020] Then the fitted image data is combined with the three-dimensional coordinate points, that is, each pixel point in the image is combined with the three-dimensional contour coordinate point. After fitting, each three-dimensional contour coordinate point obtained will have a corresponding image pixel point (i.e., texture information). These pixel points not only record the color information and brightness information in the image, but also contain rich surface information such as texture and edge. These information are associated with the three-dimensional coordinate points to form "image distribution coordinates" to construct the image distribution coordinates of the three-dimensional contour of the measured object. Through geometric fitting, the texture information in the two-dimensional image can be accurately matched to the three-dimensional contour coordinate points to ensure that the texture, color, brightness and other data of each coordinate point are saved. Each three-dimensional coordinate point (X, Y, Z) not only contains position data, but also combines the pixel features in the image (such as color value, brightness, surface texture, etc.). By constructing the image distribution coordinates of the three-dimensional contour, not only the geometric shape of the object surface can be accurately captured, but also the surface texture, color, brightness and other features of the object can be displayed, which not only improves the ability to capture surface details, but also enhances the ability to identify and analyze complex or irregular surface defects.

[0021] Step three: construct a filter group, and use the filter group to perform texture recognition on the distribution image of the three-dimensional contour to obtain texture features and their distribution coordinates.

[0022] Specifically, first, a filter bank is constructed, which is composed of multiple filters. The function of the filter bank is to effectively identify the textures of various scales and directions in the image. The filter used in this application is a Gabor filter. The Gabor filter can analyze the image at different scales and directions to extract the details of the surface texture, and can be optimized for specific texture features (such as frequency, directionality, roughness, etc.) at the same time, so as to ensure that diversified texture information is extracted. The design of the Gabor filter is usually set based on parameters such as frequency, direction, and scale. These parameters can be optimized for specific texture features, so that the filter can perform texture analysis at different scales and directions.

[0023] Then, the three-dimensional contour distribution image is subjected to texture recognition by the filter group, that is, the three-dimensional contour image is taken as input and processed by the pre-constructed filter group in sequence, wherein each filter responds to different texture features in the image, thereby identifying the texture details on the surface, ensuring that various complex surface features can be identified, for example, a rough texture may produce a strong response in a higher frequency filter, while a fine texture may show a strong feature in a low frequency filter. After being processed by the filter, texture features of various scales and directions in the three-dimensional contour image can be extracted, including surface roughness, directionality, frequency features, etc., wherein roughness refers to the roughness of the surface texture. The roughness on the surface can be evaluated by the strength of the filter response; directionality refers to the arrangement direction of the surface texture of an object, and the Gabor filter can effectively capture texture features in different directions, especially linear or strip features on the surface of an object; frequency features refer to the periodicity of the surface texture, such as high-frequency textures corresponding to small surface features, while low-frequency textures correspond to larger surface structures.

[0024] Then, while performing texture recognition, the identified texture features are associated with their three-dimensional coordinate positions through the previously constructed mapping relationship between the image and the three-dimensional coordinates, that is, each texture feature will be accompanied by corresponding three-dimensional coordinate information, forming a complete texture distribution coordinate system, and finally outputting the texture features and their distribution coordinates. By constructing texture features and their distribution coordinates, the texture features can not only have the position and feature information in the two-dimensional image, but also represent their precise position in space through three-dimensional coordinates, which can not only provide accurate three-dimensional surface detail information, but also accurately locate subtle defects on the surface, even those details that are difficult to detect in traditional two-dimensional images.

[0025] By constructing a filter group to perform texture recognition on the three-dimensional contour image, the texture features of the object surface can be effectively extracted, and then the texture features can be associated with the coordinate information in the three-dimensional space. This not only enhances the perception of the surface details of the contact instrument being tested, but also provides accurate data support for subsequent defect location and analysis.

[0026] Step 4: Identify and separate defect areas according to the texture features, obtain defect identification areas, evaluate the defect impact of the defect identification areas according to distribution coordinates, and feed back defect detection results.

