A method and system for detecting appearance defects of rubber integral molds based on machine vision

Abstract

The present invention relates to the technical field of appearance defect monitoring, and particularly to a method and system for detecting appearance defects of a rubber integral mold based on machine vision. The method includes the following steps: obtaining rubber integral mold parameter data; setting the inner and outer diameters of the O-ring of the rubber integral mold based on the rubber integral mold parameter data to obtain the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold; performing an overall photographing of the first surface of the rubber integral mold according to the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold; dividing the detection area of the standard first overall appearance image of the rubber integral mold to generate rubber integral mold detection area data; comparing the rubber integral mold detection area data with a preset detection area quantity range to generate rubber integral mold product specification data. The present invention improves the accuracy and comprehensiveness of defect detection through precise parameter setting, automatic area division, focal length adjustment, high-precision defect detection and comprehensive report generation.

Description

Technical Field

[0001] The present invention relates to the technical field of appearance defect monitoring, and particularly to a method and system for detecting appearance defects of a rubber integral mold based on machine vision. Background Art

[0002] Early rubber defect detection mainly relied on manual inspection and simple mechanical detection. These methods were not only time-consuming and laborious, but also easily affected by human factors, making it difficult to ensure consistency and high precision. With the rapid development of computer vision technology, machine vision systems began to be introduced into the rubber detection field. Early machine vision systems used simple image processing algorithms such as edge detection and template matching. Although these methods could detect some basic defects, their ability to identify complex defect types and minute flaws was limited. After entering the 21st century, the introduction of deep learning technology significantly improved the detection ability of machine vision systems. Through convolutional neural networks (CNNs) and image classification models, modern detection systems can achieve automatic recognition and classification of various defects. These deep learning models can learn from a large amount of labeled data, identify more complex defect patterns, and have high accuracy and robustness. However, currently, traditional defect detection methods cannot accurately detect and locate minute defects, and lack comprehensive detection of multiple angles or surfaces, resulting in the inability to comprehensively identify all potential defects, and thus the accuracy and comprehensiveness of defect detection are relatively low. Summary of the Invention

[0003] Based on this, it is necessary to provide a method and system for detecting appearance defects of a rubber integral mold based on machine vision to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for detecting appearance defects of a rubber integral mold based on machine vision, the method includes the following steps:

[0005] Step S1: Obtain rubber integral mold parameter data; set the inner and outer diameters of the O-ring based on the rubber integral mold parameter data to obtain the adjusted inner and outer diameter data of the O-ring built in the rubber integral mold; perform an overall shooting of the first surface of the rubber integral mold according to the adjusted inner and outer diameter data of the O-ring built in the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold;

[0006] Step S2: Divide the detection area of the standard first overall appearance image of the rubber integral mold to generate rubber integral mold detection area data; compare the rubber integral mold detection area data with a preset detection area quantity range to generate rubber integral mold product specification data; perform trajectory positioning on the standard first overall appearance image of the rubber integral mold based on a trajectory positioning camera according to the rubber integral mold product specification data to generate rubber integral mold product defect shooting trajectory data;

[0007] Step S3: Adjust the lens focal length of the defect camera based on the rubber integral mold product specification data to generate defect camera focal length adjustment data; perform rubber integral mold appearance defect detection on the rubber integral mold product defect shooting trajectory data based on the defect camera focal length adjustment data to obtain the rubber integral mold overall appearance defect detection image; perform color marking on the rubber integral mold overall appearance defect detection image to generate defective product positioning marking data and normal product positioning marking data; perform color mapping on the standard rubber integral mold first overall appearance image through the defective product positioning marking data and the normal product positioning marking data to generate the rubber integral mold first overall appearance defect detection image;

[0008] Step S4: Perform a second overall shooting on the rubber integral mold according to the rubber integral mold first overall appearance defect detection image to generate the rubber integral mold second overall appearance image and perform second side appearance defect detection until the rubber integral mold second overall appearance defect detection image is generated; upload the rubber integral mold first overall appearance defect detection image and the rubber integral mold second overall appearance defect detection image to the central processing system for appearance defect data storage, thereby generating a rubber integral mold appearance defect detection report.

[0009] The present invention accurately sets the inner and outer diameters of the O-ring to ensure that the size of the rubber integral mold meets the specifications, thereby improving the standardization of subsequent images. By generating a standard first overall appearance image through overall shooting, a stable benchmark is provided for subsequent detection, which helps in comparative analysis and accurate defect positioning. The system automatically divides the detection area, reducing manual operation errors and ensuring that each area is properly inspected. By comparing the data of the detection area with the preset specifications, accurate product specification data is generated, which helps in positioning specific areas of the rubber integral mold and improving the accuracy of defect detection. Based on the product specification data, trajectory positioning is performed to ensure the accurate shooting trajectory of the defect camera, reducing the possibility of missed or false detections. The focal length of the defect camera is adjusted according to the rubber integral mold product specification data to ensure the clarity and detail accuracy of the image, improving the accuracy of defect detection. By generating positioning data for defective and normal areas through color marking and performing color mapping on the standard image, a high-precision defect detection image is obtained, enhancing the ability to identify defects. Through color mapping technology, defective and normal areas of the standard image are marked, making the defect location and type more intuitive and facilitating subsequent analysis and processing. The second side of the rubber integral mold is overall shot and defect detected to ensure that all defective areas are detected, avoiding missed detections. The detection images of the first side and the second side are uploaded to the central processing system to centrally store and manage defect data, facilitating subsequent analysis, statistics, and review. By comprehensively analyzing the uploaded detection images, a detailed rubber integral mold appearance defect detection report is generated, providing data support for quality control and improvement and enhancing the traceability and reliability of the detection results. Therefore, the present invention improves the accuracy and comprehensiveness of defect detection through precise parameter setting, automated area division, focal length adjustment, high-precision defect detection, and comprehensive report generation.

[0010] Preferably, step S1 includes the following steps:

[0011] Step S11: Obtain rubber integral mold parameter data;

[0012] Step S12: Based on the rubber integral mold parameter data, set the inner and outer diameters of the O-ring of the rubber integral mold to obtain the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold;

[0013] Step S13: According to the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold, perform an overall shot of the first side of the rubber integral mold to generate the first overall appearance image of the rubber integral mold;

[0014] Step S14: Perform image preprocessing on the first overall appearance image of the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold, where the image preprocessing includes image denoising, image filtering, and brightness enhancement.

[0015] The present invention ensures accurate basic data for subsequent operations by obtaining rubber integral mold parameter data. Obtaining detailed integral mold parameter data can provide necessary references, enabling more precise setting of the inner and outer diameters of the O-ring, thereby improving the quality of the final rubber integral mold. By precisely adjusting the inner and outer diameters of the O-ring, the sealing performance and fit of the integral mold are optimized, avoiding production problems caused by improper parameter settings, such as air leakage or dimensional non-conformance. This can enhance the performance and service life of the rubber integral mold. Generate the first overall appearance image of the surface, providing a visual basis for subsequent image analysis and quality inspection. Through systematic photographing, detailed appearance information can be obtained, which helps to detect existing defects or inconsistencies. Image preprocessing (including denoising, filtering, and brightness enhancement) can improve the clarity and contrast of the image, making subsequent image analysis more accurate. The standardized image is more conducive to quality assessment and defect detection, thereby improving the overall detection efficiency and accuracy.

[0016] Preferably, step S2 includes the following steps:

[0017] Step S21: Divide the detection area of the first overall appearance image of the standard rubber integral mold to generate rubber integral mold detection area data;

[0018] Step S22: Calculate the number of detection areas for the rubber integral mold detection area data to obtain detection area number data;

[0019] Step S23: Compare the detection area number data with the preset detection area number range. When the detection area number data is less than the preset detection area number range, the corresponding rubber integral mold is marked as a small-scale rubber integral mold product; when the detection area number data is within the preset detection area number range, the corresponding rubber integral mold is marked as a medium-scale rubber integral mold product; when the detection area number data is greater than the preset detection area number range, the corresponding rubber integral mold is marked as a large-scale rubber integral mold product;

[0020] Step S24: Integrate the small-scale rubber integral mold products, medium-scale rubber integral mold products, and large-scale rubber integral mold products to generate rubber integral mold product specification data;

[0021] Step S25: Based on the trajectory positioning camera, perform trajectory positioning on the first overall appearance image of the standard rubber integral mold according to the rubber integral mold product specification data to generate rubber integral mold product defect shooting trajectory data.

[0022] Through the division of the detection area for the image, the present invention can systematically perform local detection on the rubber integral mold. This method can help refine the quality inspection process, making the detection more accurate and ensuring that each detection area is fully analyzed. Calculating the number of detection areas provides basic data for further classification and analysis. By quantifying the number of detection areas, an objective basis can be provided for subsequent product scale classification and defect detection. By comparing the actual number of detection areas with the preset range, rubber integral mold products can be classified into different scales (small scale, medium scale, large scale). This classification helps formulate appropriate detection and processing plans according to products of different scales, thereby optimizing production management and quality control. Integrating product data of different scales to generate rubber integral mold product specification data helps standardize product management. This integration can support the standardization of the production process, improve the organization and efficiency of product production, and facilitate the formulation of targeted quality inspection standards. Through the trajectory positioning camera to perform trajectory positioning on the rubber integral mold image, accurate defect shooting trajectory data can be generated. This positioning helps accurately capture the defect position, improve the accuracy of defect detection, and provide specific data support for subsequent quality improvement and defect repair.

