Industrial product detection method and system based on image data analysis and processing

Abstract

The invention discloses an industrial product detection method and system based on image data analysis and processing, and particularly relates to the technical field of industrial product detection. According to the method, refined analysis logic is designed for three types of typical defects, linear damage of the raceway integrates length, width, depth and position parameters, surface peeling covers indexes such as a two-dimensional projection area, a three-dimensional volume and edge features, and quenching micro-cracks are subjected to quantitative analysis through a cooling rate anomaly ratio and a thermal diffusion ratio. The evaluation report integrates complete information such as bearing identification codes, timestamps, fault types, quantization parameters and causes, breaks through limitation of traditional qualitative evaluation, and provides data support for quality judgment, product tracing and production optimization.

Description

Technical Field

[0001] This invention relates to the field of industrial product testing technology, and more specifically, to an industrial product testing method and system based on image data analysis and processing. Background Technology

[0002] In the industrial manufacturing sector, bearings are core transmission components, and their quality directly determines the operating accuracy, stability, and service life of mechanical equipment. As the high-end equipment manufacturing industry rapidly develops towards high precision and high reliability, the market's requirements for bearing quality are becoming increasingly stringent, especially in key areas such as aerospace, precision machine tools, and new energy vehicles. Even minor bearing defects can lead to serious safety accidents and huge economic losses.

[0003] The current bearing production process encompasses multiple key steps, including grinding, heat treatment, assembly, and final inspection. Each step can produce typical defects such as raceway linear damage, surface spalling, and quenching microcracks due to factors such as fluctuations in process parameters, decreased equipment precision, and deviations in operating procedures. However, traditional bearing inspection methods still have the following shortcomings in practical applications: In traditional inspection, image data from different processes such as grinding, heat treatment, and assembly lack effective correlation and are mostly stored independently. When a defect is found, it is necessary to check the data of each process one by one, making it difficult to quickly locate the target process and original image corresponding to the defect, resulting in low traceability efficiency and delayed cause analysis. Defect assessments are mostly qualitative descriptions, lacking quantitative analysis of defect geometric parameters and physical characteristics, and failing to integrate key information such as defect location, edge features, and causes. As a result, assessment reports have limited support for production optimization.

[0004] To address this, an industrial product inspection method and system based on image data analysis and processing has been developed. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an industrial product inspection method and system based on image data analysis and processing.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Industrial product inspection methods based on image data analysis and processing include: Step 1: Collect multiple modal image data of the bearing at different process stages, including grinding, heat treatment, assembly and final inspection. Step 2: Using the bearing identification code as the primary key and the time sequence and process as the index, store the multi-modal image data of the bearing at different process stages in a hierarchical manner; Step 3: After the final inspection of the bearing, the management personnel conduct quality inspection on the bearing and selectively output the fault type and fault code of the bearing based on the quality inspection results; Step 4: In response to the fault code output by the bearing, use the bearing's identification code and fault code in the image database to locate and extract the target image data collected at the target process stage corresponding to the fault code. Step 5: Based on the extracted target image data and fault codes, perform collaborative analysis to generate a bearing defect assessment report.

[0007] Specifically, the logic for setting the fault type and its fault code; The types of failures include raceway linear damage, surface spalling, and quenching microcracks. Pre-set the fault codes corresponding to different fault types.

[0008] Specifically, the target image data collected at the target process stage corresponding to the fault code; The target image data extracted corresponding to the linear damage of the raceway is a microscopic image of the raceway surface of the bearing during the grinding process; The image data extracted from the surface peeling are the three-dimensional point cloud data of the raceway of the bearing in the assembly process and the microscopic image of the raceway surface in the grinding process. The extracted image data corresponding to the quenching microcracks is an infrared thermographic sequence of the bearing during the heat treatment process.