[0027] Specifically, defective areas are identified and separated based on the texture features. First, the extracted texture features are further analyzed to calculate their image gradients. Image gradients can reflect the degree of change in texture features, such as changes in brightness, frequency, and direction. If the texture features of a certain area suddenly change significantly (such as suddenly becoming rough or the brightness changes dramatically), then the area may be a potential defect. The image gradient is obtained by calculating the rate of change of each pixel in the image in the horizontal and vertical directions. Commonly used image gradient calculation methods include the Sobel operator, the Prewitt operator, etc., that is, the gradient change is calculated for each texture feature pixel in the three-dimensional contour image to obtain a gradient map. Through gradient analysis, areas with smooth texture and small changes can be distinguished from areas with drastic texture changes, providing a basis for separating defective areas.

[0028] Then, texture feature segmentation is performed based on the image gradient, and the image is divided into different regions, each of which represents relatively consistent texture features. Image segmentation algorithms can be used for processing, for example, edge detection algorithms are used for image segmentation, that is, by detecting regions with large gradient changes, the portions with clear edges are separated. After image segmentation, several segmented regions are obtained, each of which represents a different texture feature pattern; defect types are further identified in the segmented feature regions, and common defect types include cracks, dents, scratches, etc. Cracks refer to thin and continuous linear defects on the surface; dents refer to local sunken areas on the surface, usually accompanied by sudden changes in lighting or texture; scratches refer to superficial lines caused by external physical effects, usually manifested as drastic changes in local brightness or texture. Then, based on the identified defect type and defect degree, a complete defect identification result is generated, which contains information such as the category, size, shape, etc. of each defect area; then, based on the identified regional defect category, adjacent feature segmentation areas with the same defect category are found, that is, for adjacent areas of the same category, texture features are spliced ​​to form a larger and unified defect identification area. The splicing method can be achieved by detecting the edge similarity of the segmented area or the consistency of the defect type, and finally a defect identification area is generated, in which the defect identification area contains information such as the defect category, size, shape, etc.

[0029] Then, the defect impact assessment is performed on the defect identification area according to the distribution coordinates, including defect location assessment and defect severity assessment. For example, according to the specific position of the defect identification area in the three-dimensional coordinate space, it is judged whether the defect is located in the key area. For example, the area where important components such as probes and support structures are located has a greater impact on the performance of the equipment, so the greater the impact of the defect in this area; combined with the distribution coordinates, geometric characteristics (such as area, depth) and importance of the location of the defect, the impact of the defect is assessed. Defects with greater depth or wider area usually have a greater impact on the structural integrity or function of the equipment, and the greater the impact of the defect. Then, a comprehensive defect impact score is generated by combining the defect location assessment results and the defect severity assessment results. Finally, a complete defect detection result is output, which contains information such as the coordinate position, area, geometric characteristics, and impact score of the defect. At the same time, the coordinate information of the defect area is marked in the three-dimensional model, and the defect areas of different severity are clearly marked. They can be classified and marked by color, size, etc. to generate a three-dimensional visual model with defect marks. Finally, the defect detection results are fed back to the staff for subsequent maintenance and management of the contact instrument under test. By evaluating the impact of defects based on their distribution coordinates and geometric features, a comprehensive defect detection report is generated, and defect areas are visually marked in three-dimensional space, providing accurate data support for subsequent repair and management.

[0030] By providing a contact instrument three-dimensional contour defect detection method based on texture feature analysis, the traditional contact instrument three-dimensional contour defect detection method can solve the technical problem that the traditional contact instrument three-dimensional contour defect detection method can only capture the geometric shape and contour of the object, and has weak ability to accurately extract and analyze the surface texture. At the same time, it is difficult to accurately locate defects on complex or irregular three-dimensional surfaces, resulting in insufficient accuracy, comprehensiveness and effectiveness of defect detection results. By collecting images according to the three-dimensional contour coordinate data of the contact instrument under test, and fitting according to the mapping relationship between the image and the three-dimensional contour coordinates, the image distribution coordinates of the three-dimensional contour are constructed; then, the texture recognition of the distribution image of the three-dimensional contour is performed by the filter group to obtain the texture features and their distribution coordinates; then, the defect area is identified and separated according to the texture features to obtain the defect recognition area; finally, the defect impact of the defect recognition area is evaluated according to the distribution coordinates, and the defect detection results are fed back; through the above steps, the surface texture features of the contact instrument can be effectively extracted and analyzed, and the technical goal of accurately locating the defect area can be achieved. At the same time, the degree of defect impact can be effectively evaluated, thereby improving the accuracy, comprehensiveness and reliability of the defect detection results, and providing effective support for the subsequent maintenance and management of the contact instrument.