[0023] Preferably, step S25 includes the following steps:

[0024] Step S251: Based on the rubber integral mold product specification data, adjust the pixel value of the trajectory positioning camera to generate trajectory positioning camera pixel adjustment data;

[0025] Step S252: Use the trajectory positioning camera pixel adjustment data to perform initial product positioning on the standard rubber integral mold first overall appearance image to generate rubber integral mold product initial positioning data; perform target overlap detection on the rubber integral mold product initial positioning data to generate overlapping detection target positioning data and non-overlapping detection target positioning data;

[0026] Step S253: Perform multi-dimensional feature detection on the overlapping detection target positioning data to generate overlapping target multi-dimensional feature data, where the multi-dimensional feature detection includes area range detection, aspect ratio range detection, and color feature detection; calculate the confidence level for the overlapping target multi-dimensional feature data to obtain the overlapping target confidence level;

[0027] Step S254: Filter the overlapping detection target positioning data according to the overlapping target confidence level to generate overlapping target optimized positioning data; integrate the non-overlapping detection target positioning data and the overlapping target optimized positioning data to generate a rubber integral mold product positioning data set;

[0028] Step S255: Generate rubber integral mold product defect shooting trajectory data through trajectory generation for the rubber integral mold product positioning data set.

[0029] The present invention can optimize the image resolution and detail capture ability by adjusting the camera pixel value, making subsequent positioning and detection more accurate. Precise pixel adjustment helps improve the image quality, thereby enhancing the accuracy of defect detection. The initial product positioning provides an accurate starting point, enabling subsequent detection and analysis to be carried out on the correct basis. Overlap detection can help distinguish and identify different detection areas, ensuring that the detection and analysis of each area are effective and avoiding detection errors caused by target overlap. Multi-dimensional feature detection (including area range, aspect ratio, and color features) can provide a comprehensive understanding of the target, making the identification and classification of the target more accurate. Confidence calculation helps evaluate the reliability of the detection results, ensuring that only highly credible detection results are included in subsequent processing. By filtering overlapping target positioning, inaccurate or redundant target data can be eliminated, improving the quality of the positioning data. Integrating non-overlapping and optimized overlapping target positioning data to form a comprehensive positioning data set helps build a complete product defect data foundation and improve the overall efficiency of data processing. Generating defect shooting trajectory data can guide the camera to shoot along a predetermined trajectory during actual detection. This not only improves the systematicness and comprehensiveness of the detection but also helps accurately capture the location of the defect, providing reliable data support for subsequent quality analysis and improvement.

[0030] Preferably, step S255 includes the following steps:

[0031] Step S2551: Confirm the product detection center point of the rubber integral mold product positioning data set to obtain the center point coordinate data of the rubber integral mold product;

[0032] Step S2552: Connect the detection areas of the center point coordinate data of the rubber integral mold product and the rubber integral mold detection area data to generate the detection area connection data of the rubber integral mold product;

[0033] Step S2253: Generate a single-line flying shooting trajectory for the detection area connection data of the rubber integral mold product to generate the defect shooting trajectory data of the rubber integral mold product.

[0034] The present invention can provide an accurate reference point by confirming the center point coordinate data of the rubber integral mold product, making the subsequent detection and shooting trajectory planning more accurate. The confirmation of the center point helps to ensure the consistent positioning of the entire product during the detection process, effectively avoiding detection problems caused by position deviation. Connecting the product center point coordinate data with the detection area data can form a complete detection area layout. This connection helps to ensure that all detection areas are covered and the relationships between areas are effectively managed, making the defect detection more systematic and comprehensive. Generating a single-line flying shooting trajectory helps to formulate an optimized shooting path, thereby ensuring that each key area is effectively covered during the detection process. By planning the single-line flying shooting trajectory, the efficiency and accuracy of shooting can be improved, reducing repeated shooting or missed areas and enhancing the quality and speed of defect detection.

[0035] Preferably, step S3 includes the following steps:

[0036] Step S31: Adjust the lens focal length of the defect camera based on the rubber integral mold product specification data to generate defect camera focal length adjustment data; perform rubber integral mold appearance defect detection on the rubber integral mold product defect shooting trajectory data based on the defect camera focal length adjustment data to obtain the rubber integral mold overall appearance defect detection image;

[0037] Step S32: Extract the color defect features from the rubber integral mold overall appearance defect detection image to generate rubber integral mold defect feature extraction data; locate the defective products in the rubber integral mold appearance defect detection image according to the rubber integral mold defect feature extraction data to generate defective product location data and non-defective product location data;

[0038] Step S33: Perform the first color marking on the defective product location data to generate defective product location marking data; perform the second color marking on the non-defective product location data to generate normal product location marking data;

[0039] Step S34: Perform color mapping on the standard rubber integral mold first overall appearance image through the defective product location marking data and the normal product location marking data to generate the rubber integral mold first overall appearance defect detection image.

[0040] The present invention ensures that the camera can clearly capture the details of the rubber whole mold through accurate adjustment of the lens focal length, improving the resolution and clarity of the defect detection image. The accurate focal length setting makes the defect detection more precise, thus effectively identifying and locating the appearance defects. The accurate adjustment of the lens focal length ensures that the camera can clearly capture the details of the rubber whole mold, improving the resolution and clarity of the defect detection image. The accurate focal length setting makes the defect detection more precise, thus effectively identifying and locating the appearance defects. By using different color markings for defective products and normal products, the defective areas and normal areas can be visually presented. This visualization effect helps to quickly identify the defective areas, reduce human errors, and facilitate further analysis and processing. By applying the defect marking information to the standard rubber whole mold image through color mapping, the defects become more prominent in the overall appearance image. This processing can effectively integrate the defect information with the product appearance, providing a clear defect detection image to support more effective quality control and problem location.

[0041] Preferably, step S32 includes the following steps:

[0042] Step S321: Conduct rough positioning by searching for a template on the rubber whole mold appearance defect detection image to generate rough positioning data of the rubber whole mold product; extract the shape features of the rubber whole mold appearance defect detection image to obtain the appearance shape feature data of the rubber whole mold product; perform fine positioning of the rubber whole mold product on the rough positioning data of the rubber whole mold product according to the appearance shape feature data of the rubber whole mold product to generate fine positioning data of the rubber product.

[0043] Step S322: Encircle the inner and outer contours of the rubber whole mold on the rubber whole mold appearance defect detection image through the fine positioning data of the rubber product to generate a locally extracted image of the rubber whole mold product; perform color conversion on the red, green, and blue layers of the locally extracted image of the rubber whole mold product to generate a color conversion image of the locally extracted rubber whole mold product.

[0044] Step S323: Analyze the color difference area of the color conversion image of the locally extracted rubber whole mold product through the rubber whole mold defect feature extraction data to generate the color difference area of the locally extracted rubber whole mold product, where the color difference area of the locally extracted rubber whole mold product includes the color difference area of the red layer of the locally extracted rubber whole mold product, the color difference area of the green layer of the locally extracted rubber whole mold product, and the color difference area of the blue layer of the locally extracted rubber whole mold product; screen the repeated color difference areas of the color difference area of the red layer of the locally extracted rubber whole mold product, the color difference area of the green layer of the locally extracted rubber whole mold product, and the color difference area of the blue layer of the locally extracted rubber whole mold product to obtain the image of the obvious appearance defect area of the rubber whole mold product; perform traditional visual defect detection on the image of the obvious appearance defect area of the rubber whole mold product to generate the obvious appearance defect feature data of the rubber whole mold product.

[0045] Step S324: Screen the non-repeated color difference regions of the local rubber whole-mold product for the color difference of the rubber whole-mold product appearance recessive defect region image based on the rubber whole-mold product appearance dominant defect region image to obtain the rubber whole-mold product appearance recessive defect region image; perform deep AI vision defect detection on the rubber whole-mold product appearance recessive defect region image to generate the rubber whole-mold product appearance recessive defect feature data;

[0046] Step S325: Integrate the rubber whole-mold product appearance recessive defect feature data and the rubber whole-mold product appearance dominant defect feature data to generate the rubber whole-mold defect feature extraction data; perform defect product positioning on the rubber whole-mold appearance defect detection image according to the rubber whole-mold defect feature extraction data to generate defect product positioning data and non-defect product positioning data.

[0047] The present invention can initially determine the defect region by performing rough positioning and shape feature extraction on the rubber whole-mold appearance defect detection image, and perform precise positioning through the shape feature. The rough positioning quickly locates the potential problem region, while the precise positioning provides a more detailed defect position to ensure the accuracy and comprehensiveness of the detection. By precisely delineating the inner and outer contours of the product, the detection region can be further divided for subsequent detailed analysis. The color conversion separates the different color layers of the image to help distinguish and analyze the color-related defects, making the color feature extraction clearer and more effective. Analyzing the color difference region can identify the regions with color deviation. By screening the repeated color difference regions, the dominant defects can be determined and further confirmed by traditional vision detection. This helps to clearly identify and classify the dominant defects to ensure the accuracy of quality control. By screening the non-repeated color difference regions, the recessive defects can be identified. Using the deep AI vision detection technology can more efficiently and accurately discover these recessive defects that are not easily detected by traditional methods. Ensure comprehensive detection and analysis of all potential defects. Integrating the dominant and recessive defect feature data can provide a comprehensive defect characteristic data set. This integration can comprehensively reflect the defect situation of the product, making the defect analysis more in-depth and providing comprehensive data support for subsequent product improvement and quality control.

[0048] Preferably, performing deep AI vision defect detection on the rubber whole-mold product appearance recessive defect region image includes:

[0049] Perform regional focusing analysis on the rubber whole-mold product appearance recessive defect region image to generate the rubber whole-mold product focused recessive defect feature data; divide the rubber whole-mold product focused recessive defect feature data into data sets to generate a model training set and a model test set;

[0050] The model training set is trained through a deep learning classification algorithm to generate a pre-model for recessive defect classification; the pre-model for recessive defect classification is tested using the model test set to generate a recessive defect classification model; the image of the recessive defect area on the appearance of the rubber integral mold product is imported into the recessive defect classification model for recessive defect classification to generate recessive defect classification data;

[0051] The image of the recessive defect area on the appearance of the rubber integral mold product is verified for multi-modal defects through the recessive defect classification data to generate recessive defect feature data for the appearance of the rubber integral mold product, where the multi-modal defect verification includes optical verification and thermal imaging verification.

