Industrial part surface defect identification system and identification method

Abstract

The invention relates to the technical field of part surface defects, in particular to an industrial part surface defect recognition system and method, and the system comprises a high-resolution industrial camera which is used for collecting an industrial part surface image and is provided with an automatic focusing and light source compensation module; therefore, clear images can be obtained in different light environments; the image preprocessing unit has the functions of denoising, contrast enhancement and edge detection and ensures the saliency of defect features in the original image; a defect detection model based on deep learning, wherein the model is designed through a convolutional neural network structure and can automatically identify and classify various surface defects including but not limited to scratches, pits, cracks, air holes and the like; the defect positioning module is used for accurately marking the specific position of the defect on the surface of the part in combination with an image processing algorithm and calibration information; the system further comprises a data storage and management unit, a system control unit and an abnormity alarm module.

Description

Technical Field

[0001] The present invention relates to the technical field of part surface defects, and in particular to an industrial part surface defect recognition system and recognition method. Background Art

[0002] With the continuous improvement of modern industrial manufacturing, the quality control of industrial parts has become particularly important. Surface defects on parts not only affect the appearance of the parts, but also seriously affect their mechanical properties and service life, which is directly related to the safety and reliability of the entire mechanical equipment. Traditional surface defect detection of parts mainly relies on manual visual inspection. Due to human fatigue, experience differences and environmental factors, the detection results are highly subjective and unstable. Especially on high-capacity assembly lines, manual inspection is difficult to meet the dual requirements of efficiency and accuracy, resulting in a high rate of missed defect detection and increased production costs. In addition, parts with complex shapes and diverse materials also make manual inspection more difficult. For this reason, the industrial field urgently needs to introduce automated and intelligent defect recognition technology. Through high-precision image acquisition and advanced image processing methods, it can achieve rapid and accurate detection of surface defects on parts, improve the quality control level of the production process, and ensure product quality and production efficiency.

[0003] In recent years, with the rapid development of computer vision and artificial intelligence technologies, machine vision-based surface defect recognition systems for industrial parts have become a hot topic in research and application. These systems typically combine high-resolution industrial cameras, multi-angle lighting techniques, and deep learning algorithms to achieve automatic defect detection and classification. Deep learning models, particularly convolutional neural networks (CNNs), have demonstrated excellent feature extraction and classification capabilities in image recognition, effectively identifying a variety of complex defects such as scratches, pits, cracks, and pores. However, practical applications still face numerous challenges, such as varying lighting conditions in complex environments, large variations in image features due to diverse surface materials, insufficient defect samples, and difficulty in labeling. These challenges limit improvements in system detection accuracy and stability. Therefore, designing a defect recognition system that adapts to the changing conditions of industrial sites while achieving both high accuracy and high efficiency has become a key technical challenge that urgently needs to be addressed. This has prompted the industry to continuously explore advanced technologies such as multimodal data fusion, intelligent preprocessing, and online learning to meet the actual needs of industrial production.

[0004] In view of the above situation, in order to overcome the above technical problems, the present invention designs an industrial parts surface defect recognition system and recognition method to solve the above technical problems. Summary of the Invention

[0005] The technical objective of this invention is to design a defect recognition system that adapts to the changing conditions of industrial sites and takes into account both high precision and high efficiency, combining advanced technologies such as multimodal data fusion, intelligent preprocessing, and online learning to meet the actual needs of industrial production.

[0006] In order to achieve the above technical objectives, the present invention provides the following technical solutions:

[0007] This system, designed through a multi-module collaborative design, enables efficient and accurate identification of surface defects on industrial parts. The system's core components include a high-resolution industrial camera, an image preprocessing unit, a deep learning-based defect detection model, a defect localization module, a data storage and management unit, a system control unit, and an anomaly alarm module. Each module offers independent yet complementary functions, forming a complete automated defect detection solution.

[0008] First, the system is equipped with a high-resolution industrial camera for capturing surface images of industrial parts. This industrial camera not only provides high-definition imaging capabilities but also features an autofocus module and a light source compensation module. The autofocus function ensures that the camera can quickly and accurately adjust the focus when capturing parts of varying sizes and shapes, ensuring clear image detail. The light source compensation module automatically adjusts the camera's exposure parameters based on ambient lighting conditions, effectively preventing image blur or loss of detail caused by excessive or insufficient lighting. This ensures high-quality images are captured even in industrial environments with drastic changes in brightness, providing a solid foundation for subsequent analysis.