[0031] Furthermore, the present application also includes:

[0032] Establish image distribution coordinates of a first contour and image distribution coordinates of a second contour, wherein the image distribution coordinates of the first contour are the image distribution coordinates of the contact instrument in an initial state, and the image distribution coordinates of the second contour are the image distribution coordinates of the contact instrument in a measuring state; establish a component coordinate mapping relationship between the first contour and the second contour according to the distribution coordinates; and align and compare the texture features of the image distribution coordinates of the first contour with the texture features of the image distribution coordinates of the second contour based on the component coordinate mapping relationship to obtain difference texture features and their distribution coordinates.

[0033] Specifically, when the contact instrument is in the initial state (unmeasured state), a high-precision contact measurement device is used to obtain its three-dimensional contour data, and its texture features and geometric information are extracted based on the three-dimensional contour data of the initial state, and image distribution coordinates are established. These coordinates include the position and texture information of the instrument surface in three-dimensional space, which are the reference benchmarks of the device in normal working state, and the image distribution coordinates of the first contour are established. On the other hand, when the contact instrument is in use or measurement state, its three-dimensional contour data is re-collected to capture the surface contour information and texture features in the current state, and the image distribution coordinates of the second contour are established, wherein the image distribution coordinates of the second contour reflect the surface changes of the device due to wear or other factors during actual use.

[0034] Then, the image distribution coordinates of the two groups of contours in the initial state and the measured state are aligned to establish a component coordinate mapping relationship between the first contour and the second contour. The mapping relationship is used to align the contour data in different states to ensure that the texture features can be compared one-to-one in different states. The mapping can be achieved through a geometric alignment algorithm (such as rigid body transformation, ICP algorithm) to ensure the precise alignment of the two groups of contour data. Then, based on the component coordinate mapping relationship, the texture features of the image distribution coordinates of the first contour are aligned with the texture features of the image distribution coordinates of the second contour. By comparing the texture features of the two groups of contours, the difference texture features in the two states are obtained, and the distribution coordinates of the difference texture features in three-dimensional space are calculated, where the difference texture features correspond to the changes caused by wear, scratches or use on the surface of the equipment. The specific location and area where the difference occurs can be clarified through the three-dimensional distribution coordinates, and the difference texture features and their distribution coordinates can be obtained. The above-mentioned difference texture feature comparison is suitable for scenarios such as equipment wear, detection accuracy analysis and status monitoring. For example, by comparing the contour and texture features in different states, the degree of wear on the surface of the contact instrument caused by long-term use or external force can be accurately identified, helping maintenance and repair decisions; by comparing the surface contour data in the initial state and the use state, the deformation or damage of the equipment in use can be analyzed to ensure that the equipment accuracy meets the requirements; by regularly comparing the contour data, early damage or abnormalities can be identified and maintenance measures can be taken in advance.

[0035] The above steps can accurately identify the surface texture differences of contact instruments in different states, especially the changes caused by wear during the use of the equipment. It can not only provide accurate wear location and degree, but also play an important role in various equipment monitoring and maintenance scenarios.

[0036] Furthermore, the present application also includes the following steps:

[0037] A dynamic sliding time window is set to switch between the image distribution coordinates of the first contour and the image distribution coordinates of the second contour, wherein the adaptive adjustment condition of the dynamic sliding time window corresponds to the change in the difference texture feature and its distribution coordinates. When the difference texture feature and its distribution coordinates change, the length of the dynamic sliding time window is adjusted; nodes are adjusted according to the length of the dynamic sliding time window to construct time series characteristics of the difference texture feature and the change in its distribution coordinates; the frequency of the difference texture feature is calculated according to the time series characteristics, and warning information is generated according to the frequency of difference occurrence, and the warning information is used to remind of measurement abnormalities of contour defects.

[0038] Specifically, first, a dynamic sliding time window is set between the image distribution coordinates of the first contour (initial state) and the second contour (measurement state), and the dynamic sliding time window is used to switch between the detection data in different states, collect multiple contour data and analyze their changes over time, wherein the dynamic sliding time window can adaptively adjust the window length according to the change amount of the difference texture feature and its distribution coordinates. When the difference texture feature and its distribution coordinates change, the length of the sliding time window is dynamically adjusted. When the change is large, the time window is shortened and data is collected frequently; when the change is small, the time window is extended and the data collection frequency is reduced. Through adaptive adjustment, higher frequency detection can be performed when the equipment wears faster or the texture changes greatly, ensuring that key changes in the equipment status are captured in time.