[0009] Specifically, the collaborative analysis logic for linear damage to the raceway; Adaptive threshold segmentation is performed on the microscopic image of the rolling surface to separate the defect region from the normal raceway surface, resulting in a binarized defect image. A single-pixel skeleton of the defect is extracted from the binarized image, with the skeleton representing the centerline of the defect. The number of pixels in the skeleton is counted, and the pixel length is converted to the actual physical length to obtain the defect length. A cross-section of the defect is cut along the normal direction of the defect skeleton with a set step size. The pixel width of the defect at each cross-section is counted, and the average pixel width is calculated. The average pixel width is then converted to the actual physical width to obtain the defect width. A standard grayscale value is preset, and the difference between the grayscale value of the defect area and the standard grayscale value is calculated as the grayscale difference; a mapping rule between the grayscale difference and depth is established and converted into defect depth. Based on the additional coefficients derived from the defect location, a weighted fusion is performed using defect length, defect width, and defect depth to output a linear damage assessment coefficient. The bearing's identification code, timestamp, fault type, defect location of raceway linear damage, and linear damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report.

[0010] Specifically, the conversion logic of the additional coefficient; Define the center coordinates of the raceway section, calculate the centroid coordinates of the defect area, calculate the distance from the centroid to the raceway center, and convert it to physical distance; Defect locations are determined by combining physical distance with pre-set location judgment criteria; curve locations include the central area, shoulder area, and intermediate transition area; and a set of additional coefficients are set for different defect locations.

[0011] Specifically, the collaborative analysis logic for surface peeling; A two-dimensional mask image of the peeled area was obtained by using a combination of grayscale thresholding and morphological segmentation algorithm on the microscopic image of the raceway surface. The number of defective pixels in the peeling area of ​​the mask image is counted, and combined with the pixel equivalent of the two-dimensional image, it is converted into the actual projected area. Extract the gradient magnitude image of the defect edge, count the average gradient magnitude of the defect edge region, and calculate the ratio of the average gradient of the defect edge region to the average gradient of the normal region, which is denoted as the gradient ratio. If the gradient ratio is greater than the preset reference gradient ratio, it is determined to be a clear brittle fracture edge; otherwise, it is determined to be a fuzzy plastic deformation edge. For the 3D point cloud data of the raceway, calculate the normal distance from each point cloud point in the peeling area to the preset ideal surface, define it as the single-point depth, and extract the maximum depth; generate a 3D mesh model of the peeling area, calculate the volume through mesh voxelization, and obtain the peeling volume; Two-dimensional cross-sectional contour lines are extracted along the axial and circumferential directions of the center of the spalling pit. The contour lines are composed of the point cloud depth values ​​within the cross-section. The edge steepness of the cross-sectional contour is calculated. The actual projected area, gradient ratio, maximum depth, peeling volume, and edge steepness are combined and weighted to output the peeling damage assessment coefficient. The bearing's identification code, timestamp, fault type, spalling pit shape, edge features, and spalling damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report. The edge features include clear brittle fracture edges and blurred plastic deformation edges.

[0012] Specifically, the edge features include clear brittle fracture edges and blurred plastic deformation edges; If the gradient ratio is greater than the preset reference gradient ratio, it is determined to be a clear brittle fracture edge; otherwise, it is determined to be a fuzzy plastic deformation edge. Logic for determining the shape of spalling pits; Extract the edge contour points of the curve, calculate the sum of the Euclidean distances of adjacent contour points, and convert it into the actual perimeter; use the reciprocal of the roundness to represent the irregularity, compare the irregularity with the set reference regularity, if the irregularity is greater than the reference regularity, it is determined that the spalling pit shape is irregular, and the feature is brittle fracture.

[0013] Specifically, the collaborative analysis logic for quenching microcracks; The digital output value of the thermal image is converted into the real temperature value to obtain the spatiotemporal temperature field data; the crack region is extracted by threshold segmentation, and the instantaneous cooling rate is calculated for each pixel in the crack region based on its temperature-time curve. The average cooling rate of the neighborhood is calculated by selecting a defined range of neighborhoods centered on each pixel. Define the cooling rate anomaly coefficient The calculated cooling rate anomaly coefficient for each pixel is compared with a set anomaly threshold. If a pixel exceeds the anomaly threshold, it is marked as a cooling rate anomaly point. The ratio of abnormal cooling rate points to the total number of points in the cracked region is used to obtain the rate anomaly ratio. Two sets of time points during the cooling process are selected, along with the average temperatures of the cracked region and the normal region. The thermal diffusivity of the cracked region and the normal region are calculated respectively. The diffusivity is obtained by calculating the ratio of the thermal diffusivity of the cracked region to the thermal diffusivity of the normal region. The cooling evaluation coefficient is output by weighted fusion of the rate anomaly ratio and the diffusion ratio. If the cooling evaluation coefficient is higher than the preset cooling threshold coefficient, the quenching microcracks are determined to be caused by cooling differences; otherwise, they are determined to be caused by unknown reasons. The bearing's identification code, timestamp, fault type, and the cause of the quenching microcracks, including cooling differences and unknown causes, are integrated into a pre-built report template to generate a defect assessment report.