[0009] Secondly, the image preprocessing unit is responsible for optimizing the captured images, primarily including functions such as denoising, contrast enhancement, and edge detection. Through advanced image filtering algorithms, the system effectively removes background noise and stray light interference common in industrial environments, improving image clarity and signal-to-noise ratio. Contrast enhancement technology also emphasizes the grayscale difference between defective and normal areas in the image, making defect features more prominent. The edge detection function further highlights the morphological characteristics of defects by extracting object boundaries and texture contours in the image, ensuring that subsequent deep learning models can accurately capture defect details and enhance recognition effectiveness.

[0010] The deep learning-based defect detection model is the system's intelligent core. Designed using a convolutional neural network (CNN) architecture, it boasts powerful automatic feature extraction and classification capabilities. During the training phase, the model learns from a large amount of labeled industrial part surface image data, automatically identifying a variety of common and complex surface defects, including scratches, pits, cracks, and pores. This end-to-end automated recognition process significantly reduces manual intervention and improves detection speed and accuracy. The model output includes not only the defect type but also a confidence score, enabling operators to assess defect severity.

[0011] The defect localization module combines image processing algorithms with camera calibration information to accurately convert defect locations in 2D images into 3D surface coordinates of the part, achieving precise spatial localization of the defect. This localization information can assist in subsequent manual re-inspection or automated repair processes, improving overall production line quality management and operational efficiency.

[0012] The data storage and management unit is responsible for storing inspection results, raw images, and processed image data, supporting defect statistics and historical data analysis. This module also features data backup and retrieval capabilities, facilitating production quality tracking and continuous improvement. It also supports cloud-based data synchronization, enabling remote monitoring and big data analysis applications.

[0013] The system control unit serves as the overall coordination center, managing the coordinated operation of various functional modules. Through a human-machine interface, the system displays real-time inspection results, defect information, and statistical data, facilitating operator monitoring and management. Furthermore, the system features an abnormality alarm module. When a serious defect is detected, it automatically triggers audible and visual alarms and remote notifications, prompting maintenance personnel to intervene promptly to ensure production safety and product quality.

[0014] In summary, this industrial parts surface defect recognition system achieves fast, accurate and intelligent detection of part defects in industrial sites by integrating high-performance imaging equipment, intelligent image processing and deep learning algorithms. It has good adaptability and practicality and meets the needs of modern industrial automation quality control.

[0015] The industrial camera utilizes a multi-angle, multi-light source combination lighting design. By flexibly adjusting the illumination angle and intensity of different light sources, it can capture images of industrial parts from multiple perspectives and under multiple lighting conditions. This design not only effectively reduces shadows and reflections caused by a single light source, but also enhances the visibility of subtle surface structures, making surface defects such as dents, cracks, and scratches easier to detect. The detection of tiny cracks and shallow dents is particularly effective, greatly improving the accuracy and stability of the system's overall defect recognition, meeting the needs of high-precision inspection in complex industrial environments.

[0016] The image preprocessing unit integrates an advanced noise suppression algorithm based on adaptive filtering, effectively filtering and suppressing the complex and changing background noise found in industrial sites, significantly improving image clarity and signal-to-noise ratio. Furthermore, the unit employs histogram equalization technology to optimize the image's grayscale distribution, enhancing contrast and detail, particularly highlighting textures and edges in defect areas. These preprocessing steps ensure higher-quality image data input to the subsequent deep learning defect recognition model, thereby improving the accuracy and robustness of defect detection and adapting to diverse industrial inspection environments.

[0017] During the training phase, the deep learning defect detection model utilizes a rich dataset with multi-category annotations, covering a variety of common and complex surface defect types, including scratches, pits, cracks, and pores. By introducing data augmentation techniques such as rotation, scaling, and flipping, the model can learn a wider range of defect manifestations, enhancing its ability to recognize different defect characteristics. Furthermore, the application of transfer learning technology enables the model to leverage common features from the pre-trained network, rapidly adapting to variations in part materials and surface conditions, improving generalization and robustness. These combined techniques ensure that the model maintains high-precision and stable defect recognition performance in complex industrial environments.