[0052] The present invention can concentrate on detecting the characteristics of the recessive defect area through regional focus analysis, providing higher details and accuracy. The generated focused recessive defect feature data can help accurately locate and identify recessive defects, improving the sensitivity and reliability of detection. Dividing the recessive defect feature data into a model training set and a model test set helps to construct and evaluate the performance of the deep learning model. The training set is used for the learning and training of the model, while the test set is used to verify the accuracy and generalization ability of the model. This method ensures that the model can effectively identify recessive defects and perform well in practical applications. Generating a recessive defect classification model through a deep learning classification algorithm enables the model to identify and classify different types of recessive defects. This model can improve the detection ability of recessive defects, reduce missed detections and false detections, and improve the overall detection accuracy. Using the classification model to classify the image of the recessive defect area can generate specific recessive defect classification data. This data can provide detailed information about the defect type and location, helping with subsequent quality analysis and improvement measures. Through multi-modal verification methods such as optical verification and thermal imaging verification, the accuracy of recessive defects can be verified from different detection angles and technical means. Optical verification provides a detailed perspective on the image, while thermal imaging verification can detect temperature changes to further confirm the existence of defects. This multi-modal verification helps to improve the reliability and comprehensiveness of the detection results, ensuring the accuracy and consistency of defect detection.

[0053] Preferably, step S4 includes the following steps:

[0054] Step S41: The second side of the rubber integral mold is taken as a whole according to the first overall appearance defect detection image of the rubber integral mold to generate a second overall appearance image of the rubber integral mold;

[0055] Step S42: The second overall appearance image of the rubber integral mold is preprocessed to generate a standard second overall appearance image of the rubber integral mold; based on the second overall appearance image of the rubber integral mold, return to step S21 to start the second side appearance defect detection until a second overall appearance defect detection image of the rubber integral mold is generated;

[0056] Step S43: Upload the first overall appearance defect detection image and the second overall appearance defect detection image of the rubber integral mold to the central processing system for storing appearance defect data, thereby generating an appearance defect detection report for the rubber integral mold.

[0057] In the present invention, by taking an overall photograph of the second surface of the rubber integral mold to obtain appearance images from different angles, the comprehensive detection of appearance defects is ensured. This multi-angle photographing method helps to discover hidden defects that cannot be detected on the first surface, improving the detection coverage rate. Preprocessing the second overall appearance image of the rubber integral mold to generate a standard image helps to improve the accuracy and consistency of subsequent detections. Return to step S21 to repeat the detection process of the first surface to ensure that the second surface also undergoes fine defect detection. This method can ensure that each surface of the rubber integral mold undergoes the same strict detection procedure, thereby reducing omissions and enhancing the integrity of the detection. Uploading the appearance defect detection images of the first surface and the second surface to the central processing system for data storage helps to build a comprehensive defect data file. The generated appearance defect detection report for the rubber integral mold provides a detailed record of the product quality, which can be used for quality analysis, problem tracing, and subsequent production improvement. This systematic data management method helps to improve the transparency and traceability of the production process, ensuring the continuous improvement of product quality.

[0058] In this specification, an appearance defect detection system for a rubber integral mold based on machine vision is provided, which is used to execute the above-mentioned appearance defect detection method for a rubber integral mold based on machine vision. The appearance defect detection system for a rubber integral mold based on machine vision includes:

[0059] An overall appearance photographing module, which is used to obtain rubber integral mold parameter data; set the inner and outer diameters of the O-ring of the rubber integral mold based on the rubber integral mold parameter data to obtain the adjusted inner and outer diameter data of the built-in O-ring of the rubber integral mold; take an overall photograph of the first surface of the rubber integral mold according to the adjusted inner and outer diameter data of the built-in O-ring of the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold;

[0060] A product trajectory positioning module, which is used to divide the detection area of the standard first overall appearance image of the rubber integral mold to generate rubber integral mold detection area data; compare the rubber integral mold detection area data with the preset detection area quantity range to generate rubber integral mold product specification data; perform trajectory positioning on the standard first overall appearance image of the rubber integral mold based on the trajectory positioning camera according to the rubber integral mold product specification data to generate rubber integral mold product defect photographing trajectory data;

[0061] The product defect detection module is used to adjust the lens focal length of the defect camera based on the rubber whole-mold product specification data, generating defect camera focal length adjustment data; detecting the rubber whole-mold appearance defects of the rubber whole-mold product defect shooting trajectory data based on the defect camera focal length adjustment data, thereby obtaining the rubber whole-mold overall appearance defect detection image; performing color marking on the rubber whole-mold overall appearance defect detection image, generating defective product positioning marking data and normal product positioning marking data; performing color mapping on the standard rubber whole-mold first overall appearance image through the defective product positioning marking data and the normal product positioning marking data, generating the rubber whole-mold first overall appearance defect detection image;

[0062] The product defect report module is used to perform a second-side overall shooting on the rubber whole-mold according to the rubber whole-mold first overall appearance defect detection image, generating the rubber whole-mold second overall appearance image and performing second-side appearance defect detection until the rubber whole-mold second overall appearance defect detection image is generated; uploading the rubber whole-mold first overall appearance defect detection image and the rubber whole-mold second overall appearance defect detection image to the central processing system for appearance defect data storage, thereby generating a rubber whole-mold appearance defect detection report.

[0063] The beneficial effects of the present invention are as follows. By ensuring that the inner and outer diameters of the O-ring are set to meet the design requirements, a high-standard first overall appearance image is generated. This helps to accurately detect and evaluate the appearance of the rubber whole-mold in subsequent steps. By standardizing the shooting conditions and parameters, a consistent reference image is provided, serving as a reliable reference for subsequent defect detection. By precisely dividing the detection area, it is ensured that each area is properly inspected, reducing the error of manual operation and improving the comprehensiveness and accuracy of detection. Automatically comparing the preset range of the number of areas, accurate product specification data is generated, thereby optimizing the subsequent trajectory positioning and defect detection processes. By adjusting the focal length of the defect camera, the best image clarity is ensured, improving the detection accuracy of defects. Defect detection is performed based on the focal length adjustment data, and the defective and normal areas are accurately located through color marking, improving the accuracy of defect recognition and the readability of detection results. Shooting the second side of the rubber whole-mold ensures a comprehensive inspection of all defective areas, avoiding omission. Uploading the defect detection images of the first side and the second side to the central processing system to generate a comprehensive appearance defect detection report helps to systematically manage and analyze defect data, improving the detection efficiency and the comprehensiveness of the report. Therefore, the present invention improves the accuracy and comprehensiveness of defect detection through precise parameter setting, automated area division, focal length adjustment, high-precision defect detection, and comprehensive report generation. Description of the Drawings

[0064] Figure 1 It is a schematic diagram of the step flow of a rubber whole-mold appearance defect detection method based on machine vision;

[0065] Figure 2 is Figure 1 a detailed schematic flowchart of the implementation steps of step S2 in

[0066] Figure 3 is Figure 1 a detailed schematic flowchart of the implementation steps of step S3 in

[0067] Figure 4 is Figure 1 a detailed schematic flowchart of the implementation steps of step S4 in

[0068] The realization, functional characteristics and advantages of the object of the present invention will be further described in conjunction with the embodiments with reference to the accompanying drawings. Specific Embodiments

[0069] The technical method of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0070] In addition, the accompanying drawings are only schematic diagrams of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings represent the same or similar parts, so repeated descriptions thereof will be omitted. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor methods and / or microcontroller methods.

[0071] It should be understood that although the terms "first", "second", etc. may be used here to describe each unit, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, the first unit can be called the second unit, and similarly the second unit can be called the first unit. The term "and / or" used here includes any and all combinations of one or more of the listed related items.

[0072] To achieve the above object, please refer to Figures 1 to 4 , a method for detecting appearance defects of rubber whole molds based on machine vision, the method comprising the following steps:

[0073] Step S1: Obtain the parameter data of the rubber integral mold; set the inner and outer diameters of the O-ring of the rubber integral mold based on the parameter data of the rubber integral mold to obtain the adjustment data of the inner and outer diameters of the O-ring built in the rubber integral mold; perform an overall shooting of the first surface of the rubber integral mold according to the adjustment data of the inner and outer diameters of the O-ring built in the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold;

[0074] Step S2: Divide the detection area of the standard first overall appearance image of the rubber integral mold to generate the detection area data of the rubber integral mold; compare the detection area data of the rubber integral mold with the preset range of the number of detection areas to generate the product specification data of the rubber integral mold; perform trajectory positioning on the standard first overall appearance image of the rubber integral mold based on the trajectory positioning camera according to the product specification data of the rubber integral mold to generate the product defect shooting trajectory data of the rubber integral mold;

[0075] Step S3: Adjust the lens focal length of the defect camera based on the product specification data of the rubber integral mold to generate the focal length adjustment data of the defect camera; perform the appearance defect detection of the rubber integral mold on the product defect shooting trajectory data of the rubber integral mold based on the focal length adjustment data of the defect camera to obtain the overall appearance defect detection image of the rubber integral mold; perform color marking on the overall appearance defect detection image of the rubber integral mold to generate the defect product positioning marking data and the normal product positioning marking data; perform color mapping on the standard first overall appearance image of the rubber integral mold through the defect product positioning marking data and the normal product positioning marking data to generate the first overall appearance defect detection image of the rubber integral mold;

[0076] Step S4: Perform an overall shooting of the second surface of the rubber integral mold according to the first overall appearance defect detection image of the rubber integral mold to generate the second overall appearance image of the rubber integral mold and perform the appearance defect detection of the second surface until the second overall appearance defect detection image of the rubber integral mold is generated; upload the first overall appearance defect detection image of the rubber integral mold and the second overall appearance defect detection image of the rubber integral mold to the central processing system for storage of the appearance defect data, thereby generating the appearance defect detection report of the rubber integral mold.