[0014] Specifically, Definition logic; That is, to obtain by using the formula ,in The instantaneous cooling rate of the target pixel. The average cooling rate of the neighborhood centered on the target pixel; Logic for calculating thermal diffusivity; According to the formula Calculate the thermal diffusivity for the cracked and normal regions, respectively; r is the characteristic size of the region. The ambient temperature; For the two selected time sets, When calculating the thermal diffusivity of the cracked region, the average temperature of the cracked region at two different time points is represented. When calculating the thermal diffusivity of the normal region, the average temperature of the normal region at two different time points is represented.

[0015] The technical effects and advantages of this invention are as follows: (1) This invention uses the bearing identification code as the main key and the time sequence and process as the index to store multi-process and multi-modal image data in layers, and establishes a precise mapping relationship. The linear damage of the raceway is directly associated with the microscopic image of the grinding process, the surface peeling is associated with the three-dimensional point cloud data of the assembly process and the microscopic image of the grinding process, and the quenching microcracks are associated with the infrared thermograph sequence of the heat treatment process. Non-target processes can be skipped through the fault code, the core data can be extracted quickly, the source of defects can be accurately traced, and the efficiency of cause analysis can be greatly improved. (2) This invention designs a refined analysis logic for three types of typical defects. The linear damage of the raceway integrates the length, width, depth and position parameters. The surface spalling covers indicators such as two-dimensional projected area, three-dimensional volume and edge features. The quenching microcracks are quantitatively analyzed through the abnormal cooling rate ratio and thermal diffusion ratio. The evaluation report integrates complete information such as bearing identification code, timestamp, fault type, quantitative parameters and causes, breaking through the limitations of traditional qualitative evaluation and providing data support for quality judgment, product traceability and production optimization. Attached Figure Description

[0016] Figure 1 This is a flowchart of the industrial product inspection method based on image data analysis and processing according to the present invention; Figure 2 This is a schematic diagram of the industrial product inspection system based on image data analysis and processing according to the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0018] Example 1

[0019] like Figure 1 As shown, the industrial product inspection method based on image data analysis and processing is as follows: Multi-process image acquisition: Quality monitoring points pre-deployed in the bearing production line are used to collect multi-modal image data of bearings at different process stages; quality monitoring points include post-grinding process points, post-heat treatment process points, pre-assembly process points, and final inspection process points; image data includes at least surface microscopic images, infrared thermographs, three-dimensional morphology images, and multi-angle appearance images; each image data is bound to a unique bearing identification code, process code, timestamp, and acquisition parameters; process stages include grinding process, heat treatment process, assembly process, and final inspection process; A line scan camera system (resolution ≥ 20 million pixels) is installed after the grinding station, and a coaxial light source is configured to acquire microscopic images of the raceway surface; An infrared thermal imager (thermal sensitivity ≤ 0.05℃) was installed at the end of the heat treatment line to acquire a sequence of images showing the temperature field distribution of the rings. A 3D structured light sensor (accuracy ±5μm) is installed at the pre-assembly station to collect three-dimensional point cloud data of the raceway. A multi-angle vision system (4 high dynamic range cameras) is installed at the final inspection station to capture 360° images of the bearing's appearance.

[0020] Image database: Using bearing identification codes as the primary key and time sequence and process as the index, it stores multi-modal image data of bearings at different process stages in a hierarchical manner; before storage, the images are processed by denoising, grayscale conversion, image enhancement and edge correction.

