A precision ball screw pair detection system and method thereof

Abstract

The present application relates to the technical field of screw detection, and discloses a precision ball screw pair detection system and method, which comprises an image enhancement module, a feature extraction module, a reference image generation module, a defect image determination module, a defect area positioning module and a monitoring report generation module, wherein: a multi-angle original image sequence is subjected to visual enhancement processing to obtain an enhanced image; multi-dimensional feature extraction is performed on the enhanced image to obtain a multi-dimensional feature image set; the ball screw pair is reversely mapped into a pixel grid of the enhanced image to obtain an optical simulation reference image; the optical simulation reference image and the multi-dimensional feature image set are subjected to causal inference, and an initial defect atlas is determined according to the inference result; conditional area decoupling is performed on the enhanced image to obtain a defect image area; and the multi-dimensional feature image set is subjected to multi-element correlation analysis to generate a defect detection report; the present application can improve the accuracy of precision ball screw pair detection.

Description

Technical Field

[0001] This invention relates to the field of ball screw testing technology, and in particular to a precision ball screw pair testing system and method. Background Technology

[0002] In the field of precision ball screw assembly inspection, the structural precision requirements are extremely high, and the accuracy of the inspection results directly affects the reliability of subsequent applications. However, existing inspection technologies have significant shortcomings in the image preprocessing stage. They lack effective visual enhancement and fusion schemes for multi-angle original images of ball screw assemblies, making it difficult to eliminate possible interference factors in the images, resulting in low utilization of image information. Furthermore, the feature extraction dimension is limited, failing to comprehensively cover key information such as geometric contours, texture gradients, and topological structures, leaving insufficient feature support for subsequent defect identification, thus affecting the accuracy of the inspection.

[0003] Existing technologies lack rationality in the generation of reference images and defect judgment. They struggle to accurately reverse map the vector geometric elements and tolerance zones of design drawings to the pixel grid of the actual enhanced image. The construction of coordinate mapping relationships lacks rigorous spatial alignment and scaling mechanisms, resulting in low matching between optical simulation reference images and actual inspection images. In the defect inference process, the analysis of feature differences lacks systematic causal inference and comprehensive evaluation, making it impossible to scientifically quantify defect confidence. This leads to poor accuracy of the initial defect map, making it difficult to guarantee the reliability of subsequent defect area location and inspection report generation. Consequently, it fails to meet the high-standard inspection requirements of precision ball screw pairs. Therefore, improving the accuracy of precision ball screw pair inspection has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a precision ball screw pair detection system and method to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides a precision ball screw pair inspection system, characterized in that the system includes an image enhancement module, a feature extraction module, a reference image generation module, a defect image determination module, a defect region localization module, and a monitoring report generation module, wherein:

[0006] The image enhancement module is used to perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair.

[0007] The feature extraction module is used to perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair;

[0008] The reference image generation module is used to reverse map the vector geometric elements and tolerance bands of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair.

[0009] The defect image determination module is used to perform causal inference on the optical simulation reference image and the multi-dimensional feature image set, and determine the initial defect map of the ball screw pair based on the inference result.

[0010] The defect region localization module is used to perform conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair.

[0011] The monitoring report generation module is used to perform multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair.

[0012] In a preferred embodiment, when the image enhancement module performs visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair, it is specifically used for:

[0013] Acquire multi-angle raw image sequences of the ball screw assembly;

[0014] Feature filtering is performed on the multi-angle original image sequence to obtain the initial image set of the ball screw pair;

[0015] The initial image set is fused across different viewpoints to obtain the fused image of the ball screw pair;

[0016] Spatial structure analysis is performed on the fused image to obtain the grayscale distribution features of the fused image;

[0017] Based on the grayscale distribution characteristics, the fused image is adaptively intensity adjusted to obtain an enhanced image of the ball screw pair.

[0018] In a preferred embodiment, when the feature extraction module performs multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair, it is specifically used for:

[0019] Edge detection is performed on the enhanced image to obtain the geometric contour information of the ball screw pair;

[0020] Based on the geometric contour information, multi-scale texture gradient analysis is performed on the enhanced image to obtain the texture gradient information of the ball screw pair;

[0021] Structural morphology analysis is performed on the enhanced image to obtain the topological information of the ball screw pair;

[0022] By integrating the geometric contour information, the texture gradient information, and the topological structure information, a multidimensional feature image set of the ball screw pair is obtained.

[0023] In a preferred embodiment, when the reference image generation module performs the reverse mapping of the vector geometric elements and tolerance bands of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair, it is specifically used for:

[0024] The design drawings of the ball screw assembly are analyzed to obtain the vector geometric elements and tolerance zones of the design drawings;

[0025] Perform multi-view geometric analysis on the enhanced image to obtain the actual imaging perspective of the enhanced image;

[0026] The spatial coordinate system of the enhanced image is determined based on the enhanced image and the actual imaging viewpoint;

[0027] Based on the actual imaging perspective and spatial coordinate system, construct the coordinate mapping relationship between the design drawing and the enhanced image;

[0028] Based on the coordinate mapping relationship, the vector geometric elements and the annotation tolerance zone are mapped to the pixel position set and allowable pixel offset range of the ball screw pair;

[0029] Image reconstruction is performed on the pixel position set and the allowed pixel offset range to obtain the optical simulation reference image of the ball screw pair.

[0030] In a preferred embodiment, when the reference image generation module constructs the coordinate mapping relationship between the design drawing and the enhanced image based on the actual imaging viewpoint and spatial coordinate system, it is specifically used for:

[0031] Optimal spatial transformation estimation is performed on the enhanced image and the design drawing to obtain the spatial alignment parameters of the ball screw pair;

[0032] The image ratio of the enhanced image and the design drawing is used as the conversion ratio of the ball screw pair;

[0033] By associating and coupling the spatial alignment parameters and the transformation ratio, the coordinate mapping relationship between the design drawing and the enhanced image is obtained.