[0077] The present invention accurately sets the inner and outer diameters of the O-ring to ensure that the size of the rubber integral mold meets the specifications, thereby improving the standardization of subsequent images. By generating a standard first overall appearance image through overall shooting, a stable benchmark is provided for subsequent inspections, which helps in comparative analysis and accurate defect localization. The system automatically divides the inspection area, reducing manual operation errors and ensuring that each area is properly inspected. By comparing the data of the inspection area with the preset specifications, accurate product specification data is generated, which helps in locating specific areas of the rubber integral mold and improving the accuracy of defect detection. Based on the product specification data, trajectory positioning is performed to ensure the accurate shooting trajectory of the defect camera, reducing the possibility of missed or misdetected defects. The focal length of the defect camera is adjusted according to the product specification data of the rubber integral mold to ensure the clarity and detail accuracy of the image, improving the accuracy of defect detection. By generating color markers to obtain the positioning data of the defect and normal areas and performing color mapping on the standard image, a high-precision defect detection image is obtained, enhancing the defect recognition ability. Through color mapping technology, defect markers and normal area markers are made on the standard image, making the defect positions and types more intuitive and facilitating subsequent analysis and processing. The second side of the rubber integral mold is overall shot and defect inspected to ensure that all defect areas are detected, avoiding missed inspections. The inspection images of the first side and the second side are uploaded to the central processing system to centrally store and manage the defect data, facilitating subsequent analysis, statistics, and review. By comprehensively analyzing the uploaded inspection images, a detailed inspection report on the appearance defects of the rubber integral mold is generated, providing data support for quality control and improvement and enhancing the traceability and reliability of the inspection results. Therefore, the present invention improves the accuracy and comprehensiveness of defect detection through precise parameter setting, automated area division, focal length adjustment, high-precision defect detection, and comprehensive report generation.

[0078] In the embodiment of the present invention, referring to Figure 1 as described, it is a schematic diagram of the step flow of a method for detecting the appearance defects of a rubber integral mold based on machine vision according to the present invention. In this example, the method for detecting the appearance defects of a rubber integral mold based on machine vision includes the following steps:

[0079] Step S1: Obtain the parameter data of the rubber integral mold; based on the parameter data of the rubber integral mold, set the inner and outer diameters of the O-ring of the rubber integral mold to obtain the adjusted data of the inner and outer diameters of the O-ring built in the rubber integral mold; according to the adjusted data of the inner and outer diameters of the O-ring built in the rubber integral mold, perform an overall shot of the first side of the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold;

[0080] In the embodiments of the present invention, by obtaining the design drawings or CAD files of the rubber integral mold. Obtain the technical specifications and dimensional data of the integral mold from the manufacturer. Use measuring tools to conduct on-site measurement of the actual integral mold to obtain accurate dimensional data. Record the inner diameter, outer diameter and related dimensions of the O-ring. Record the physical properties of the rubber material, such as hardness, elasticity, etc. Record the overall dimensions and geometric shape of the mold. Input the inner diameter and outer diameter parameters of the O-ring into the design software and set the correct dimensions. Use the software to simulate the installation and adjustment process of the O-ring to ensure that the parameter settings meet the design requirements. Calculate the adjustment amount of the O-ring according to the design requirements and actual dimensions, record and save the adjustment data of the O-ring for subsequent shooting and detection. Select a high-resolution camera and set appropriate lighting conditions to avoid shadows and reflections. Fix the rubber integral mold on the shooting table to ensure that the first side is completely visible. Select multiple angles or use a panoramic camera to shoot the first side of the rubber integral mold to ensure that comprehensive images are captured. Conduct preliminary processing on the captured images (such as denoising, contrast adjustment) to ensure the image quality. Preprocess the captured first-side images to ensure that the images meet the standard requirements (such as denoising, enhancing contrast), and adjust the images to a unified size and format for subsequent analysis.

[0081] Step S2: Divide the detection areas of the first overall appearance image of the standard rubber integral mold to generate rubber integral mold detection area data; compare the rubber integral mold detection area data with the preset detection area quantity range to generate rubber integral mold product specification data; based on the trajectory positioning camera, perform trajectory positioning on the first overall appearance image of the standard rubber integral mold according to the rubber integral mold product specification data to generate rubber integral mold product defect shooting trajectory data;

[0082] In the embodiments of the present invention, by ensuring that the first overall appearance image of the standard rubber integral mold has been obtained and preprocessed, use image segmentation algorithms (such as k-means clustering, superpixel segmentation, edge detection, etc.) to divide the image into areas. According to the division results, determine the boundaries and features of each detection area. Record the coordinates, sizes and area features of each detection area, and save the detection area data in a format suitable for analysis and further processing (such as JSON, CSV). Set the specification range of the detection areas (such as the minimum and maximum area or size of the areas). Measure the features such as the area and shape of each detection area, compare the actual area features with the preset specifications, confirm whether they meet the specification requirements, generate a specification-compliant data report, including the number of detection areas, area sizes, etc., and save the specification data in a standard format (such as JSON, CSV). Set the trajectory positioning camera, including the camera position, shooting angle and trajectory control parameters. Plan the shooting trajectory of the camera according to the product specification data to ensure that all detection areas are covered. Use the camera to shoot according to the preset trajectory, record the images at each position, and record the trajectory data of the camera, including the shooting positions, angles and image information.

[0083] Step S3: Based on the rubber whole-mold product specification data, adjust the lens focal length of the defect camera to generate defect camera focal length adjustment data; based on the defect camera focal length adjustment data, perform rubber whole-mold appearance defect detection on the rubber whole-mold product defect shooting trajectory data, so as to obtain the rubber whole-mold overall appearance defect detection image; perform color marking on the rubber whole-mold overall appearance defect detection image to generate defective product positioning marking data and normal product positioning marking data; perform color mapping on the standard rubber whole-mold first overall appearance image through the defective product positioning marking data and the normal product positioning marking data to generate the rubber whole-mold first overall appearance defect detection image;

[0084] In the embodiment of the present invention, the rubber whole-mold product specification data is obtained from step S2, including the main dimensions and detection area information. According to the product specification data and shooting requirements, adjust the focal length of the camera lens to ensure that the detection area is clearly visible. Use the automatic focusing function of the camera to perform automatic focusing adjustment within the detection area, and record the focal length adjustment data, including the actual focal length value and setting parameters. According to the focal length adjustment data, set the focal length and shooting parameters of the camera, and perform shooting according to the trajectory data in step S2 to ensure that all detection areas are covered. Shoot the overall appearance image of the rubber whole-mold to ensure that the image quality meets the requirements of defect detection, and save the shot defect detection image for subsequent analysis. Use image processing techniques (such as segmentation algorithms, edge detection, color thresholding, etc.) to mark the defect areas, assign specific colors to the defect areas to distinguish the normal areas from the defect areas, and generate defective product positioning marking data and normal product positioning marking data, including the coordinates and color information of the marked areas. Apply the marking data to the standard image to generate an image marked with defective and normal areas, fuse the defective and normal marking information into the standard rubber whole-mold image, and apply the generated marking data to the standard image to obtain an image containing defect markings.

[0085] Step S4: According to the rubber whole-mold first overall appearance defect detection image, perform a second-side overall shooting on the rubber whole-mold to generate the rubber whole-mold second overall appearance image and perform second-side appearance defect detection until the rubber whole-mold second overall appearance defect detection image is generated; upload the rubber whole-mold first overall appearance defect detection image and the rubber whole-mold second overall appearance defect detection image to the central processing system for storage of appearance defect data, so as to generate a rubber whole-mold appearance defect detection report.

[0086] In the embodiments of the present invention, according to the experience of photographing the first side, the angle, focal length, and light source settings of the camera are adjusted to ensure the clarity and consistency of the second-side photographed image. The second side of the rubber integral mold is comprehensively photographed using the camera to ensure that the entire surface of the rubber integral mold is covered. A second overall appearance image of the rubber integral mold is generated and saved in a standard format (such as JPEG, PNG). The previously trained defect detection model or algorithm is used to analyze the second-side image, identify and mark the defect areas in the image, generate an image containing defect marks, and display the location information of the defect areas. The defect detection images of the first side and the second side are integrated into a dataset, and the image data is converted into a format suitable for uploading (such as a compressed package, a JSON description file, etc.). Through a network connection to the central processing system, the image data and related defect detection information are uploaded. The uploaded image data and detection information are extracted from the central processing system, and a standardized detection report format (such as PDF, HTML) is generated. The report includes image data, defect marks, detection analysis results, and suggestions. The report is saved as an electronic document and distributed to relevant personnel.

[0087] Preferably, step S1 includes the following steps:

[0088] Step S11: Obtain the parameter data of the rubber integral mold;

[0089] Step S12: Based on the parameter data of the rubber integral mold, set the inner and outer diameters of the O-ring of the rubber integral mold to obtain the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold;

[0090] Step S13: According to the inner and outer diameter adjustment data of the O-ring built in the rubber integral mold, conduct an overall photograph of the first side of the rubber integral mold to generate a first overall appearance image of the rubber integral mold;

[0091] Step S14: Perform image preprocessing on the first overall appearance image of the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold, where the image preprocessing includes image denoising, image filtering, and brightness enhancement.