[0039] Then, the detection data of the node is adjusted according to the length of the dynamic sliding time window, and the time series characteristics of the difference texture features and the change amount of their distribution coordinates are constructed, that is, the changes of texture features are recorded in multiple time periods, and these changes are associated with the time when they occur to form a time series curve, wherein the time series characteristics can help identify the change trend of texture differences, and judge whether the texture features are gradually deepening, whether they are continuously expanding, whether there are mutations, etc. The frequency of the difference texture features is further calculated based on the time series characteristics, which can be determined based on the texture change amount within the time window. The higher the frequency, the more frequent the change of texture differences, indicating that the surface of the equipment may be more quickly worn or abnormal. The frequency analysis of the difference texture can help determine whether the wear or defects show an accelerating trend and evaluate whether the equipment status deteriorates in a short period of time.

[0040] Then, warning information is generated based on the frequency of the difference. If the frequency of the difference exceeds the preset frequency threshold, a warning is issued. The warning information is used to remind the user that the equipment may have contour defects, increased wear or measurement abnormalities. By generating warning information, potential problems of the equipment can be warned in time, and corresponding maintenance plans can be formulated to repair the equipment or adjust the operation in time to ensure equipment performance and measurement accuracy.

[0041] Furthermore, the present application includes:

[0042] Based on historical abnormal cases, the contact instrument is analyzed for measurement influencing features, and components are partitioned according to the recognition accuracy and division granularity of the influencing features; filter parameters of each component partition are configured to construct the filter group, which corresponds to the component partition.

[0043] Specifically, by analyzing the abnormal cases of contact instruments in historical measurements, we can identify which parts have the greatest impact on the measurement results. For example, the probe part and the area that contacts the measured object usually have a greater impact on the measurement accuracy, so the defects in these parts need to be identified with high precision. At the same time, the influencing features in the historical data are extracted, and combined with the importance of the parts (such as probes, contact surfaces, supports, etc.), it is clear which parts have a greater impact on the measurement results, and further feature division is performed to determine the key parts that affect the measurement. Then, the parts are partitioned according to the recognition accuracy and division granularity of the influencing features. For key parts with a large measurement impact, such as the probe and contact area, high-granularity division is performed, that is, these areas are divided into multiple small partitions, so as to identify the subtle defects in these areas. The defects in each partition, even if they are tiny flaws, may have a significant impact on the measurement results, so high-precision identification parameters and detailed analysis are required; for parts with low measurement accuracy requirements, such as supports or structural parts, the defects in these areas have a small impact on the measurement results, so low-granularity division can be performed. The area partition is larger, mainly identifying large defects, and the identification parameters can also be relatively simplified, reducing the focus on subtle defects, thereby improving detection efficiency.

[0044] Then, according to the matching filter parameters of the component partition configuration, that is, according to the accuracy requirements of each component partition, a corresponding filter group is constructed, and the filter group contains filter configurations adapted to different partition granularities. For example, for high-granularity partition areas, such as the partition of the probe or contact surface, high-precision filter parameters are configured. The filter used in this application is a Gabor filter to ensure that small defects such as small cracks and scratches can be identified. These filters need to be analyzed at multiple scales and in multiple directions to ensure high-precision defect identification; for low-granularity partition areas, such as support parts, simpler filters are configured, focusing mainly on the identification of large defects. The filter parameters can be set to detect texture changes in larger areas or rougher surface defects, reducing attention to small defects and saving computing resources and time.

[0045] By partitioning the components and configuring matching filter groups based on the component partitioning results, high-precision defect identification can be performed in key areas, and large defects can be quickly identified in secondary areas, which not only improves the detection efficiency but also ensures the detection accuracy of key components.

[0046] Furthermore, the present application also includes the following steps:

[0047] The filter parameters include wavelength, direction, scale, and aspect ratio, wherein the recognition accuracy is inversely proportional to the wavelength, scale, and aspect ratio, and the number of directions is proportional to the recognition accuracy. The filter parameters of the component partitions are configured according to the recognition accuracy, and a filter group for each component partition is constructed; the parameter gradient of the granularity partition is set according to the division granularity of the component partition, and the filter group of each component partition is parameter matched according to the parameter gradient of the granularity partition, and a mapping association between the granularity partition and the filter group is established.