[0021] Fault diagnosis: After the final inspection of the bearing, the management personnel conduct quality inspection on the bearing and selectively output the fault type and fault code of the bearing based on the quality inspection results; Personnel do not need to measure each fault individually; they only need to make an initial judgment on the type of fault. Subsequent quantitative assessments are automatically completed by retrieving relevant image data and using pre-built image algorithms, reducing the workload of quality inspectors.

[0022] The types of failures include raceway linear damage, surface spalling, and quenching microcracks. Pre-set the fault codes corresponding to different fault types; Examples include raceway linear damage (F101), surface spalling (F102), and quenching microcracks (F103).

[0023] Intelligent traceability: In response to the fault code output by the bearing, the bearing identification code and fault code are used in the image database to locate and extract the target image data collected at the target process stage corresponding to the fault code; Image data corresponding to different faults are distributed across different processes (e.g., quenching microcracks can only be effectively identified in infrared thermographic images of the heat treatment process). Fault codes output by personnel quality inspection can be directly used as indexes to skip images of non-target processes, accurately retrieve core modal data, and significantly improve analysis efficiency.

[0024] The target image data extracted corresponding to the linear damage of the raceway is a microscopic image of the raceway surface of the bearing during the grinding process; The image data extracted from the surface peeling are the three-dimensional point cloud data of the raceway of the bearing in the assembly process and the microscopic image of the raceway surface in the grinding process. The extracted image data corresponding to the quenching microcracks is an infrared thermogram sequence of the bearing during the heat treatment process; Correlation analysis: Based on the extracted target image data and fault codes, a collaborative analysis is performed to generate a bearing defect assessment report; Specifically: If the fault type is raceway linear damage; Adaptive threshold segmentation is performed on the microscopic image of the rolling surface; the maximum inter-class variance method is used for segmentation; the defect region is separated from the normal raceway surface to obtain a binarized defect image; a morphological thinning algorithm is performed on the binarized image to extract the single-pixel skeleton of the defect, with the skeleton being the centerline of the defect; the number of pixels in the skeleton is counted, and the pixel length is converted to the actual physical length to obtain the defect length; system calibration parameters based on the line scan camera are obtained. Along the normal direction of the defect skeleton, the defect cross section is cut with a set step size; the step size is set to 5 pixels; the pixel width of the defect at each cross section is counted, and the average pixel width is calculated after averaging. The average pixel width is then converted to the actual physical width to obtain the defect width. Preset a standard grayscale value; that is, the grayscale value of the normal area; calculate the difference between the grayscale value of the defect area and the standard grayscale value as the grayscale difference; establish a mapping rule between the grayscale difference and depth, and convert it into defect depth; That is, by fitting a gray-scale difference-depth curve using a known standard curve sample, the relative temperature of the defect is estimated by substituting the gray-scale difference into the curve.

[0025] Define the center coordinates of the raceway section Calculate the centroid coordinates of the defect region. Calculate the distance from the center of mass to the center of the raceway and convert it to physical distance. ; distance .

[0026] Defect locations are determined by combining physical distance with pre-set location judgment criteria; the curved location includes the center area, shoulder area, and intermediate transition area. Location determination criteria: Central area: ; Shoulder area: ; Intermediate transition area: ; in The physical width is the result of converting the pixel width of the scroll rail region. Shoulder area: The raceway shoulder is the edge transition zone where rolling elements contact. Stress concentration is obvious here, and edge loads and impact damage are likely to occur. Defects here can accelerate fatigue crack initiation and lubrication failure, and the degree of harm is the highest. Central area: The center of the raceway is the main contact area of ​​the rolling elements. It has a large contact area and uniform stress distribution. Minor defects have little impact on overall performance, but serious defects will still affect motion accuracy and lifespan, with a moderate degree of harm. Intermediate transition zone: Located between the center and the shoulder, the probability of rolling element contact and the stress level are in the middle, and the degree of harm caused by defects is the lowest.

[0027] Set a set of additional coefficients corresponding to different defect locations; For example, the additional coefficient for the shoulder area is 1.3, the additional coefficient for the central area is 1.2, and the additional coefficient for the intermediate transition area is 1.0. These can be dynamically adjusted according to the actual situation.