[0092] In the embodiments of the present invention, by using a digital measurement device or a three-dimensional scanning device to accurately measure the rubber integral mold, key dimension data thereof is obtained, including the inner and outer diameters of the O-ring, the overall shape of the mold, the surface smoothness, and the material properties, etc. By sorting and archiving these parameter data, a rubber integral mold parameter data set is generated. According to the rubber integral mold parameter data obtained in step S11, a dedicated design software (such as CAD) is used to optimize the inner and outer diameters of the O-ring built in the rubber integral mold. This optimization setting includes adjusting the sizes of the inner and outer diameters of the O-ring to ensure that they meet the actual application requirements. Through simulation and iterative optimization, finally, the adjustment data of the inner and outer diameters of the O-ring built in the rubber integral mold is generated, which is used to guide the subsequent shooting and processing operations. After completing the optimization adjustment of the inner and outer diameters of the O-ring, the rubber integral mold is fixed on a dedicated shooting platform to ensure that its first surface is completely exposed in the field of view of the imaging device. A high-resolution industrial camera is used to take an overall picture of the first surface of the rubber integral mold to capture its appearance details. During the shooting process, it is necessary to ensure uniform illumination and avoid shadows and reflections. After the shooting is completed, a first overall appearance image of the rubber integral mold is generated, which serves as the basis for subsequent image processing. The first overall appearance image of the rubber integral mold obtained in step S13 is input into an image processing system for a series of image preprocessing operations. First, a denoising algorithm is used to remove the noise in the image to improve the clarity of the image. Then, image filtering technology (such as Gaussian filtering) is applied to smooth the image to eliminate texture non-uniformity. Finally, brightness enhancement is performed to make the brightness and contrast of the image reach the best state and ensure clear details. After these preprocessing steps, a standard first overall appearance image of the rubber integral mold is generated, providing a high-quality data basis for subsequent analysis and processing.

[0093] As an example of the present invention, with reference to Figure 2 as shown, in this example, step S2 includes:

[0094] Step S21: Divide the detection area of the standard first overall appearance image of the rubber integral mold to generate rubber integral mold detection area data;

[0095] Step S22: Calculate the number of detection areas for the rubber integral mold detection area data to obtain detection area number data;

[0096] Step S23: Compare the detection area number data with a preset detection area number range. When the detection area number data is less than the preset detection area number range, the corresponding rubber integral mold is marked as a small-scale rubber integral mold product; when the detection area number data is within the preset detection area number range, the corresponding rubber integral mold is marked as a medium-scale rubber integral mold product; when the detection area number data is greater than the preset detection area number range, the corresponding rubber integral mold is marked as a large-scale rubber integral mold product;

[0097] Step S24: Integrate the small-scale rubber integral mold products, medium-scale rubber integral mold products, and large-scale rubber integral mold products to generate rubber integral mold product specification data;

[0098] Step S25: Based on the trajectory positioning camera, perform trajectory positioning on the standard rubber integral mold first overall appearance image according to the rubber integral mold product specification data to generate rubber integral mold product defect shooting trajectory data.

[0099] In the embodiment of the present invention, the standard rubber integral mold first overall appearance image generated in step S14 is imported into the image processing system. The image is analyzed using image segmentation algorithms (such as algorithms based on edge detection or threshold segmentation) to identify and divide the key detection areas on the surface of the rubber integral mold. These detection areas include the inner and outer surfaces of the O-ring, the edges of the mold, the interfaces of key components, etc. After the division is completed, rubber integral mold detection area data is generated, identifying the specific position, size, and shape information of each detection area. After the detection area division is completed, the system will count the generated rubber integral mold detection area data to calculate the total number of detection areas. This calculation can be completed through a simple counting operation to obtain the detection area quantity data. This data represents the total number of areas on the surface of the rubber integral mold that need to be focused on for detection, providing a basis for subsequent product scale classification. According to the actual production requirements, preset the range of the number of detection areas corresponding to rubber integral mold products of different scales. Then, compare the detection area quantity data obtained in step S22 with these preset ranges: If the detection area quantity data is less than the lower limit of the preset range, mark this rubber integral mold product as a "small-scale rubber integral mold product"; if the detection area quantity data is within the preset range, mark it as a "medium-scale rubber integral mold product"; if the detection area quantity data is greater than the upper limit of the preset range, mark it as a "large-scale rubber integral mold product". This classification process helps to optimize the production and detection processes according to the complexity and quantity of the detection areas. After the classification is completed, integrate all the marked rubber integral mold products according to their scales to form a complete product specification data set. This data set includes the specific specification information of each scale product, such as size, number of detection areas, applicable scenarios, etc. The rubber integral mold product specification data will be used to guide subsequent production, quality control, and inventory management work. Using the trajectory positioning camera system, according to the rubber integral mold product specification data generated in step S24, perform detailed trajectory positioning on the standard rubber integral mold first overall appearance image. The trajectory positioning camera will plan the optimal shooting path according to the position, size, and shape of the detection areas to ensure that all key areas can be accurately captured. During the shooting process, the system will record the shooting trajectory and the corresponding detection areas to generate rubber integral mold product defect shooting trajectory data. This data will be used for subsequent automated detection to help identify potential defects on the surface of the rubber integral mold.

[0100] Preferably, step S25 includes the following steps:

[0101] Step S251: Adjust the pixel value of the trajectory positioning camera based on the rubber integral mold product specification data to generate trajectory positioning camera pixel adjustment data;

[0102] Step S252: Perform initial product positioning on the standard rubber integral mold first overall appearance image through the trajectory positioning camera pixel adjustment data to generate rubber integral mold product initial positioning data; perform target overlap detection on the rubber integral mold product initial positioning data to generate overlapping detection target positioning data and non-overlapping detection target positioning data;

[0103] Step S253: Perform multi-dimensional feature detection on the overlapping detection target positioning data to generate overlapping target multi-dimensional feature data, where the multi-dimensional feature detection includes area range detection, aspect ratio range detection, and color feature detection; calculate the confidence level for the overlapping target multi-dimensional feature data to obtain the overlapping target confidence level;

[0104] Step S254: Filter the overlapping detection target positioning data according to the overlapping target confidence level to generate overlapping target optimized positioning data; integrate the non-overlapping detection target positioning data and the overlapping target optimized positioning data to generate a rubber integral mold product positioning data set;

[0105] Step S255: Generate a trajectory for the rubber integral mold product positioning data set to obtain rubber integral mold product defect shooting trajectory data.

[0106] In the embodiments of the present invention, the rubber integral mold product specification data generated in step S24 is imported into the trajectory positioning camera system. According to rubber integral mold products of different specifications, the pixel values of the camera are adjusted accordingly to ensure that the best resolution and image quality can be obtained during shooting. These adjustments include resolution setting, focal length adjustment, and optimization of exposure parameters. After these adjustments, trajectory positioning camera pixel adjustment data is generated, providing technical support for subsequent image capture and positioning. After the pixel value adjustment is completed, the trajectory positioning camera is used for the initial positioning of the rubber integral mold product. The system generates the initial positioning data of the rubber integral mold product by analyzing the first overall appearance image of the standard rubber integral mold and combining the pixel adjustment data. Subsequently, the system performs an overlap detection on the initial positioning data to identify the existing overlapping regions or targets. The overlap detection will distinguish the overlapping part and the non-overlapping part in the image and generate two sets of data: the overlapping detection target positioning data and the non-overlapping detection target positioning data. After identifying the overlapping regions, the system will perform multi-dimensional feature detection on these overlapping regions. Specifically, the multi-dimensional feature detection includes: area range detection: measuring the area of the overlapping region and determining whether it is within a preset reasonable range; aspect ratio range detection: analyzing the aspect ratio of the overlapping region and determining whether it conforms to the standard ratio of the rubber integral mold product; color feature detection: analyzing the color of the overlapping region to ensure that its color features are consistent with the product standard. Through these detections, the system generates the multi-dimensional feature data of the overlapping targets. Next, the system will calculate the confidence level of these feature data to obtain the confidence level of each overlapping target, which is used to evaluate the reliability of its detection. According to the confidence level of the overlapping targets calculated in step S253, the overlapping detection target positioning data is filtered. Only those targets with a higher confidence level will be retained to generate the optimized positioning data of the overlapping targets. The overlapping targets with a lower confidence level will be marked as abnormal for subsequent further inspection. Then, the optimized positioning data of the overlapping targets and the non-overlapping detection target positioning data are integrated to generate a complete rubber integral mold product positioning data set, covering all regions that need to be key detected. Using the rubber integral mold product positioning data set, the shooting trajectory of the camera is planned. The system will generate the optimal shooting path according to the characteristics such as the area distribution, size, and shape in the positioning data set. This trajectory will ensure that the camera can cover all key regions in the most effective way and capture high-quality images. Finally, the rubber integral mold product defect shooting trajectory data is generated, providing a basis for subsequent defect detection and quality control.

[0107] Preferably, step S255 includes the following steps:

[0108] Step S2551: Confirm the product detection center point of the rubber integral mold product positioning data set to obtain the center point coordinate data of the rubber integral mold product;

[0109] Step S2552: Detect area connection is performed on the center point coordinate data of the rubber integral mold product and the rubber integral mold detection area data to generate rubber integral mold product detection area connection data;

[0110] Step S2253: Generate a single-line flying shooting trajectory for the rubber integral mold product detection area connection data to generate rubber integral mold product defect shooting trajectory data.