[0048] Specifically, the filter used in the present application is a Gabor filter, and the filter parameters include wavelength, direction, scale, and aspect ratio, wherein the wavelength determines the response frequency of the filter to texture features, and a shorter wavelength can capture subtle features, so the wavelength is inversely proportional to the recognition accuracy. The higher the accuracy requirement, the shorter the wavelength should be; the more filter directions there are, the more comprehensive the texture information from different directions can be captured, so the number of directions is proportional to the recognition accuracy, and the higher the recognition accuracy, the more filter directions there should be; the scale parameter is used to control the size of the filter action area, a smaller scale can capture local features and is suitable for high-precision areas, while a larger scale is suitable for detecting overall features, so the scale is inversely proportional to the recognition accuracy; the aspect ratio of the filter controls its sensitivity to textures in a specific direction, a smaller aspect ratio can capture subtle changes more sensitively and is suitable for high-precision detection areas, so the aspect ratio is inversely proportional to the recognition accuracy.

[0049] Then, filter parameters are configured for the component partitions according to the recognition accuracy. For example, the probe tip is a key part in contact measurement, and the accuracy requirement is extremely high. Any tiny defect may affect the overall measurement accuracy, so it is necessary to configure a high-precision filter group (short wavelength, multi-direction, small aspect ratio, small scale); the shell surface is mainly used for the detection of the overall structure, and the defects do not directly affect the measurement accuracy, so lower-precision detection is required, so it is necessary to configure a low-precision filter group (long wavelength, few directions, large aspect ratio, large scale); tiny wear may occur near the contact point, which has a certain impact on the measurement, so medium-precision detection is required, and a medium-precision filter group is configured.

[0050] Then, the parameter gradient of the granularity partition is set according to the granularity of the component partition, wherein the granularity of the granularity partition directly affects the parameter accuracy of the filter group. When setting the parameter gradient, the accuracy requirements of the partition should be considered, and then the parameters of the filter group are adjusted. The finer the granularity of the granularity partition, the higher the required filter parameter accuracy, and the coarser the granularity, the lower the accuracy requirements of the filter parameters. The granularity partition includes fine-grained partition (high accuracy) and coarse-grained partition (low accuracy). Fine-grained partition is suitable for key components or important surface areas, such as probe tips, contact points, etc., and can set short wavelength, multi-direction, small aspect ratio, and small scale filter parameters, such as wavelength of 2 to 6 pixels, scale of 1 to 3 pixels, aspect ratio of 0.5 to 0.8, and number of directions of 4 to 8. Coarse-grained partitioning is suitable for non-critical components or large structural components, such as shell surfaces, supports, etc., and can set filter parameters with long wavelength, few directions, large aspect ratio, and large scale, such as a wavelength of 10 to 20 pixels, a scale of 5 to 10 pixels, an aspect ratio of 1.0, and 2 to 4 directions.

[0051] Then, according to the set granularity partition parameter gradient, the filter groups of each component partition are matched with parameters. The matching process selects the corresponding filter parameters according to the granularity characteristics of the partition to achieve accurate detection. For example, according to the characteristics of the fine-grained partition, a filter group with a smaller wavelength and scale and more directions is selected to ensure that the subtle features of the area are captured. At the same time, for each fine-grained partition, according to the preset parameter gradient, the wavelength, scale, number of directions and aspect ratio of the filter group are gradually adjusted to ensure that they match the accuracy requirements of the partition; according to the characteristics of the coarse-grained partition, a filter group with a larger wavelength and scale and fewer directions is selected to capture macroscopic features. At the same time, in each coarse-grained partition, a preset coarse-grained parameter gradient is used to select a suitable filter combination. Then a mapping association between the granularity partition and the filter group is established, where the granularity details of each partition determine the filter group parameters used.

[0052] By matching the parameters of the filter group according to the parameter gradient of the granular partition, it is possible to ensure that the detection accuracy of different component areas matches the actual needs. While ensuring the detection accuracy of key areas, it is also possible to improve the overall detection efficiency so that the detection of each partition meets the accuracy requirements.