[0034] In a preferred embodiment, when the defect image determination module performs causal inference on the optical simulation reference image and the multidimensional feature image set, and determines the initial defect map of the ball screw pair based on the inference result, it is specifically used for:

[0035] The optical simulation reference image and the multi-dimensional feature image set are compared dimension by dimension to obtain the feature differences between the optical simulation reference image and the multi-dimensional feature image set;

[0036] The feature differences are normalized and then fused to obtain the preliminary comprehensive difference value of the ball screw pair.

[0037] The spatial distribution pattern of the preliminary differential composite value is analyzed, and the consistency of the spatial distribution pattern is evaluated to obtain the morphological consistency coefficient of the ball screw pair.

[0038] The defect confidence level of the ball screw pair is assessed based on the preliminary differential composite value and the morphological consistency coefficient.

[0039] Based on the defect confidence level, salient regions are extracted from the multidimensional feature image set to obtain the initial defect map.

[0040] In a preferred embodiment, the formula for calculating the defect confidence level is:

[0041] ;

[0042] in, This indicates the confidence level of the defect. Indicates the first Normalized feature differences across each feature dimension Indicates the first Preset weighting factors for each feature dimension, Indicates the total number of feature dimensions. This represents the preliminary composite value of the differences. This represents the threshold value for the allowed pixel offset range. This represents the preset positive adjustment coefficient. This represents the morphological consistency coefficient. This represents an exponential function.

[0043] In a preferred embodiment, when the defect region localization module performs conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair, it is specifically used for:

[0044] Identify potential defect regions in the initial defect map and perform structured analysis on the potential defect regions to obtain their spatial boundaries and confidence attributes.

[0045] Based on the confidence attribute, the potential defect regions are prioritized to obtain the candidate defect regions of the ball screw pair;

[0046] Using the spatial boundary as an initial guide, the candidate defect region is refined to obtain the precise candidate region of the ball screw pair;

[0047] The distribution consistency of the precise candidate regions is checked to obtain the defect image region of the ball screw pair.

[0048] In a preferred embodiment, when the monitoring report generation module performs multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair, it is specifically used for:

[0049] Based on the defective image region, correlation analysis is performed on the multidimensional feature image set to obtain the correlation relationship of the ball screw pair;

[0050] Based on the aforementioned correlation, cluster matching is performed on the defect image region to obtain the defect pattern category of the ball screw pair;

[0051] A comprehensive judgment is made on the defect image region and the defect pattern category to obtain a defect detection report for the ball screw pair.

[0052] To address the above problems, the present invention also provides a method for detecting precision ball screw pairs, the method comprising:

[0053] S1. Perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair;

[0054] S2. Perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair;

[0055] S3. Reverse map the vector geometric elements and tolerance zones of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair.

[0056] S4. Perform causal inference on the optical simulation reference image and the multidimensional feature image set, and determine the initial defect spectrum of the ball screw pair based on the inference result;

[0057] S5. Based on the initial defect map, conditional region decoupling is performed on the enhanced image to obtain the defect image region of the ball screw pair;

[0058] S6. Based on the defective image region, perform multivariate correlation analysis on the multidimensional feature image set to generate a defect detection report for the ball screw pair.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] 1. This invention performs visual enhancement processing on the multi-angle original image sequence of the ball screw assembly, and obtains a comprehensive multi-dimensional feature image set by combining multi-dimensional feature extraction. Then, it reverse maps the vector geometric elements of the design drawing with the marked tolerance zone to generate an optical simulation reference image. By determining the initial defect map through causal inference and accurately locating the defect area, it significantly improves the accuracy of defect identification and ensures that it can comprehensively and accurately capture various defect information of the ball screw assembly, making the detection results more reliable.

[0061] 2. This invention significantly improves the efficiency of inspection work through a seamless automated processing flow, from image enhancement to multivariate correlation analysis to generate inspection reports, completing the entire inspection process without complex manual intervention. Simultaneously, the generated defect inspection report integrates defect pattern categories and related information, providing comprehensive and effective data support for the quality assessment and subsequent processing of ball screw assemblies, further enhancing the practicality and application value of the inspection technology. Attached Figure Description

[0062] Figure 1 A system architecture diagram of a precision ball screw pair detection system provided in an embodiment of the present invention;

[0063] Figure 2 This is a flowchart illustrating a method for detecting a precision ball screw pair according to an embodiment of the present invention.

[0064] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.

[0067] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0068] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.

[0069] In practice, the server-side equipment deployed in a precision ball screw pair testing system may consist of one or more devices. This precision ball screw pair testing system can be implemented as: a business instance, a virtual machine, or hardware devices. For example, this precision ball screw pair testing system can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this precision ball screw pair testing system can be understood as software deployed on a cloud node, used to provide a precision ball screw pair testing system to various user terminals. Alternatively, this precision ball screw pair testing system can also be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing various user terminals. Alternatively, this precision ball screw pair testing system can also be implemented as a server composed of numerous identical or different types of hardware devices, with one or more hardware devices configured to provide a precision ball screw pair testing system to various user terminals.

[0070] In terms of implementation, the precision ball screw pair detection system and the user terminal are mutually compatible. That is, if the precision ball screw pair detection system is implemented as an application installed on a cloud service platform, the user terminal is implemented as a client that establishes a communication connection with the application; or if the precision ball screw pair detection system is implemented as a website, the user terminal is implemented as a webpage; or if the precision ball screw pair detection system is implemented as a cloud service platform, the user terminal is implemented as a mini-program in an instant messaging application.