[0028] Based on the additional coefficients derived from the defect location, a weighted fusion is performed using defect length, defect width, and defect depth to output a linear damage assessment coefficient. That is, the defect length, defect width, and defect depth are respectively marked as After normalization, the formula is used. Calculate the linear damage assessment coefficient ;in This is an additional coefficient converted based on the defect location; , as well as These are the preset weighting coefficients.

[0029] The bearing's identification code, timestamp, fault type, defect location of raceway linear damage, and linear damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report. Defect levels can be classified based on linear damage assessment coefficients. Three sets of coefficient intervals are set for each linear damage assessment coefficient, and each set of coefficient intervals corresponds to a defect level. The defect levels include minor scratches, moderate strains, and severe grooves. The higher the linear damage assessment coefficient, the higher the probability of matching a severe groove.

[0030] If the fault type is surface peeling; A combination of grayscale thresholding and morphological segmentation algorithm was used to segment the microscopic image of the raceway surface. The initial segmentation mask was obtained by the OTSU method, and then the morphological closing operation (3×3 rectangle structuring element) was performed to fill the holes. The opening operation was used to remove isolated noise points, resulting in a two-dimensional mask image of the peeled area. The number of defective pixels in the peeling area of ​​the mask image is counted, and combined with the pixel equivalent of the two-dimensional image, it is converted into the actual projected area. The edge contour points of the curve are extracted using a chain code tracking algorithm, and the sum of the Euclidean distances between adjacent contour points is calculated and converted into the actual perimeter. Irregularity is characterized by the reciprocal of circularity. ; This represents the actual projected area. Represents the square of the actual perimeter; The irregularity is compared with the set reference regularity. If the irregularity is greater than the reference regularity, it is determined that the shape of the spalling pit is irregular and the characteristic is brittle fracture. Canny edge detection was performed on the microscopic image of the raceway surface to extract the gradient amplitude image of the defect edge. The average gradient amplitude of the defect edge region was statistically analyzed and the ratio of the average gradient of the defect edge region to that of the normal region was calculated and denoted as the gradient ratio. If the gradient ratio is greater than the preset reference gradient ratio, it is determined to be a clear brittle fracture edge; otherwise, it is determined to be a fuzzy plastic deformation edge. For example If it is, then it is determined to be a clear brittle fracture edge.

[0031] For the three-dimensional point cloud data of the raceway, calculate the normal distance from each point cloud point in the peeling area to the preset ideal surface, define it as the single-point depth, and extract the maximum depth; A 3D mesh model of the spalled area is generated; the Poisson reconstruction algorithm is used to generate the model; the volume is calculated by mesh voxelization to obtain the spalled volume. Obtain; among them The number of voxels to fill the exfoliated area, The volume of a single voxel.

[0032] Two-dimensional cross-sectional contour lines are extracted along the axial and circumferential directions of the center of the spalling pit. The contour lines are composed of the point cloud depth values ​​within the cross-section. The edge steepness of the cross-sectional contour is calculated. It is defined as the depth change rate at the edge of the contour; it is calculated by the ratio of the depth change to the path length change. The depth change is calculated using the difference between the normal surface depth on the edge side and the depth of the spalling pit on the inner edge side; the path length change is calculated using the ratio of the arc length to the straight line segment length corresponding to the depth change on the cross-sectional profile.

[0033] The actual projected area, gradient ratio, maximum depth, peeling volume, and edge steepness are combined and weighted to output the peeling damage assessment coefficient. That is, after normalizing the actual projected area, gradient ratio, maximum depth, peeling volume, and edge steepness, the formula is used. Calculate the peeling damage assessment coefficient ;in These represent the actual projected area, gradient ratio, maximum depth, peeling volume, and edge steepness, respectively. These are the preset weighting coefficients.

[0034] The bearing's identification code, timestamp, fault type, spalling pit shape, edge features, and spalling damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report. The edge features include clear brittle fracture edges and blurred plastic deformation edges.

[0035] The spalling level can be divided according to the spalling damage assessment coefficient. Three sets of coefficient intervals are set for each spalling damage assessment coefficient, and each set of coefficient intervals corresponds to a spalling level. The spalling level includes slight spalling, moderate spalling and severe spalling. The higher the spalling damage assessment coefficient, the higher the probability of matching severe spalling.