[0111] In the embodiment of the present invention, based on the rubber integral mold product positioning data set generated according to step S254, the detection center point of the rubber integral mold product is determined. The system will analyze each detection area in the positioning data set and calculate the geometric center or mass center position of the entire rubber integral mold product. This position will be used as a reference point for subsequent detection and shooting to ensure that the camera can be accurately positioned based on the center point. After calculation, the system generates the center point coordinate data of the rubber integral mold product. After the center point is confirmed, the system connects the center point coordinate data of the rubber integral mold product with the rubber integral mold detection area data. By analyzing the relative positions between the center point and each detection area, the system generates an optimal connection path. This path sequentially connects the center point with each detection area to ensure that the camera can smoothly move from one area to the next without missing any key areas. Finally, the system generates rubber integral mold product detection area connection data, laying a foundation for subsequent trajectory generation. After the connection path is determined, the system generates a single-line flying shooting trajectory based on the rubber integral mold product detection area connection data. The single-line flying shooting trajectory means that the camera shoots along a continuous path without stopping or backing up in the middle. This trajectory design helps to improve the shooting efficiency and reduce the motion blur or error generated during the shooting process. Through this process, the system finally generates the defect shooting trajectory data of the rubber integral mold product to ensure that the camera can cover all important detection areas according to the preset trajectory for efficient defect detection.

[0112] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0113] Step S31: Adjust the lens focal length of the defect camera based on the rubber integral mold product specification data to generate defect camera focal length adjustment data; perform rubber integral mold appearance defect detection on the rubber integral mold product defect shooting trajectory data based on the defect camera focal length adjustment data to obtain a rubber integral mold overall appearance defect detection image;

[0114] Step S32: Extract color defect features from the rubber integral mold overall appearance defect detection image to generate rubber integral mold defect feature extraction data; perform defect product positioning on the rubber integral mold appearance defect detection image according to the rubber integral mold defect feature extraction data to generate defect product positioning data and non-defect product positioning data;

[0115] Step S33: Perform a first color marking on the defective product positioning data to generate defective product positioning marking data; perform a second color marking on the non-defective product positioning data to generate normal product positioning marking data.

[0116] Step S34: Perform color mapping on the first overall appearance image of the standard rubber whole mold through the defective product positioning marking data and the normal product positioning marking data to generate a first overall appearance defect detection image of the rubber whole mold.

[0117] In the embodiment of the present invention, by adjusting the lens focal length of the defective camera according to the rubber whole mold product specification data, it is ensured that a clear appearance image of the rubber whole mold can be captured during shooting. The system uses these adjustment data to generate defective camera focal length adjustment data and applies it to the actual shooting process. Subsequently, the system uses these focal length adjustment data to gradually perform appearance defect detection on the rubber whole mold along the preset rubber whole mold product defect shooting trajectory data. Through a series of image capturing and processing, a first overall appearance defect detection image of the rubber whole mold is finally generated. After obtaining the first overall appearance defect detection image of the rubber whole mold, the system will extract the color defect features in the image. These color defect features include color anomalies, uneven brightness, or color differences, etc. By extracting these features, the system generates rubber whole mold defect feature extraction data. Then, based on these feature data, the system further analyzes and locates the first overall appearance defect detection image of the rubber whole mold. The system will locate the defective area and its specific position to generate defective product positioning data; at the same time, for the parts where no defects are detected, non-defective product positioning data is generated. According to the positioning data generated in step S32, the system performs a first color marking on the detected defective product positioning data, usually using a more eye-catching color, such as red, to mark the defective area. At the same time, a second color marking is performed on the non-defective product positioning data, usually using green or other contrasting colors to distinguish the normal area from the defective area. Finally, the system generates defective product positioning marking data and normal product positioning marking data, and these marking data will be used for further image processing and analysis. Finally, the system applies the defective product positioning marking data and the normal product positioning marking data to the first overall appearance image of the standard rubber whole mold for color mapping. The color mapping process will fuse the differently marked data into one image, so that the finally generated first overall appearance defect detection image of the rubber whole mold can intuitively display the defective and normal areas. In this way, users can quickly identify the appearance defects of the rubber whole mold products through this defect detection image and take corresponding treatment measures.

[0118] Preferably, step S32 includes the following steps:

[0119] Step S321: Conduct rough positioning by searching for templates on the rubber whole-mold appearance defect detection image to generate rough positioning data of the rubber whole-mold product; extract shape features from the rubber whole-mold appearance defect detection image to obtain appearance shape feature data of the rubber whole-mold product; perform fine positioning of the rubber whole-mold product on the rough positioning data of the rubber whole-mold product according to the appearance shape feature data of the rubber whole-mold product to generate fine positioning data of the rubber product.

[0120] Step S322: Encircle the inner and outer contours of the rubber whole-mold product on the rubber whole-mold appearance defect detection image through the fine positioning data of the rubber product to generate an extracted image of the local rubber whole-mold product; perform color conversion on the red, green, and blue layers of the extracted image of the local rubber whole-mold product to generate a color-converted image of the local rubber whole-mold product.

[0121] Step S323: Conduct color difference area analysis on the color-converted image of the local rubber whole-mold product through the rubber whole-mold defect feature extraction data to generate a color difference area of the local rubber whole-mold product, where the color difference area of the local rubber whole-mold product includes the color difference area of the red layer of the local rubber whole-mold product, the color difference area of the green layer of the local rubber whole-mold product, and the color difference area of the blue layer of the local rubber whole-mold product; screen for repeated color difference areas in the color difference area of the red layer of the local rubber whole-mold product, the color difference area of the green layer of the local rubber whole-mold product, and the color difference area of the blue layer of the local rubber whole-mold product to obtain an image of the obvious appearance defect area of the rubber whole-mold product; conduct traditional visual defect detection on the image of the obvious appearance defect area of the rubber whole-mold product to generate obvious appearance defect feature data of the rubber whole-mold product.

[0122] Step S324: Screen for non-repeated color difference areas in the color difference area of the local rubber whole-mold product according to the image of the obvious appearance defect area of the rubber whole-mold product to obtain an image of the hidden appearance defect area of the rubber whole-mold product; conduct deep AI visual defect detection on the image of the hidden appearance defect area of the rubber whole-mold product to generate hidden appearance defect feature data of the rubber whole-mold product.

[0123] Step S325: Integrate the hidden appearance defect feature data and the obvious appearance defect feature data of the rubber whole-mold product to generate rubber whole-mold defect feature extraction data; conduct defect product positioning on the rubber whole-mold appearance defect detection image according to the rubber whole-mold defect feature extraction data to generate defect product positioning data and non-defect product positioning data.

[0124] In the embodiments of the present invention, the rough positioning of the rubber integral mold appearance defect detection image is carried out by using the search template technology to quickly determine the approximate position of the rubber integral mold product and generate the rough positioning data of the rubber integral mold product. Then, the system extracts the appearance shape feature data of the rubber integral mold from the image and analyzes its key shape features such as edges and contours. Based on the extracted shape features, the system further refines the rough positioning data and finally generates the precise fine positioning data of the rubber product. According to the fine positioning data, the system precisely delineates the inner and outer contours of the rubber integral mold product to obtain the rubber integral mold local product extraction image. Then, the color conversion of the red, green, and blue (RGB) layers is performed on the local extraction image to enhance the color information representation of the image and generate the rubber integral mold local product color conversion image, providing a basis for subsequent color difference analysis. Based on the color conversion image, the system performs color difference region analysis on each RGB layer of the rubber integral mold to identify the color difference regions in each layer, including the color difference regions of the red, green, and blue layers. Then, the system screens these color difference regions to exclude duplicate color difference regions to identify the appearance dominant defect regions of the rubber integral mold. Finally, the system performs traditional visual defect detection on these dominant defect region images to generate the appearance dominant defect feature data of the rubber integral mold product. After screening out the dominant defects, the system further screens those non-duplicate color difference regions to identify potential recessive defect regions. For these recessive defect regions, the system conducts a more in-depth analysis through the AI visual detection technology of deep learning and finally generates the appearance recessive defect feature data of the rubber integral mold product. The recessive defect feature data and the dominant defect feature data are integrated to obtain the comprehensive rubber integral mold defect feature extraction data. Based on these comprehensive extraction data, the system performs defect positioning on the rubber integral mold appearance defect detection image to generate defect product positioning data and non-defect product positioning data respectively, providing a basis for subsequent processing and analysis.

[0125] Preferably, the deep AI visual defect detection of the rubber integral mold product appearance recessive defect region image includes:

[0126] Performing regional focus analysis on the rubber integral mold product appearance recessive defect region image to generate the rubber integral mold product focus recessive defect feature data; dividing the rubber integral mold product focus recessive defect feature data into a data set to generate a model training set and a model test set;

[0127] Training the model training set through a deep learning classification algorithm to generate a recessive defect classification pre-model; using the model test set to test the recessive defect classification pre-model to generate a recessive defect classification model; importing the rubber integral mold product appearance recessive defect region image into the recessive defect classification model for recessive defect classification to generate recessive defect classification data;

[0128] Perform multi-modal defect verification on the image of the hidden defect area of the rubber integral mold product through hidden defect classification data to generate the hidden defect feature data of the rubber integral mold product appearance, where the multi-modal defect verification includes optical verification and thermal imaging verification.