[0053] Furthermore, the present application also includes the following steps:

[0054] According to the influencing characteristics of the historical abnormal cases, recognition accuracy clustering is performed to determine the highest threshold, lowest threshold and mean of the accuracy area; for the highest threshold, lowest threshold and mean of each accuracy area, filter parameter test data optimization is performed to obtain parameter values ​​of the accuracy area, which include the highest threshold, lowest threshold and mean, and a parameter configuration library is constructed; the recognition accuracy is used as the input index, and traversal matching is performed in the parameter configuration library to determine the filter configuration parameters.

[0055] Specifically, according to the influencing characteristics of historical abnormal cases, the recognition accuracy of the components is clustered and analyzed, that is, components with high recognition accuracy (such as probes) and components with low recognition accuracy (such as supports) are divided into different categories, and in the clustering process, the highest threshold, lowest threshold and mean are determined for each precision area. For example, the area with the highest recognition accuracy has a smaller texture change detection threshold, while the low-precision area has a larger threshold. Then, according to the highest threshold, lowest threshold and mean of each precision area, the filter parameters are experimentally tuned, and the combination of different wavelengths, scales, aspect ratios and the number of directions is measured through multiple experiments to evaluate the detection effect of each area; in the experimental process, the best filter parameter combination is optimized according to the experimental results, and the most suitable wavelength, scale, aspect ratio and number of directions are determined for each precision area for the highest, lowest and mean thresholds respectively; at the same time, a set of parameter values ​​is generated for each precision area, including the filter parameters corresponding to the highest threshold, the lowest threshold and the mean, which are used to configure the filter of the corresponding area to obtain the parameter values ​​of the precision area. Then, the filter parameter values ​​(including wavelength, scale, aspect ratio and number of directions) of different precision areas are stored in the parameter configuration library to generate a parameter configuration library, in which each record in the library corresponds to an identification precision area, and the parameter values ​​include the highest, lowest and mean values.

[0056] Finally, the recognition accuracy is used as an input index to perform traversal matching in the parameter configuration library, and the corresponding filter configuration parameters are determined according to the input recognition accuracy, and the filter configuration parameters are output.

[0057] Further, as attached Figure 2 As shown, this application includes:

[0058] Calculate the image gradient of the texture feature, perform texture feature segmentation according to the image gradient, and obtain a feature segmentation area; identify the defect type of the feature segmentation area, determine the regional defect category, generate a defect recognition result according to the regional defect category and the degree of defect recognition, and fit the defect recognition result to the feature segmentation area to obtain the defect recognition area; wherein, according to the regional defect category, select adjacent feature segmentation areas with the same defect category for texture feature splicing to construct the same family defect recognition area.

[0059] Specifically, first, the image gradient of the texture feature is calculated to detect the edges and mutation points of texture changes. The image gradient can effectively identify changes in texture direction and intensity, thereby helping to identify potential defect areas; then, based on the image gradient information, the texture features in the image are segmented, that is, the surface area is divided into multiple feature segmentation areas, each segmentation area represents a different texture change interval, and a feature segmentation area is obtained.

[0060] The defect type is identified for the feature segmentation area. Common defect types include scratches, cracks, dents, peeling, etc. By analyzing parameters such as texture features, morphology, direction, depth, etc., each feature segmentation area can be classified into different defect categories, such as classifying irregular surface cracks as "cracks" and classifying subtle changes in surface texture as "scratches". Then, the severity of the defect is evaluated based on the geometric features of the defect (such as depth, width, length), and different defect types are classified, such as shallow scratches, deep scratches, or small cracks, large cracks, etc.; defect recognition results are generated based on the regional defect category and the degree of defect recognition. The defect recognition results are then fitted to the feature segmentation area so that each defect result accurately corresponds to its segmentation area, and a defect recognition area is obtained.

[0061] Further, according to the regional defect category, adjacent feature segmentation regions with the same defect category are selected for texture feature stitching. First, the correlation between adjacent feature segmentation regions is identified through connectivity analysis. If the adjacent segmentation regions have the same defect category, the system will analyze whether they belong to the same overall defect. For example, if the adjacent scratch regions have different depths but are connected in texture direction and features, these regions can be regarded as the same defect. Then, for the defect regions of the same category found in the connectivity analysis, these adjacent regions are spliced ​​and merged into an overall region. For example, different parts of a scratch may have different depths in different regions. These regions are spliced ​​into the same scratch through information such as texture direction and depth changes. Then, the same family defect recognition region is constructed based on the regional splicing results.