[0071] like Figure 1 The figure shown is a system architecture diagram of a precision ball screw pair detection system provided in an embodiment of the present invention.

[0072] The precision ball screw pair detection system 100 described in this invention can be installed on a cloud server. In terms of implementation, it can be used as one or more service devices, or as an application installed on the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed into a website. Depending on the functions implemented, the precision ball screw pair detection system 100 may include an image enhancement module 101, a feature extraction module 102, a reference image generation module 103, a defect image determination module 104, a defect area localization module 105, and a monitoring report generation module 106. The modules described in this invention can also be called units, referring to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.

[0073] In this embodiment of the invention, in a precision ball screw pair testing system, each of the above-mentioned modules can be implemented independently and called upon other modules. This "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. The precision ball screw pair testing system provided by this embodiment of the invention allows for adjustment of the system's applicability by adding modules and directly calling them, without modifying the program code. This enables cluster-based horizontal expansion, facilitating quick and flexible expansion of the precision ball screw pair testing system. In practical applications, the above modules can be located in the same or different devices, or in virtual devices, such as service instances on a cloud server.

[0074] The following describes, with reference to specific embodiments, each component and its specific workflow of a precision ball screw pair detection system:

[0075] The image enhancement module 101 is used to perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair.

[0076] In this embodiment of the invention, when the image enhancement module performs visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair, it is specifically used for:

[0077] Acquire multi-angle raw image sequences of the ball screw assembly;

[0078] Feature filtering is performed on the multi-angle original image sequence to obtain the initial image set of the ball screw pair;

[0079] The initial image set is fused across different viewpoints to obtain the fused image of the ball screw pair;

[0080] Spatial structure analysis is performed on the fused image to obtain the grayscale distribution features of the fused image;

[0081] Based on the grayscale distribution characteristics, the fused image is adaptively intensity adjusted to obtain an enhanced image of the ball screw pair.

[0082] By deploying multiple high-resolution industrial cameras, images are acquired from different directions, including the circumferential, axial, and radial directions, around the ball screw assembly. The shooting angle of each camera is fixed and does not overlap, ensuring that the structural information of each surface of the ball screw assembly can be completely captured, and finally obtaining a multi-angle original image sequence covering the overall appearance of the ball screw assembly.

[0083] Each of the acquired multi-angle raw image sequences was examined, with a focus on judging the clarity, contrast, and completeness of the key structure of the ball screw pair. Images that were blurry, heavily reflective, obscured by foreign objects, or missing key structures were removed. All images that met the requirements of clarity and completeness were selected to form the initial image set of the ball screw pair.

[0084] First, image registration is performed on each image in the initial image set. By identifying the common feature points of the ball screw pair in each image, the positional correspondence between images from different perspectives is determined. Then, the pixel information of the corresponding positions in each image is integrated to retain the clear structural details in each image, make up for the information blind spots that may exist under a single perspective, and form a fused image that can fully reflect the three-dimensional structural features of the ball screw pair.

[0085] The fused image is scanned row by row and column by column to analyze the pixel arrangement pattern of each structural region of the ball screw pair in the image, clarify the gray value range corresponding to different structures, and record the gray value change trend in the image, such as the gray value gradient from the top of the thread to the bottom of the thread, so as to obtain the gray value distribution characteristics of the fused image completely.

[0086] Based on the obtained grayscale distribution characteristics, regions with excessively low grayscale values ​​leading to blurred details and regions with excessively high grayscale values ​​leading to loss of details in the fused image are identified. For regions with excessively low grayscale values, the grayscale intensity of each pixel in that region is gradually increased. For regions with excessively high grayscale values, the grayscale intensity of each pixel in that region is appropriately decreased. Throughout the adjustment process, the goal is to maintain the grayscale differences and detail integrity of each structural region, ultimately resulting in an enhanced image of a ball screw pair with a clear structure and balanced grayscale.

[0087] The beneficial effects are that through a coherent process of multi-dimensional acquisition, screening, fusion, analysis and adaptive adjustment, invalid image information is effectively eliminated, comprehensive and clear structural data is integrated, the image clarity and contrast are improved, the key structural features of the ball screw pair are fully preserved, and a high-quality image foundation is provided for subsequent multi-dimensional feature extraction, ensuring that the subsequent detection process can accurately capture relevant feature information.

[0088] The feature extraction module 102 is used to perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair;

[0089] In this embodiment of the invention, when the feature extraction module performs multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair, it is specifically used for:

[0090] Edge detection is performed on the enhanced image to obtain the geometric contour information of the ball screw pair;

[0091] Based on the geometric contour information, multi-scale texture gradient analysis is performed on the enhanced image to obtain the texture gradient information of the ball screw pair;

[0092] Structural morphology analysis is performed on the enhanced image to obtain the topological information of the ball screw pair;

[0093] By integrating the geometric contour information, the texture gradient information, and the topological structure information, a multidimensional feature image set of the ball screw pair is obtained.

[0094] The enhanced image is scanned pixel by pixel to analyze the gray value difference between each pixel and its neighboring pixels. When there is a clear boundary between the gray values ​​of neighboring pixels, the boundary position is marked as an edge point. The entire enhanced image is continuously traversed to collect all edge points and connect them in order of their spatial position in the image to form a set of lines that can clearly show the outline of the ball screw pair, the thread helical trajectory, the end profile and the key structural boundary, thereby obtaining the geometric contour information of the ball screw pair.

[0095] Based on the obtained geometric contour information, the core area of ​​the analysis is clearly defined as the main body of the ball screw pair surrounded by the contour. Starting from different observation scales, the general trend of the overall texture change is first analyzed from a larger perspective, and then the observation range is gradually narrowed to focus on the subtle texture changes in local areas. The direction and magnitude of the gray value change at each position are recorded point by point, such as the direction of the gray gradient between the thread crest and the root, and the texture depth change pattern of the inner wall of the ball groove. By combining these texture change data at different scales, the texture gradient information of the ball screw pair is obtained.