[0036] If the fault type is quenching microcrack; Based on the blackbody calibration curve of the infrared thermal imager, the digital output value of the thermal image is converted into the real temperature value to obtain the spatiotemporal temperature field data. Crack regions are extracted using threshold segmentation. For each pixel within a crack region, the instantaneous cooling rate is calculated using the forward difference method based on its temperature-time curve. Instantaneous cooling rate ;in The time interval between two adjacent frames is determined by the thermal imager's frame rate; This represents the temperature value of two adjacent frames.

[0037] Calculate the average cooling rate of a neighborhood within a defined range centered on each pixel; for example, 7×7. Define the cooling rate anomaly coefficient That is, to obtain by using the formula This characterizes the degree of difference between the target pixel and its neighborhood; among which The instantaneous cooling rate of the target pixel. The average cooling rate of the neighborhood centered on the target pixel; The calculated cooling rate anomaly coefficient for each pixel is compared with a set anomaly threshold. If a pixel exceeds the anomaly threshold, it is marked as a cooling rate anomaly point. The ratio of abnormal cooling rate points to the total number of points in the cracked region is used to obtain the rate anomaly ratio. Two sets of time points were selected during the cooling process, representing the average temperature of the cracked region and the normal region, and the formula was applied. Calculate the thermal diffusivity for the cracked region and the normal region, respectively; r is the size of the region feature. The ambient temperature; For the two selected time sets, When calculating the thermal diffusivity of the cracked region, the average temperature of the cracked region at two sets of time points is represented; when calculating the thermal diffusivity of the normal region, the average temperature of the normal region at two sets of time points is represented. The diffusion ratio is obtained by calculating the ratio between the thermal diffusivity of the cracked region (numerator) and the thermal diffusivity of the normal region (denominator). The lower the diffusion ratio, the more significant the decrease in the thermal diffusion capacity of the crack zone (the crack obstructs the heat conduction path), indicating that the cooling difference is caused by quenching microcracks.

[0038] The cooling evaluation coefficient is output by weighted fusion of rate anomaly ratio and diffusion ratio. That is, after normalizing the rate anomaly ratio and diffusion ratio, the formula is used. Calculate the cooling evaluation coefficient Z; where and These represent the rate anomaly ratio and the diffusion ratio, respectively. and These are the preset weighting coefficients.

[0039] If the cooling evaluation coefficient is higher than the preset cooling threshold coefficient, the quenching microcracks are determined to be caused by cooling differences; otherwise, they are determined to be caused by unknown reasons. The bearing's identification code, timestamp, fault type, and the cause of the quenching microcracks, including cooling differences and unknown causes, are integrated into a pre-built report template to generate a defect assessment report.

[0040] Example 2

[0041] Please see Figure 2 As shown, based on the industrial product inspection method based on image data analysis and processing provided in Embodiment 1 of this application, Embodiment 2 of this application proposes an industrial product inspection system based on image data analysis and processing. Embodiment 2 is merely a preferred embodiment of Embodiment 1, and the implementation of Embodiment 2 will not affect the individual implementation of Embodiment 1.

[0042] Specifically, the industrial product inspection system based on image data analysis and processing provided in Embodiment 2 of this application includes: The image acquisition module is used to transmit multiple modal image data of the bearing acquired at different stages of the process; The image storage module is used for hierarchical storage of multiple modal image data of bearings at different process stages; The fault output module is used to input the fault type and fault code of the bearing after the final inspection of the bearing, based on the quality inspection results of the bearing by the management personnel. The fault tracing module is used to locate and extract the target image data collected at the target process stage corresponding to the fault code in the image database by using the bearing's identification code and fault code; The report generation module is used to perform collaborative analysis of the extracted target image data and fault codes to generate a bearing defect assessment report.