[0062] Furthermore, the present application also includes the following steps:

[0063] The image is preprocessed to mark the foreground and background, wherein the background is marked as 0 and the foreground is marked as the corresponding gradient value; a gradient amplitude map is constructed according to the image gradient, wherein the minimum gradient amplitude region is the initial extension point, and region filling is performed starting from the initial extension point, and all pixels with the same gradient in the initial extension point and its neighborhood are marked with the same label, and a first segmentation boundary is determined according to the encounter of regions with the same label; the minimum gradient is screened outside the region filling and the region is re-filled to obtain a second segmentation boundary, and so on until the region filling with the maximum gradient amplitude is completed, and the nth segmentation boundary is obtained, where n is the number of gradients; the first segmentation boundary, the second segmentation boundary, and the nth segmentation boundary are used to perform segmentation to obtain the feature segmentation region.

[0064] Specifically, first, the image is preprocessed, including grayscale processing, denoising and edge enhancement processing; then the gradient in the image is calculated to highlight the edges and features of the image. Commonly used gradient calculation methods include Sobel, Scharr or Laplacian operators; then the gradient amplitude map is calculated based on the gradient map, where the gradient amplitude of each pixel represents its gradient size. The larger the gradient amplitude, the more obvious the edge of the point; further, based on the gradient amplitude map, the area with a smaller gradient is marked as the background, and the value of the background pixel is set to 0. For the foreground area, each pixel is marked according to its gradient amplitude to ensure that the marking value of the foreground can reflect its gradient size.

[0065] Then, a gradient amplitude map is constructed based on the image gradient, where the area with the minimum gradient amplitude is the initial extension point; then, starting from the initial extension point, the area is filled, and all pixels with the same gradient in the initial extension point and its neighborhood are marked with the same label. When the labels of adjacent areas meet, the first segmentation boundary is formed. Then, the minimum gradient is screened outside the area filling and the area is filled again, that is, outside the initial filling, the next minimum gradient value point is screened and the area is filled to obtain the second segmentation boundary, and so on until the area filling with the maximum gradient amplitude is completed, and the nth segmentation boundary is obtained, where n is the number of gradients. Finally, the image is segmented using the first segmentation boundary, the second segmentation boundary, and the nth segmentation boundary to obtain the feature segmentation area.

[0066] In summary, the contact instrument three-dimensional contour defect detection method based on texture feature analysis provided by the present application has the following technical effects:

[0067] By acquiring images according to the three-dimensional contour coordinate data of the contact instrument under test and fitting them according to the mapping relationship between the image and the three-dimensional contour coordinates, the image distribution coordinates of the three-dimensional contour are constructed; then, texture recognition is performed on the distribution image of the three-dimensional contour through a filter group to obtain texture features and their distribution coordinates; then, defect area recognition and separation are performed according to the texture features to obtain the defect recognition area; finally, the defect impact of the defect recognition area is evaluated according to the distribution coordinates, and the defect detection results are fed back; through the above steps, the surface texture features of the contact instrument can be effectively extracted and analyzed, and the technical goal of accurately locating the defect area can be achieved. At the same time, the degree of defect impact can be effectively evaluated, thereby improving the accuracy, comprehensiveness and reliability of the defect detection results, and providing effective support for the subsequent maintenance and management of the contact instrument.

[0068] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

[0069] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application belong to the scope of the present application and its equivalent technology, the present application is also intended to include these modifications and variations.