[0096] By comprehensively observing the overall structure of the ball screw pair in the enhanced image, analyzing the morphological characteristics of each component and their interconnections, clarifying the helical shape of the thread structure, the cross-sectional shape of the ball groove, and the specific style of the end structure, and simultaneously sorting out the positional relationships between the structures, such as the connection method between the thread and the end, the uniform distribution of the ball groove, and the relative spacing of each key structure, it is determined which structures are continuous and which are independent. Through a systematic sorting out of these morphologies and relationships, the topological structure information of the ball screw pair is obtained.

[0097] The obtained geometric contour information, texture gradient information, and topological structure information are precisely mapped according to the pixel coordinate system of the enhanced image. The line data of the geometric contour, the change data of the texture gradient, and the relationship data of the topological structure are mapped to the corresponding pixel positions, so that each pixel contains contour belonging information, has texture gradient features, and is associated with topological structure relationships. Through this superposition and integration of multi-dimensional information, a set that can comprehensively reflect the shape, texture, and structural relationship features of the ball screw pair is formed, resulting in the multi-dimensional feature image set of the ball screw pair.

[0098] The beneficial effects are that by extracting three core features—geometric contour, texture gradient, and topological structure—in a step-by-step manner, the comprehensiveness and relevance of feature extraction are ensured. The feature information complements each other and corresponds precisely. The resulting multi-dimensional feature image set completely preserves the key structural features of the ball screw pair, providing rich and accurate feature basis for subsequent comparison with optical simulation benchmark images and defect identification, effectively improving the accuracy of subsequent detection steps.

[0099] The reference image generation module 103 is used to reverse map the vector geometric elements and tolerance bands of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair.

[0100] In this embodiment of the invention, when the reference image generation module performs the reverse mapping of the vector geometric elements and tolerance bands of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair, it is specifically used for:

[0101] The design drawings of the ball screw assembly are analyzed to obtain the vector geometric elements and tolerance zones of the design drawings;

[0102] Perform multi-view geometric analysis on the enhanced image to obtain the actual imaging perspective of the enhanced image;

[0103] The spatial coordinate system of the enhanced image is determined based on the enhanced image and the actual imaging viewpoint;

[0104] Based on the actual imaging perspective and spatial coordinate system, construct the coordinate mapping relationship between the design drawing and the enhanced image;

[0105] Based on the coordinate mapping relationship, the vector geometric elements and the annotation tolerance zone are mapped to the pixel position set and allowable pixel offset range of the ball screw pair;

[0106] Image reconstruction is performed on the pixel position set and the allowed pixel offset range to obtain the optical simulation reference image of the ball screw pair.

[0107] When the reference image generation module constructs the coordinate mapping relationship between the design drawing and the enhanced image based on the actual imaging viewpoint and spatial coordinate system, it is specifically used for:

[0108] Optimal spatial transformation estimation is performed on the enhanced image and the design drawing to obtain the spatial alignment parameters of the ball screw pair;

[0109] The image ratio of the enhanced image and the design drawing is used as the conversion ratio of the ball screw pair;

[0110] By associating and coupling the spatial alignment parameters and the transformation ratio, the coordinate mapping relationship between the design drawing and the enhanced image is obtained.

[0111] The digital file of the ball screw pair design drawing is read, and the graphic elements in the file are identified one by one. The vector geometric elements such as lines, curves, arcs, and polygons that constitute the ball screw pair structure are extracted. At the same time, various tolerance information such as dimensional tolerances, geometric tolerances, and surface roughness tolerances marked on the drawing are sorted out. These tolerance information are integrated to form a complete annotation tolerance zone, and finally the vector geometric elements and annotation tolerance zone of the design drawing are obtained.

[0112] By observing the projection shape of each structure of the ball screw pair in the enhanced image, analyzing the distortion and relative positional relationship of key structures such as threads and ball grooves in the image, and combining the camera installation layout information during image acquisition, the actual shooting angle corresponding to the enhanced image is determined, including horizontal angle, vertical angle and pitch angle, thereby obtaining the actual imaging angle of the enhanced image.

[0113] Using the center of the end of the ball screw pair in the enhanced image as the origin of the coordinate system, the x-axis is set to be parallel to the width direction of the enhanced image, the y-axis is set to be parallel to the height direction of the enhanced image, and the z-axis is set to be perpendicular to the plane of the enhanced image and pointing towards the imaging lens. A three-dimensional spatial coordinate system is established. This coordinate system can accurately describe the spatial position of each pixel in the enhanced image, thereby determining the spatial coordinate system of the enhanced image.

[0114] By comparing the structural morphology of the ball screw pair in the enhanced image with the geometric structure of the design drawing, different spatial transformation methods are tried to find the transformation method that maximizes the overlap between the geometric contour of the design drawing and the contour of the ball screw pair in the enhanced image. The adjustment parameters corresponding to this transformation method are recorded, and these parameters are the spatial alignment parameters of the ball screw pair.

[0115] The number of pixels corresponding to the key structural dimensions of the ball screw pair in the enhanced image is measured, and the annotation dimensions of the same key structure on the design drawing are read. The ratio of the number of pixels in the enhanced image to the unit physical dimension on the design drawing is calculated. This ratio is the conversion ratio of the ball screw pair.

[0116] By integrating the position adjustment rules specified by the spatial alignment parameters with the size correspondence rules determined by the conversion ratio, it is clear how any geometric coordinate point in the design drawing can be converted into a pixel coordinate point in the enhanced image through the conversion ratio after spatial alignment adjustment, forming a complete set of coordinate correspondence rules. Through this set of rules, the coordinates between the design drawing and the enhanced image can be mutually converted, and the coordinate mapping relationship between the design drawing and the enhanced image can be obtained.