[0043] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0044] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0045] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0046] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0047] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0048] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0049] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0050] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0051] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An industrial product inspection method based on image data analysis and processing, characterized in that, include: Step 1: Collect multiple modal image data of the bearing at different process stages, including grinding, heat treatment, assembly and final inspection. Step 2: Using the bearing identification code as the primary key and the time sequence and process as the index, store the multi-modal image data of the bearing at different process stages in a hierarchical manner; Step 3: After the final inspection of the bearing, the management personnel conduct quality inspection on the bearing and selectively output the fault type and fault code of the bearing based on the quality inspection results; Step 4: In response to the fault code output by the bearing, use the bearing's identification code and fault code in the image database to locate and extract the target image data collected at the target process stage corresponding to the fault code. Step 5: Based on the extracted target image data and fault codes, perform collaborative analysis to generate a bearing defect assessment report.

2. The industrial product inspection method based on image data analysis and processing according to claim 1, characterized in that: The logic for setting the fault type and its fault code; The types of failures include raceway linear damage, surface spalling, and quenching microcracks. Pre-set the fault codes corresponding to different fault types.

3. The industrial product inspection method based on image data analysis and processing according to claim 2, characterized in that: The target image data collected during the target process stage corresponding to the fault code; The target image data extracted corresponding to the linear damage of the raceway is a microscopic image of the raceway surface of the bearing during the grinding process; The image data extracted from the surface peeling are the three-dimensional point cloud data of the raceway of the bearing in the assembly process and the microscopic image of the raceway surface in the grinding process. The extracted image data corresponding to the quenching microcracks is an infrared thermographic sequence of the bearing during the heat treatment process.

4. The industrial product inspection method based on image data analysis and processing according to claim 3, characterized in that: Collaborative analysis logic for linear damage to raceways; Adaptive threshold segmentation is performed on the microscopic image of the rolling surface to separate the defect region from the normal raceway surface, resulting in a binarized defect image; the single-pixel skeleton of the defect is extracted from the binarized image, with the skeleton being the center line of the defect; The number of pixels in the skeleton is counted, and the pixel length is converted into the actual physical length to obtain the defect length. The defect section is cut along the normal direction of the defect skeleton with a set step size; The pixel width of the defect at each cross section is counted, and the average pixel width is obtained by averaging the values. The average pixel width is then converted to the actual physical width to obtain the defect width. A standard grayscale value is preset, and the difference between the grayscale value of the defect area and the standard grayscale value is calculated as the grayscale difference; a mapping rule between the grayscale difference and depth is established and converted into defect depth. Based on the additional coefficients derived from the defect location, a weighted fusion is performed using defect length, defect width, and defect depth to output a linear damage assessment coefficient. The bearing's identification code, timestamp, fault type, defect location of raceway linear damage, and linear damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report.

5. The industrial product inspection method based on image data analysis and processing according to claim 4, characterized in that: The conversion logic of the additional coefficient; Define the center coordinates of the raceway section, calculate the centroid coordinates of the defect area, calculate the distance from the centroid to the raceway center, and convert it to physical distance; Defect locations are determined by combining physical distance with pre-set location judgment criteria; curve locations include the central area, shoulder area, and intermediate transition area; and a set of additional coefficients are set for different defect locations.

6. The industrial product inspection method based on image data analysis and processing according to claim 3, characterized in that: Collaborative analysis logic for surface peeling; A two-dimensional mask image of the peeled area was obtained by using a combination of grayscale thresholding and morphological segmentation algorithm on the microscopic image of the raceway surface. The number of defective pixels in the peeling area of ​​the mask image is counted, and combined with the pixel equivalent of the two-dimensional image, it is converted into the actual projected area. Extract the gradient magnitude image of the defect edge, count the average gradient magnitude of the defect edge region, and calculate the ratio of the average gradient of the defect edge region to the average gradient of the normal region, which is denoted as the gradient ratio. If the gradient ratio is greater than the preset reference gradient ratio, it is determined to be a clear brittle fracture edge; otherwise, it is determined to be a fuzzy plastic deformation edge. For the 3D point cloud data of the raceway, calculate the normal distance from each point cloud point in the peeling area to the preset ideal surface, define it as the single-point depth, and extract the maximum depth; generate a 3D mesh model of the peeling area, calculate the volume through mesh voxelization, and obtain the peeling volume; Two-dimensional cross-sectional contour lines are extracted along the axial and circumferential directions of the center of the spalling pit. The contour lines are composed of the point cloud depth values ​​within the cross-section. The edge steepness of the cross-sectional contour is calculated. The actual projected area, gradient ratio, maximum depth, peeling volume, and edge steepness are combined and weighted to output the peeling damage assessment coefficient. The bearing's identification code, timestamp, fault type, spalling pit shape, edge features, and spalling damage assessment coefficient are integrated and filled into a pre-built report template to generate a defect assessment report. The edge features include clear brittle fracture edges and blurred plastic deformation edges.