Claims

1. A three-dimensional contour defect detection method of a contact instrument based on texture feature analysis, characterized in that the method include: Extracting the three-dimensional contour coordinate data of the contact instrument under test; According to the three-dimensional contour coordinate data, image acquisition is performed, and fitting is performed according to the mapping relationship between the image and the three-dimensional contour coordinates to construct image distribution coordinates of the three-dimensional contour; Constructing a filter group, and performing texture recognition on the distribution image of the three-dimensional contour through the filter group to obtain texture features and distribution coordinates thereof; Identify and separate defect areas according to the texture features, obtain defect identification areas, evaluate the defect impact of the defect identification areas according to distribution coordinates, and feed back defect detection results; The image distribution coordinates and the distribution image of the three-dimensional contour are formed by associating each three-dimensional contour coordinate point with the texture information contained in the corresponding image pixel point; The constructing of the filter group, performing texture recognition on the distribution image of the three-dimensional contour by the filter group, and obtaining texture features and distribution coordinates thereof, comprises: Based on historical abnormal cases, the contact instrument is analyzed for measurement influencing features, and components are partitioned according to the recognition accuracy and division granularity of the influencing features; Configuring filter parameters of each component partition to construct the filter group, wherein the filter group corresponds to the component partition; The configuring filter parameters of each component partition to construct the filter bank includes: The filter parameters include wavelength, direction, scale, and aspect ratio, wherein the recognition accuracy is inversely proportional to the wavelength, scale, and aspect ratio, and the number of directions is proportional to the recognition accuracy. The filter parameters of the component partitions are configured according to the recognition accuracy to construct a filter group for each component partition. Setting a parameter gradient of a granularity partition according to the division granularity of the component partition, performing parameter matching on the filter groups of each component partition according to the parameter gradient of the granularity partition, and establishing a mapping association between the granularity partition and the filter group; Configuring filter parameters for the component partitions according to the recognition accuracy includes: According to the impact characteristics of the historical abnormal cases, identification accuracy clustering is performed to determine the highest threshold, the lowest threshold and the mean of the accuracy area; According to the highest threshold, lowest threshold and mean value of each precision area, the filter parameter test data is optimized respectively to obtain the parameter values ​​of the precision area. The parameter values ​​include the highest threshold, the lowest threshold and the mean value, and the parameter configuration library is constructed. Taking the recognition accuracy as an input index, performing traversal matching in the parameter configuration library to determine the filter configuration parameters; Defective region identification and separation are performed according to the texture features to obtain defect identification regions, including: Calculating the image gradient of the texture feature, performing texture feature segmentation according to the image gradient, and obtaining a feature segmentation area; Identify the defect type of the feature segmentation region, determine the regional defect category, generate a defect recognition result according to the regional defect category and the degree of defect recognition, and fit the defect recognition result to the feature segmentation region to obtain the defect recognition region; According to the regional defect category, adjacent feature segmentation regions with the same defect category are selected for texture feature splicing to construct the same family defect recognition region; Calculating the image gradient of the texture feature, performing texture feature segmentation according to the image gradient, and obtaining a feature segmentation area, including: Preprocess the image and mark the foreground and background, where the background is marked as 0 and the foreground is marked as the corresponding gradient value; Constructing a gradient amplitude map according to the image gradient, wherein the minimum gradient amplitude region is an initial extension point, performing region filling starting from the initial extension point, marking all pixels with the same gradient in the initial extension point and its neighborhood with the same label, and determining a first segmentation boundary according to the encounter of regions with the same label; The minimum gradient is selected outside the region filling and the region is filled again to obtain the second segmentation boundary, and so on until the region with the maximum gradient amplitude is filled, and the nth segmentation boundary is obtained, where n is the number of gradients; The first segmentation boundary, the second segmentation boundary, and finally the nth segmentation boundary are used for segmentation to obtain the characteristic segmentation area.

2. The contact instrument three-dimensional contour defect detection method based on texture feature analysis according to claim 1, characterized in that: The method also includes: Establishing image distribution coordinates of a first contour and image distribution coordinates of a second contour, wherein the image distribution coordinates of the first contour are image distribution coordinates of the contact instrument in an initial state, and the image distribution coordinates of the second contour are image distribution coordinates of the contact instrument in a measuring state; Establishing a component coordinate mapping relationship between the first contour and the second contour according to the distribution coordinates; Based on the component coordinate mapping relationship, the texture features of the image distribution coordinates of the first contour are aligned and compared with the texture features of the image distribution coordinates of the second contour to obtain difference texture features and their distribution coordinates.

3. The contact instrument three-dimensional contour defect detection method based on texture feature analysis as claimed in claim 2, characterized in that: The step of obtaining the difference texture features and the distribution coordinates thereof further includes: Setting a dynamic sliding time window to switch between the image distribution coordinates of the first contour and the image distribution coordinates of the second contour, wherein the adaptive adjustment condition of the dynamic sliding time window corresponds to the change amount of the difference texture feature and its distribution coordinates, and when the difference texture feature and its distribution coordinates change, adjusting the length of the dynamic sliding time window; Adjust the nodes according to the length of the dynamic sliding time window to construct the time series features of the difference texture features and the changes in their distribution coordinates; The difference texture feature frequency is calculated according to the time series feature, and warning information is generated according to the difference occurrence frequency. The warning information is used to provide a measurement abnormality reminder for contour defects.

Images