[0117] According to the established coordinate mapping relationship, the coordinate points corresponding to each vector geometric element in the design drawing are converted into specific pixel coordinates in the enhanced image pixel grid. All converted pixel coordinates are summarized to form the pixel position set of the ball screw pair. At the same time, the physical dimension deviation range corresponding to the tolerance zone marked in the design drawing is converted into the pixel deviation range in the enhanced image pixel grid through a conversion ratio. This range is the allowable pixel offset range of the ball screw pair.

[0118] Based on the pixel position set, a vector geometric structure consistent with the design drawing is drawn in the pixel grid of the enhanced image. Then, according to the allowable pixel offset range, the allowable deviation area is marked around the drawn geometric structure to ensure that the marked deviation area can accurately reflect the tolerance range required by the design. In this way, the vector geometric elements and the allowable pixel offset range are fully presented in the image, completing the image reconstruction process and obtaining the optical simulation reference image of the ball screw pair.

[0119] The beneficial effect is that through a coherent and accurate process of analysis, coordinate establishment, mapping and reconstruction, the optical simulation reference image can accurately reproduce the design standard of the ball screw pair and maintain a high degree of consistency with the enhanced image in terms of viewpoint and scale. This provides an accurate and reliable reference for the subsequent comparison between the optical simulation reference image and the multi-dimensional feature image set, effectively ensuring the accuracy and rationality of subsequent defect identification.

[0120] The defect image determination module 104 is used to perform causal inference on the optical simulation reference image and the multi-dimensional feature image set, and determine the initial defect map of the ball screw pair based on the inference result.

[0121] In this embodiment of the invention, when the defect image determination module performs causal inference on the optical simulation reference image and the multi-dimensional feature image set, and determines the initial defect map of the ball screw pair based on the inference result, it is specifically used for:

[0122] The optical simulation reference image and the multi-dimensional feature image set are compared dimension by dimension to obtain the feature differences between the optical simulation reference image and the multi-dimensional feature image set;

[0123] The feature differences are normalized and then fused to obtain the preliminary comprehensive difference value of the ball screw pair.

[0124] The spatial distribution pattern of the preliminary differential composite value is analyzed, and the consistency of the spatial distribution pattern is evaluated to obtain the morphological consistency coefficient of the ball screw pair.

[0125] The defect confidence level of the ball screw pair is assessed based on the preliminary differential composite value and the morphological consistency coefficient.

[0126] Based on the defect confidence level, salient regions are extracted from the multidimensional feature image set to obtain the initial defect map.

[0127] The formula for calculating the defect confidence level is as follows:

[0128] ;

[0129] in, This indicates the confidence level of the defect. Indicates the first Normalized feature differences across each feature dimension Indicates the first Preset weighting factors for each feature dimension, Indicates the total number of feature dimensions. This represents the preliminary composite value of the differences. This represents the threshold value for the allowed pixel offset range. This represents the preset positive adjustment coefficient. This represents the morphological consistency coefficient. This represents an exponential function.

[0130] Each feature dimension of the optical simulation reference image and the multi-dimensional feature image set is compared one by one. First, for the geometric contour feature dimension, the differences between the two in the position, shape, and continuity of the contour lines are compared. Then, for the texture gradient feature dimension, the differences between the two in the direction and magnitude of texture changes are compared. Next, for the topological structure feature dimension, the differences between the two in the connection relationships and positional distribution of each structure are checked. In this way, a comprehensive comparison of all feature dimensions is completed, and the differences under all dimensions are collected and sorted out to obtain the feature differences between the optical simulation reference image and the multi-dimensional feature image set.

[0131] For the feature differences under each feature dimension, different types of differences are converted into directly comparable descriptive values ​​according to a unified standard, eliminating the magnitude differences between differences in different feature dimensions and bringing the differences in all feature dimensions within the same comparable range, thus completing the normalization process for the feature differences. Then, all normalized feature differences are comprehensively summarized, and the impact of each feature dimension difference on the overall detection result is considered to form a numerical value that can comprehensively reflect the degree of difference between the two, obtaining the preliminary comprehensive difference value of the ball screw pair.

[0132] Observe the distribution of the preliminary differential composite values ​​in the spatial location corresponding to the multidimensional feature image set to clarify whether these differential values ​​are concentrated in a specific region, continuously distributed in multiple related regions, or randomly discretely distributed, thus clearly understanding the spatial distribution pattern of the preliminary differential composite values. Subsequently, determine whether this distribution pattern conforms to the morphological characteristics typically possessed by defects, such as the fact that the differences corresponding to defects are often concentrated and have a certain degree of continuity. Based on this degree of conformity, a quantitative evaluation is performed to obtain the morphological consistency coefficient of the ball screw pair that reflects the rationality of the differential distribution.

[0133] The preliminary difference composite value and the morphological consistency coefficient are used as core considerations. The magnitude of the preliminary difference composite value directly reflects the degree of overall difference between the two; a larger value indicates a higher probability of defect existence. The morphological consistency coefficient reflects the degree of fit between the difference distribution and the defect morphology; a higher coefficient indicates that the difference is more likely to be caused by a defect. By comprehensively weighing the influence of these two factors, the probability of defects in the corresponding area is fully assessed, ultimately yielding the defect confidence level of the ball screw pair.

[0134] A clear defect confidence level is established, and areas whose defect confidence level meets the standard are selected. These areas are considered significant areas with a high probability of defect presence. In a multi-dimensional feature image set, the location and extent of these significant areas are precisely marked, clearly showing the boundary and coverage of each significant area. All marked significant areas are integrated to form a map that intuitively reflects the location and approximate extent of potential defects, thus obtaining the initial defect map of the ball screw pair.