[0129] In the embodiment of the present invention, the appearance hidden defect area of the rubber integral mold product is photographed by using a high-resolution camera to ensure the image quality. Image enhancement techniques, such as histogram equalization and contrast stretching, are applied to enhance the saliency of the defects. Edge detection algorithms (such as Canny edge detection) or methods based on image segmentation (such as U-Net) are used to extract potential defect areas. The extracted defect areas are refined, and morphological operations (such as dilation and erosion) are applied to remove noise and enhance the edges of the defect areas. Focused hidden defect feature data is extracted, and a deep learning feature extraction network (such as a convolutional neural network CNN) is used to extract high-dimensional features from the defect areas. The focused hidden defect feature data is divided into a training set and a test set. Usually, the training set accounts for 70%-80%, and the test set accounts for 20%-30%. Each image in the training set is labeled with the defect type or location to generate label data. A suitable deep learning classification algorithm, such as a convolutional neural network like ResNet, VGG, or EfficientNet, is selected. The model is trained using the training set, the hyperparameters are tuned, and the model performance is optimized. The model performance is evaluated on the validation set, and hyperparameter tuning is performed to prevent overfitting. The image of the hidden defect area of the rubber integral mold product appearance is input into the trained hidden defect classification model for defect classification to generate hidden defect classification data, including the defect type and its confidence score. The image of the defect area is further collected using an optical microscope or a high-resolution camera to inspect the defects in detail. By comparing the hidden defect classification data and the optical image, the accuracy of the classification result is verified. A thermal imaging camera is used to capture the thermal image of the defect area to detect the temperature change caused by the defect. Further verification is carried out by combining the thermal imaging data to confirm the existence and type of the defect.

[0130] As an example of the present invention, refer to Figure 4 As shown, in this example, step S4 includes:

[0131] Step S41: Perform a second overall photographing of the rubber integral mold according to the first overall appearance defect detection image of the rubber integral mold to generate a second overall appearance image of the rubber integral mold;

[0132] Step S42: Perform image preprocessing on the second overall appearance image of the rubber integral mold to generate a standard second overall appearance image of the rubber integral mold; Based on the second overall appearance image of the rubber integral mold, return to step S21 to start the second-side appearance defect detection until a second overall appearance defect detection image of the rubber integral mold is generated;

[0133] Step S43: Upload the first overall appearance defect detection image and the second overall appearance defect detection image of the rubber integral mold to the central processing system for storing appearance defect data, thereby generating an appearance defect detection report for the rubber integral mold.

[0134] In the embodiment of the present invention, by ensuring that the shooting environment of the second side is the same as that of the first side, with uniform illumination and a clean background, the influence of external factors on the image quality is reduced. A high-resolution camera identical to that used for the first side detection is employed to ensure consistent shooting parameters (such as exposure time, focal length). The rubber integral mold is fixed on the shooting table to ensure that the second side is fully visible. Panoramic shooting or multi-angle shooting is carried out to ensure obtaining a full-view image of the second side of the rubber integral mold. Filtering techniques (such as median filtering, Gaussian filtering) are applied to remove image noise. The image quality is improved through methods such as contrast adjustment and histogram equalization. The images are adjusted to a unified size and format for subsequent processing. Image preprocessing is performed on the second overall appearance image of the rubber integral mold to generate a standard second overall appearance image of the rubber integral mold; based on the second overall appearance image of the rubber integral mold, return to step S21 to start the appearance defect detection of the second side until a second overall appearance defect detection image of the rubber integral mold is generated. Convert the first overall appearance defect detection image and the second overall appearance defect detection image of the rubber integral mold into a standard format (such as JPEG, PNG) to ensure data compatibility, and upload the images and related data to the central processing system or database. Obtain defect detection data and images from the central processing system, and generate a detailed defect detection report based on the stored data, including information such as defect type, location, and severity.

[0135] In this specification, an appearance defect detection system for a rubber integral mold based on machine vision is provided, which is used to execute the above-mentioned appearance defect detection method for a rubber integral mold based on machine vision. The appearance defect detection system for a rubber integral mold based on machine vision includes:

[0136] An overall appearance shooting module, which is used to obtain rubber integral mold parameter data; set the inner and outer diameters of the O-ring of the rubber integral mold based on the rubber integral mold parameter data to obtain the adjusted inner and outer diameter data of the built-in O-ring of the rubber integral mold; perform the first overall shooting on the rubber integral mold according to the adjusted inner and outer diameter data of the built-in O-ring of the rubber integral mold to generate a standard first overall appearance image of the rubber integral mold;

[0137] A product trajectory positioning module, which is used to divide the detection area of the standard first overall appearance image of the rubber integral mold to generate rubber integral mold detection area data; compare the rubber integral mold detection area data with a preset detection area quantity range to generate rubber integral mold product specification data; perform trajectory positioning on the standard first overall appearance image of the rubber integral mold based on the trajectory positioning camera according to the rubber integral mold product specification data to generate rubber integral mold product defect shooting trajectory data;

[0138] The product defect detection module is used to adjust the lens focal length of the defect camera based on the rubber integral mold product specification data to generate defect camera focal length adjustment data; perform rubber integral mold appearance defect detection on the rubber integral mold product defect shooting trajectory data based on the defect camera focal length adjustment data to obtain the rubber integral mold overall appearance defect detection image; perform color marking on the rubber integral mold overall appearance defect detection image to generate defective product positioning marking data and normal product positioning marking data; perform color mapping on the standard rubber integral mold first overall appearance image through the defective product positioning marking data and the normal product positioning marking data to generate the rubber integral mold first overall appearance defect detection image;

[0139] The product defect report module is used to perform a second-side overall shooting on the rubber integral mold according to the rubber integral mold first overall appearance defect detection image, generate the rubber integral mold second overall appearance image and perform second-side appearance defect detection until the rubber integral mold second overall appearance defect detection image is generated; upload the rubber integral mold first overall appearance defect detection image and the rubber integral mold second overall appearance defect detection image to the central processing system for appearance defect data storage, thereby generating a rubber integral mold appearance defect detection report.

[0140] The beneficial effects of the present invention are as follows: By ensuring that the inner and outer diameters of the O-ring are set to meet the design requirements, a high-standard first overall appearance image is generated. This helps to accurately detect and evaluate the appearance of the rubber integral mold in subsequent steps. By standardized shooting conditions and parameters, a consistent reference image is provided, providing a reliable reference for subsequent defect detection. By precisely dividing the detection area, it is ensured that each area is properly inspected, reducing the error of manual operation and improving the comprehensiveness and accuracy of detection. Automatically compare the preset range of the number of areas, generate accurate product specification data, thereby optimizing the subsequent trajectory positioning and defect detection processes. By adjusting the focal length of the defect camera, the image clarity is ensured to be the best, improving the detection accuracy of defects. Perform defect detection based on the focal length adjustment data, and accurately locate the defective and normal areas through color marking, improving the accuracy of defect recognition and the readability of detection results. Shoot the second side of the rubber integral mold to ensure a comprehensive inspection of all defective areas and avoid omission. Upload the defect detection images of the first side and the second side to the central processing system to generate a comprehensive appearance defect detection report, which helps to systematically manage and analyze defect data, improving the detection efficiency and the comprehensiveness of the report. Therefore, the present invention improves the accuracy and comprehensiveness of defect detection through precise parameter setting, automated area division, focal length adjustment, high-precision defect detection, and comprehensive report generation.

[0141] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the application documents are intended to be encompassed within the present invention.

[0142] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. A method for detecting appearance defects of rubber molds based on machine vision, characterized in that: The following steps are involved: Step S1: Acquire parameter data of the rubber mold; set the inner and outer diameters of the O-ring of the rubber mold based on the parameter data of the rubber mold to obtain adjustment data of the inner and outer diameters of the O-ring built into the rubber mold; and take an overall photo of the first surface of the rubber mold according to the adjustment data of the inner and outer diameters of the O-ring built into the rubber mold to generate a first overall appearance image of the standard rubber mold; Step S2: dividing the first overall appearance image of the standard rubber mold into detection areas to generate rubber mold detection area data; calculating the number of detection areas on the rubber mold detection area data to obtain detection area quantity data; comparing the detection area quantity data with a preset detection area quantity range to generate rubber mold product specification data; performing trajectory positioning on the first overall appearance image of the standard rubber mold based on the trajectory positioning camera according to the rubber mold product specification data to generate rubber mold product defect shooting trajectory data; Step S3: adjusting the focal length of the defect camera lens based on the rubber mold product specification data to generate defect camera focal length adjustment data; Based on the defect camera focal length adjustment data, the rubber mold product defect shooting trajectory data is used to perform the rubber mold appearance defect detection, thereby obtaining the rubber mold overall appearance defect detection image; the rubber mold overall appearance defect detection image is color-marked to generate defect product positioning mark data and normal product positioning mark data; the standard rubber mold first overall appearance image is color-mapped through the defect product positioning mark data and the normal product positioning mark data to generate the rubber mold first overall appearance defect detection image; Step S4: photographing the second surface of the rubber mold as a whole according to the first overall appearance defect detection image of the rubber mold, generating a second overall appearance image of the rubber mold and performing second surface appearance defect detection until the second overall appearance defect detection image of the rubber mold is generated; uploading the first overall appearance defect detection image of the rubber mold and the second overall appearance defect detection image of the rubber mold to the central processing system for appearance defect data storage, thereby generating a rubber mold appearance defect detection report.

2. The method for detecting appearance defects of rubber molds based on machine vision according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Acquire rubber mold parameter data; Step S12: setting the inner and outer diameters of the O-ring of the rubber mold based on the rubber mold parameter data to obtain inner and outer diameter adjustment data of the O-ring built into the rubber mold; Step S13: photographing the first surface of the rubber mold as a whole according to the inner and outer diameter adjustment data of the O-ring built into the rubber mold to generate a first overall appearance image of the rubber mold; Step S14: performing image preprocessing on the first overall appearance image of the rubber mold to generate a first overall appearance image of a standard rubber mold, wherein the image preprocessing includes image denoising, image filtering and brightness enhancement.