7. The industrial product inspection method based on image data analysis and processing according to claim 6, characterized in that: The edge features include clear brittle fracture edges and blurred plastic deformation edges; If the gradient ratio is greater than the preset reference gradient ratio, it is determined to be a clear brittle fracture edge; otherwise, it is determined to be a fuzzy plastic deformation edge. Logic for determining the shape of spalling pits; Extract the edge contour points of the curve, calculate the sum of the Euclidean distances of adjacent contour points, and convert it into the actual perimeter; use the reciprocal of the roundness to represent the irregularity, compare the irregularity with the set reference regularity, if the irregularity is greater than the reference regularity, it is determined that the spalling pit shape is irregular, and the feature is brittle fracture.

8. The industrial product inspection method based on image data analysis and processing according to claim 3, characterized in that: Collaborative analysis logic for quenching microcracks; The digital output value of the thermal image is converted into the real temperature value to obtain the spatiotemporal temperature field data; the crack region is extracted by threshold segmentation, and the instantaneous cooling rate is calculated for each pixel in the crack region based on its temperature-time curve. The average cooling rate of the neighborhood is calculated by selecting a defined range of neighborhoods centered on each pixel. Define the cooling rate anomaly coefficient The calculated cooling rate anomaly coefficient for each pixel is compared with a set anomaly threshold. If a pixel exceeds the anomaly threshold, it is marked as a cooling rate anomaly point. The ratio of abnormal cooling rate points to the total number of points in the cracked region is used to obtain the rate anomaly ratio. Two sets of time points during the cooling process are selected, along with the average temperatures of the cracked region and the normal region. The thermal diffusivity of the cracked region and the normal region are calculated respectively. The diffusivity is obtained by calculating the ratio of the thermal diffusivity of the cracked region to the thermal diffusivity of the normal region. The cooling evaluation coefficient is output by weighted fusion of the rate anomaly ratio and the diffusion ratio. If the cooling evaluation coefficient is higher than the preset cooling threshold coefficient, the quenching microcracks are determined to be caused by cooling differences; otherwise, they are determined to be caused by unknown reasons. The bearing's identification code, timestamp, fault type, and the cause of the quenching microcracks, including cooling differences and unknown causes, are integrated into a pre-built report template to generate a defect assessment report.

9. The industrial product inspection method based on image data analysis and processing according to claim 8, characterized in that: Definition logic; That is, to obtain by using the formula ,in The instantaneous cooling rate of the target pixel. The average cooling rate of the neighborhood centered on the target pixel; Logic for calculating thermal diffusivity; According to the formula Calculate the thermal diffusivity for the cracked and normal regions, respectively; r is the characteristic size of the region. The ambient temperature; For the two selected time sets, When calculating the thermal diffusivity of the cracked region, the average temperature of the cracked region at two different time points is represented. When calculating the thermal diffusivity of the normal region, the average temperature of the normal region at two different time points is represented.

10. An industrial product inspection system based on image data analysis and processing, applied to the industrial product inspection method based on image data analysis and processing according to any one of claims 1-9, characterized in that: The image acquisition module is used to transmit multiple modal image data of the bearing acquired at different stages of the process; The image storage module is used for hierarchical storage of multiple modal image data of bearings at different process stages; The fault output module is used to input the fault type and fault code of the bearing after the final inspection of the bearing, based on the quality inspection results of the bearing by the management personnel. The fault tracing module is used to locate and extract the target image data collected at the target process stage corresponding to the fault code in the image database by using the bearing's identification code and fault code; The report generation module is used to perform collaborative analysis of the extracted target image data and fault codes to generate a bearing defect assessment report.

Images