[0135] The normalized feature differences in the feature dimensions are obtained by comparing the optical simulation benchmark image and the multi-dimensional feature image set one feature dimension at a time, and then normalizing these feature differences.

[0136] The preset weight factor is a pre-set weight value corresponding to each feature dimension.

[0137] The total number of feature dimensions is the total number of feature dimensions contained in a multidimensional feature image set that integrates geometric contour information, texture gradient information, and topological structure information.

[0138] The preliminary differential composite value is the result obtained by fusing the normalized feature differences.

[0139] The threshold for the allowable pixel offset range is derived from the coordinate mapping relationship between the design drawing and the enhanced image. The threshold corresponding to the allowable pixel offset range is obtained by mapping the vector geometric elements of the design drawing to the tolerance zone of the annotation. The coordinate mapping relationship is constructed by estimating the spatial alignment parameter by performing the optimal spatial transformation on the enhanced image and the design drawing, and combining the image ratio of the two as the transformation ratio. The spatial alignment parameter and the transformation ratio are then coupled together.

[0140] The preset positive adjustment coefficient is a pre-set positive value used to adjust the degree of influence of related items.

[0141] The morphological consistency coefficient is the result obtained by analyzing the spatial distribution pattern of the preliminary differential composite value and then evaluating the consistency of the spatial distribution pattern.

[0142] The calculation process first multiplies the normalized feature difference of each feature dimension by the corresponding preset weight factor, and then adds all the product results to obtain a value that comprehensively considers the differences and importance of each feature dimension. Next, the ratio of the preliminary difference comprehensive value to the threshold of the allowable pixel offset range is calculated. This ratio is squared and multiplied by a preset positive adjustment coefficient. The negative value of the calculation result is taken and exponentially calculated to obtain the value that constrains the impact of the preliminary difference comprehensive value exceeding the allowable range. Finally, the two values ​​obtained above are multiplied by the morphological consistency coefficient to obtain a value that can accurately reflect the existence of defects in the ball screw pair and the degree of reliability corresponding to the defect.

[0143] When the difference in normalized features of each feature dimension increases, the sum of the products of the difference in normalized features of each feature dimension and the corresponding preset weight factor will increase, thereby increasing the overall result.

[0144] When the preset weight factor increases, the sum of the products of the normalized feature differences of each feature dimension and the corresponding preset weight factor will increase, thus increasing the overall result.

[0145] When the initial difference composite value increases, the ratio of the initial difference composite value to the threshold of the allowable pixel offset range will increase. The square of this ratio multiplied by the preset positive adjustment coefficient will increase the value, while its negative value will decrease. The result of the exponential operation will decrease, thus reducing the overall result.

[0146] As the morphological consistency coefficient increases, the overall result will also increase.

[0147] When the preset positive adjustment coefficient increases, the square of the ratio of the initial difference comprehensive value to the threshold of the allowable pixel offset range multiplied by the coefficient increases, its negative value decreases, the result of the exponential operation decreases, and thus the overall result decreases.

[0148] When the threshold of the allowed pixel offset range increases, the ratio of the initial difference composite value to the threshold of the allowed pixel offset range decreases. The square of this ratio multiplied by the preset positive adjustment coefficient decreases, while its negative value increases. The result of the exponential operation increases, which in turn increases the overall result.

[0149] The beneficial effects are that through a coherent process of dimensional comparison, normalization fusion, distribution pattern analysis, confidence assessment, and significant region extraction, the differences between the optical simulation benchmark image and the multidimensional feature image set are accurately captured. The magnitude and distribution rationality of the differences are comprehensively considered, ensuring that the initial defect map can accurately locate the position and range of potential defects. This provides a reliable and effective basis for the subsequent accurate positioning of defect areas, and improves the pertinence and accuracy of defect detection.

[0150] The defect region localization module 105 is used to perform conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair.

[0151] In this embodiment of the invention, when the defect region localization module performs conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair, it is specifically used for:

[0152] Identify potential defect regions in the initial defect map and perform structured analysis on the potential defect regions to obtain their spatial boundaries and confidence attributes.

[0153] Based on the confidence attribute, the potential defect regions are prioritized to obtain the candidate defect regions of the ball screw pair;

[0154] Using the spatial boundary as an initial guide, the candidate defect region is refined to obtain the precise candidate region of the ball screw pair;

[0155] The distribution consistency of the precise candidate regions is checked to obtain the defect image region of the ball screw pair.

[0156] The system iterates through all marked salient regions in the initial defect map. These regions are the potential defect regions that need further analysis. Then, it performs structured analysis on each potential defect region, and clarifies its spatial boundary by determining the pixel coordinate range of the region's edge. At the same time, it extracts the defect confidence value corresponding to each region, which directly reflects the probability of the region having a defect. Finally, it obtains the spatial boundary and confidence attribute of the potential defect region.

[0157] Using the confidence level of each potential defect area as the core judgment criterion, all potential defect areas are sorted in descending order of confidence level. The higher the confidence level, the greater the probability that the area has a real defect. The top-ranked areas are selected as the objects of subsequent key analysis to obtain the candidate defect areas of the ball screw pair.

[0158] Using the determined spatial boundary of the candidate defect region as an initial reference, the specific location corresponding to the candidate defect region is located in the enhanced image. The pixels near the boundary are carefully examined point by point to observe whether the gray value changes and structural features of the pixels match the defect edge features. The pixel coordinates of the boundary are adjusted according to the actual pixel features so that the boundary can accurately fit the actual edge contour of the defect, eliminating possible deviations in the initial boundary, and obtaining the accurate candidate region of the ball screw pair.