3. The method for detecting appearance defects of rubber molds based on machine vision according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: dividing the first overall appearance image of the standard rubber mold into detection areas to generate rubber mold detection area data; Step S22: Calculating the number of detection areas for the rubber mold detection area data to obtain detection area number data; Step S23: comparing the detection area quantity data with the preset detection area quantity range, when the detection area quantity data is less than the preset detection area quantity range, the corresponding rubber mold is marked as a small-scale rubber mold product; when the detection area quantity data is within the preset detection area quantity range, the corresponding rubber mold is marked as a medium-scale rubber mold product; when the detection area quantity data is greater than the preset detection area quantity range, the corresponding rubber mold is marked as a large-scale rubber mold product; Step S24: integrating the small-scale rubber mold overall products, the medium-scale rubber mold products and the large-scale rubber mold products into product types to generate rubber mold product specification data; Step S25: performing trajectory positioning on the first overall appearance image of the standard rubber mold based on the trajectory positioning camera according to the rubber mold product specification data, and generating rubber mold product defect shooting trajectory data.

4. The method for detecting appearance defects of rubber molds based on machine vision according to claim 3, characterized in that: Step S25 includes the following steps: Step S251: adjusting the pixel value of the track positioning camera based on the rubber mold product specification data to generate the track positioning camera pixel adjustment data; Step S252: performing initial product positioning on the first overall appearance image of the standard rubber mold by using the track positioning camera pixel adjustment data to generate initial positioning data of the rubber mold product; performing target overlap detection on the initial positioning data of the rubber mold product to generate overlapping detection target positioning data and non-overlap detection target positioning data; Step S253: performing multi-dimensional feature detection on the overlapping detection target positioning data to generate overlapping target multi-dimensional feature data, wherein the multi-dimensional feature detection includes area range detection, aspect ratio range detection and color feature detection; performing confidence calculation on the overlapping target multi-dimensional feature data to obtain overlapping target confidence; Step S254: performing overlapping target positioning filtering on the overlapping detection target positioning data according to the overlapping target confidence level to generate overlapping target optimized positioning data; integrating the non-overlapping detection target positioning data and the overlapping target optimized positioning data to generate a rubber mold product positioning data set; Step S255: generating a trajectory for the rubber mold product positioning data set to obtain the rubber mold product defect shooting trajectory data.

5. The method for detecting appearance defects of rubber molds based on machine vision according to claim 4, characterized in that: Step S255 includes the following steps: Step S2551: confirm the product inspection center point of the rubber mold product positioning data set to obtain the center point coordinate data of the rubber mold product; Step S2552: Connect the center point coordinate data of the rubber mold product and the detection area data of the rubber mold product to generate detection area connection data of the rubber mold product; Step S2253: generating a single-line flying shooting trajectory for the connection data of the inspection area of ​​the rubber mold product, and generating defect shooting trajectory data of the rubber mold product.

6. The method for detecting appearance defects of rubber molds based on machine vision according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: adjusting the lens focal length of the defect camera based on the rubber mold product specification data to generate defect camera focal length adjustment data; performing rubber mold appearance defect detection on the rubber mold product defect shooting trajectory data based on the defect camera focal length adjustment data, thereby obtaining an overall appearance defect detection image of the rubber mold; Step S32: extracting color defect features from the rubber mold overall appearance defect detection image to generate rubber mold overall defect feature extraction data; locating defective products from the rubber mold overall appearance defect detection image based on the rubber mold overall defect feature extraction data to generate defective product location data and non-defective product location data; Step S33: marking the defective product location data with a first color to generate defective product location marking data; marking the non-defective product location data with a second color to generate normal product location marking data; Step S34: performing color mapping on the first overall appearance image of the standard rubber mold by using the defective product positioning mark data and the normal product positioning mark data to generate the first overall appearance defect detection image of the rubber mold.

7. The method for detecting appearance defects of rubber molds based on machine vision according to claim 6, characterized in that: Step S32 includes the following steps: Step S321: performing template search and rough positioning on the rubber mold appearance defect detection image to generate rough positioning data of the rubber mold product; performing shape feature extraction on the rubber mold appearance defect detection image to obtain appearance shape feature data of the rubber mold product; performing product fine positioning on the rough positioning data of the rubber mold product according to the appearance shape feature data of the rubber mold product to generate fine positioning data of the rubber product; Step S322: using the rubber product precise positioning data to delineate the inner and outer contours of the rubber mold appearance defect detection image, and generate a partial product extraction image of the rubber mold; performing red, green and blue layer color conversion on the partial product extraction image of the rubber mold, and generate a partial product color conversion image of the rubber mold; Step S323: performing color difference region analysis on the color conversion image of the local product of the rubber mold through the rubber mold defect feature extraction data to generate the color difference region of the local product of the rubber mold, wherein the color difference region of the local product of the rubber mold includes the color difference region of the red layer of the local product of the rubber mold, the color difference region of the green layer of the local product of the rubber mold, and the color difference region of the blue layer of the local product of the rubber mold; performing color difference repeated region screening on the color difference region of the red layer of the local product of the rubber mold, the color difference region of the green layer of the local product of the rubber mold, and the color difference region of the blue layer of the local product of the rubber mold to obtain an image of the dominant defect region of the appearance of the rubber mold product; performing traditional visual defect detection on the image of the dominant defect region of the appearance of the rubber mold product to generate the dominant defect feature data of the appearance of the rubber mold product; Step S324: screening the color difference non-repetitive area of ​​the local color difference area of ​​the rubber mold product according to the image of the obvious defect area of ​​the rubber mold product appearance to obtain the image of the hidden defect area of ​​the rubber mold product appearance; performing deep AI visual defect detection on the image of the hidden defect area of ​​the rubber mold product appearance to generate the hidden defect feature data of the rubber mold product appearance; Step S325: Integrate the hidden defect feature data of the rubber mold product appearance and the explicit defect feature data of the rubber mold product appearance to generate rubber mold defect feature extraction data; locate the defective product of the rubber mold appearance defect detection image according to the rubber mold defect feature extraction data to generate defective product location data and non-defective product location data.

8. The method for detecting appearance defects of rubber molds based on machine vision according to claim 7, characterized in that: Deep AI visual defect detection of hidden defect areas on the appearance of rubber mold products includes: Perform regional focusing analysis on the image of the hidden defect area of ​​the appearance of the whole rubber mold product to generate the focused hidden defect feature data of the whole rubber mold product; divide the focused hidden defect feature data of the whole rubber mold product into data sets to generate a model training set and a model test set; The model training set is trained by deep learning classification algorithm to generate a hidden defect classification pre-model; the hidden defect classification pre-model is tested by model test set to generate a hidden defect classification model; the image of the hidden defect area of ​​the appearance of the rubber mold product is imported into the hidden defect classification model to perform hidden defect classification and generate hidden defect classification data; Multimodal defect verification is performed on the image of the hidden defect area on the appearance of the rubber whole mold product through the hidden defect classification data to generate hidden defect feature data of the appearance of the rubber whole mold product, wherein the multimodal defect verification includes optical verification and thermal imaging verification.

9. The method for detecting appearance defects of rubber molds based on machine vision according to claim 3, characterized in that: Step S4 includes the following steps: Step S41: photographing the second surface of the rubber mold as a whole according to the first overall appearance defect detection image of the rubber mold to generate a second overall appearance image of the rubber mold; Step S42: performing image preprocessing on the second overall appearance image of the rubber mold to generate a second overall appearance image of the standard rubber mold; returning to step S21 based on the second overall appearance image of the rubber mold to start second surface appearance defect detection until a second overall appearance defect detection image of the rubber mold is generated; Step S43: uploading the first overall appearance defect detection image of the rubber mold and the second overall appearance defect detection image of the rubber mold to the central processing system for storing appearance defect data, thereby generating a rubber mold appearance defect detection report.

10. A rubber mold appearance defect detection system based on machine vision, characterized in that: Used to perform the method for detecting appearance defects of a rubber mold based on machine vision as claimed in claim 1, the rubber mold appearance defect detection system based on machine vision comprises: The overall appearance shooting module is used to obtain the parameter data of the rubber mold; set the inner and outer diameters of the O-ring of the rubber mold based on the parameter data of the rubber mold to obtain the inner and outer diameter adjustment data of the O-ring built into the rubber mold; shoot the first surface of the rubber mold as a whole according to the inner and outer diameter adjustment data of the O-ring built into the rubber mold to generate the first overall appearance image of the standard rubber mold; The product trajectory positioning module is used to divide the first overall appearance image of the standard rubber mold into detection areas to generate the rubber mold detection area data; calculate the number of detection areas for the rubber mold detection area data to obtain the detection area quantity data; compare the detection area quantity data with the preset detection area quantity range to generate the rubber mold product specification data; based on the trajectory positioning camera, the first overall appearance image of the standard rubber mold is tracked according to the rubber mold product specification data to generate the rubber mold product defect shooting trajectory data; A product defect detection module is used to adjust the lens focal length of a defect camera based on the product specification data of the rubber whole mold to generate defect camera focal length adjustment data; perform rubber whole mold appearance defect detection on the rubber whole mold product defect shooting trajectory data based on the defect camera focal length adjustment data, thereby obtaining an overall appearance defect detection image of the rubber whole mold; perform color marking on the overall appearance defect detection image of the rubber whole mold to generate defect product positioning marking data and normal product positioning marking data; perform color mapping on a first overall appearance image of a standard rubber whole mold through the defect product positioning marking data and the normal product positioning marking data to generate a first overall appearance defect detection image of the rubber whole mold; The product defect report module is used to shoot the second side of the rubber mold as a whole according to the first overall appearance defect detection image of the rubber mold, generate a second overall appearance image of the rubber mold and perform second side appearance defect detection until the second overall appearance defect detection image of the rubber mold is generated; upload the first overall appearance defect detection image of the rubber mold and the second overall appearance defect detection image of the rubber mold to the central processing system for appearance defect data storage, thereby generating a rubber mold appearance defect detection report.

Images