[0159] Observe the spatial distribution of all precise candidate regions in the enhanced image, analyze whether there is a correlation in position, similarity in shape, or regularity in distribution between the regions, determine whether the distribution of these regions conforms to the distribution characteristics of common defects of ball screw pairs, eliminate those precise candidate regions that exist in isolation, have no correlation, and do not conform to the distribution rules of common defects, retain regions that are reasonably distributed and have consistent characteristics, and obtain the defect image region of the ball screw pair.

[0160] The beneficial effects are that by using a multi-layered screening and precise optimization process from potential defect area identification to final defect image area determination, the interference of false defect signals is effectively eliminated, ensuring that the boundaries of defect areas are accurate and the distribution is reasonable. This provides accurate and reliable defect area data for subsequent defect pattern recognition and detection report generation, further improving the detection accuracy and reliability of the entire detection system.

[0161] The monitoring report generation module 106 is used to perform multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair.

[0162] In this embodiment of the invention, when the monitoring report generation module performs multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair, it is specifically used for:

[0163] Based on the defective image region, correlation analysis is performed on the multidimensional feature image set to obtain the correlation relationship of the ball screw pair;

[0164] Based on the aforementioned correlation, cluster matching is performed on the defect image region to obtain the defect pattern category of the ball screw pair;

[0165] A comprehensive judgment is made on the defect image region and the defect pattern category to obtain a defect detection report for the ball screw pair.

[0166] Focusing on the identified defective image regions, geometric contour information, texture gradient information, and topological structure information are extracted from the multidimensional feature image set corresponding to each region. The mutual influence between different features within the same defective image region is analyzed. For example, whether local deformation of the geometric contour causes texture gradient anomalies at the corresponding location, and whether the fracture of a certain structure in the topological structure is related to changes in the surrounding geometric contours and texture features. At the same time, the correlation between feature changes between different defective image regions is explored. For example, whether multiple defective regions have the same abnormal feature trend. Through this comprehensive correlation analysis, the intrinsic connection between features and defects, and between defects, is sorted out, and the correlation relationship of the ball screw pair is obtained.

[0167] Based on the obtained correlations, defect image regions with the same or similar feature correlation patterns are selected and classified into the same category. For example, all defect image regions with abrupt changes in texture gradient due to geometric contour depression and no obvious breakage in topological structure are classified into one category. Then, the correlation feature patterns of each category are compared with a preset common defect pattern library. The common defect pattern library contains correlation feature models corresponding to typical defects such as cracks, wear, deformation, and material shortage. Through accurate matching of feature patterns, the specific defect type corresponding to each type of defect image region is determined, and the defect pattern category of the ball screw pair is obtained.

[0168] The system integrates specific information about the defect image region, including the spatial location, boundary range, and coverage area of ​​each defect. Combined with the corresponding defect pattern category, it analyzes the severity of each defect. For example, the severity level of crack defects is determined based on the abnormality of their length, width, and associated features. The impact of wear defects is assessed based on the abnormal range of texture gradient and the wear depth of the geometric contour. At the same time, the number, distribution, pattern category, and severity of all defects are summarized and organized according to a unified report format. All detection data and analysis results are clearly recorded to form a document containing comprehensive defect information and professional assessment, resulting in the defect detection report of the ball screw pair.

[0169] The beneficial effects are that by deeply exploring the intrinsic connections between features and defects, and between defects themselves through correlation analysis, and by accurately defining defect pattern categories through clustering matching, and then comprehensively judging and integrating information such as the location, scope, and severity of defects, the generated defect detection report is detailed, logically clear, and data accurate. It not only clearly presents the defect status of the ball screw assembly, but also provides scientific and practical guidance for subsequent quality judgment, maintenance, and production process optimization, significantly improving the application value and decision-making reference significance of the detection results.

[0170] Reference Figure 2 The diagram shown is a flowchart illustrating a precision ball screw pair testing method according to an embodiment of the present invention. In this embodiment, the precision ball screw pair testing method includes:

[0171] S1. Perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair;

[0172] S2. Perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair;

[0173] S3. Reverse map the vector geometric elements and tolerance zones of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair.

[0174] S4. Perform causal inference on the optical simulation reference image and the multidimensional feature image set, and determine the initial defect spectrum of the ball screw pair based on the inference result;

[0175] S5. Based on the initial defect map, conditional region decoupling is performed on the enhanced image to obtain the defect image region of the ball screw pair;

[0176] S6. Based on the defective image region, perform multivariate correlation analysis on the multidimensional feature image set to generate a defect detection report for the ball screw pair.

[0177] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0178] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A precision ball screw pair detection system, characterized in that, The system includes an image enhancement module, a feature extraction module, a reference image generation module, a defect image determination module, a defect region localization module, and a monitoring report generation module, wherein: The image enhancement module is used to perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair. The feature extraction module is used to perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair; The reference image generation module is used to reverse map the vector geometric elements and tolerance bands of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair. The defect image determination module is used to perform causal inference on the optical simulation reference image and the multi-dimensional feature image set, and determine the initial defect map of the ball screw pair based on the inference result. The defect region localization module is used to perform conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair. The monitoring report generation module is used to perform multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair.

2. The precision ball screw pair detection system as described in claim 1, characterized in that, When the image enhancement module performs visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair, it is specifically used for: Acquire multi-angle raw image sequences of the ball screw assembly; Feature filtering is performed on the multi-angle original image sequence to obtain the initial image set of the ball screw pair; The initial image set is fused across different viewpoints to obtain the fused image of the ball screw pair; Spatial structure analysis is performed on the fused image to obtain the grayscale distribution features of the fused image; Based on the grayscale distribution characteristics, the fused image is adaptively intensity adjusted to obtain an enhanced image of the ball screw pair.

3. The precision ball screw pair detection system as described in claim 1, characterized in that, When the feature extraction module performs multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair, it is specifically used for: Edge detection is performed on the enhanced image to obtain the geometric contour information of the ball screw pair; Based on the geometric contour information, multi-scale texture gradient analysis is performed on the enhanced image to obtain the texture gradient information of the ball screw pair; Structural morphology analysis is performed on the enhanced image to obtain the topological information of the ball screw pair; By integrating the geometric contour information, the texture gradient information, and the topological structure information, a multidimensional feature image set of the ball screw pair is obtained.

4. The precision ball screw pair detection system as described in claim 1, characterized in that, When the reference image generation module performs the reverse mapping of the vector geometric elements and tolerance zones of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair, it is specifically used for: The design drawings of the ball screw assembly are analyzed to obtain the vector geometric elements and tolerance zones of the design drawings; Perform multi-view geometric analysis on the enhanced image to obtain the actual imaging perspective of the enhanced image; The spatial coordinate system of the enhanced image is determined based on the enhanced image and the actual imaging viewpoint; Based on the actual imaging perspective and spatial coordinate system, construct the coordinate mapping relationship between the design drawing and the enhanced image; Based on the coordinate mapping relationship, the vector geometric elements and the annotation tolerance zone are mapped to the pixel position set and allowable pixel offset range of the ball screw pair; Image reconstruction is performed on the pixel position set and the allowed pixel offset range to obtain the optical simulation reference image of the ball screw pair.

5. The precision ball screw pair detection system as described in claim 4, characterized in that, When the reference image generation module constructs the coordinate mapping relationship between the design drawing and the enhanced image based on the actual imaging viewpoint and spatial coordinate system, it is specifically used for: Optimal spatial transformation estimation is performed on the enhanced image and the design drawing to obtain the spatial alignment parameters of the ball screw pair; The image ratio of the enhanced image and the design drawing is used as the conversion ratio of the ball screw pair; By associating and coupling the spatial alignment parameters and the transformation ratio, the coordinate mapping relationship between the design drawing and the enhanced image is obtained.

6. The precision ball screw pair detection system as described in claim 1, characterized in that, When the defect image determination module performs causal inference on the optical simulation reference image and the multi-dimensional feature image set, and determines the initial defect map of the ball screw pair based on the inference result, it is specifically used for: The optical simulation reference image and the multi-dimensional feature image set are compared dimension by dimension to obtain the feature differences between the optical simulation reference image and the multi-dimensional feature image set; The feature differences are normalized and then fused to obtain the preliminary comprehensive difference value of the ball screw pair. The spatial distribution pattern of the preliminary differential composite value is analyzed, and the consistency of the spatial distribution pattern is evaluated to obtain the morphological consistency coefficient of the ball screw pair. The defect confidence level of the ball screw pair is assessed based on the preliminary differential composite value and the morphological consistency coefficient. Based on the defect confidence level, salient regions are extracted from the multidimensional feature image set to obtain the initial defect map.

7. The precision ball screw pair detection system as described in claim 6, characterized in that, The formula for calculating the defect confidence level is as follows: ； in, This indicates the confidence level of the defect. Indicates the first Normalized feature differences across each feature dimension Indicates the first Preset weighting factors for each feature dimension, Indicates the total number of feature dimensions. This represents the preliminary composite value of the differences. This represents the threshold value for the allowed pixel offset range. This represents the preset positive adjustment coefficient. This represents the morphological consistency coefficient. This represents an exponential function.

8. The precision ball screw pair detection system as described in claim 1, characterized in that, When the defect region localization module performs conditional region decoupling on the enhanced image based on the initial defect map to obtain the defect image region of the ball screw pair, it is specifically used for: Identify potential defect regions in the initial defect map and perform structured analysis on the potential defect regions to obtain their spatial boundaries and confidence attributes. Based on the confidence attribute, the potential defect regions are prioritized to obtain the candidate defect regions of the ball screw pair; Using the spatial boundary as an initial guide, the candidate defect region is refined to obtain the precise candidate region of the ball screw pair; The distribution consistency of the precise candidate regions is checked to obtain the defect image region of the ball screw pair.

9. The precision ball screw pair detection system as described in claim 1, characterized in that, When the monitoring report generation module performs multivariate correlation analysis on the multidimensional feature image set based on the defect image region to generate a defect detection report for the ball screw pair, it is specifically used for: Based on the defective image region, correlation analysis is performed on the multidimensional feature image set to obtain the correlation relationship of the ball screw pair; Based on the aforementioned correlation, cluster matching is performed on the defect image region to obtain the defect pattern category of the ball screw pair; A comprehensive judgment is made on the defect image region and the defect pattern category to obtain a defect detection report for the ball screw pair.

10. A method for detecting precision ball screw pairs, characterized in that, The method is used in the precision ball screw pair detection system according to claim 1. S1. Perform visual enhancement processing on the multi-angle original image sequence of the ball screw pair to obtain the enhanced image of the ball screw pair; S2. Perform multi-dimensional feature extraction on the enhanced image to obtain a multi-dimensional feature image set of the ball screw pair; S3. Reverse map the vector geometric elements and tolerance zones of the design drawing in the ball screw pair to the pixel grid of the enhanced image to obtain the optical simulation reference image of the ball screw pair. S4. Perform causal inference on the optical simulation reference image and the multidimensional feature image set, and determine the initial defect spectrum of the ball screw pair based on the inference result; S5. Based on the initial defect map, conditional region decoupling is performed on the enhanced image to obtain the defect image region of the ball screw pair; S6. Based on the defective image region, perform multivariate correlation analysis on the multidimensional feature image set to generate a defect detection report for the ball screw pair.

Images