[0018] The defect localization module, based on camera calibration technology, accurately acquires the internal and external parameters of industrial cameras. Combined with a spatial transformation algorithm, it converts the coordinates of defects detected in two-dimensional images into three-dimensional surface coordinates of the part. This conversion accurately reflects the specific location and form of the defect on the actual part surface, enabling high-precision spatial localization of the defect. This positioning information allows operators to quickly and accurately locate the defect area, greatly facilitating subsequent manual re-inspection and repair processes, improving production efficiency and repair quality. Furthermore, the module supports multi-view data fusion, further enhancing positioning accuracy and stability to meet the application requirements of complex industrial environments.

[0019] The data storage and management unit boasts powerful cloud-based synchronization capabilities, enabling real-time upload of all data generated during the inspection process, including original images, processing results, and defect analysis reports, to a cloud server. Through the cloud platform, managers can remotely monitor the production line's operating status and quality inspections, promptly identifying and addressing anomalies. Furthermore, the system supports archiving and comparative analysis of historical inspection data, helping companies track quality trends and identify potential production issues. Furthermore, combined with big data mining technology, it enables in-depth analysis of massive amounts of inspection data, optimizing production processes and quality control strategies, and thereby comprehensively improving the quality management and operational efficiency of industrial parts production.

[0020] The system control unit is equipped with multiple built-in industry-standard communication interfaces, such as Ethernet, Modbus, and Profibus, enabling seamless connection and data exchange with other automated equipment on the production line. Through these interfaces, the system can share detection data and device status information in real time, enabling coordinated control and operation across multiple devices, ensuring a highly automated and intelligent production process. This design not only improves the system's compatibility and scalability, but also effectively reduces manual intervention and operational errors, significantly improving the operational efficiency and production quality of the entire production line, and providing stable and reliable intelligent detection support for modern industrial manufacturing.

[0021] The human-machine interaction interface is designed to be intuitive and friendly, and can display a real-time visual report of surface defects of industrial parts, including key information such as defect type, location and severity, to help operators quickly understand the test results. At the same time, the interface also dynamically displays real-time detection curves to reflect quality fluctuations during the production process, facilitating the timely detection of anomalies. In addition, the system supports historical trend analysis functions, which can collect and compare past test data to assist in quality management and decision-making. In order to adapt to the needs of different regions and operators, the human-machine interface provides multi-language switching functions and allows users to customize alarm threshold settings to achieve personalized configuration and improve the convenience and flexibility of operation.

[0022] The abnormality alarm module integrates an audible and visual alarm device and remote notification function, and can automatically generate graded alarms based on the type and severity of the detected defect. For minor defects, the system triggers a low-intensity prompt, while for serious or critical defects, it activates a high-intensity audible and visual alarm to draw the attention of on-site personnel. At the same time, the module supports multiple remote notification methods such as SMS, email, and mobile application push, sending abnormality information to relevant maintenance personnel and managers in real time, ensuring that they are immediately aware of production abnormalities and can quickly take countermeasures. This function effectively improves the safety management level of the production site, reduces potential risks, and ensures the continuity of the production process and the stability of product quality.

[0023] A method for identifying surface defects of industrial parts, which is used in the above-mentioned industrial parts surface defect identification system; the steps of the method are as follows:

[0024] Step 1: Capture images of industrial part surfaces. Using a high-resolution industrial camera, the surface of the inspected part is photographed from multiple angles and with multiple light sources. Focus and light intensity are automatically adjusted to accommodate different part shapes and materials, resulting in clear, detailed raw image data, providing a high-quality image foundation for subsequent processing.

[0025] Step 2: Image preprocessing. The collected original image is denoised using an adaptive filtering algorithm to suppress noise introduced by the environment and equipment. Image enhancement techniques such as histogram equalization are used to enhance image contrast and edge information, ensuring that defect features are fully visible in the image, preparing for input into the defect detection model.

[0026] Step 3: Defect Detection and Classification. The preprocessed image is fed into a deep learning model built on a convolutional neural network. The model automatically identifies various surface defect types, including scratches, pits, cracks, and pores. It accurately classifies defects through multi-category output and simultaneously outputs a defect confidence score to improve detection reliability.

[0027] Step 4: Defect Location and Annotation. Combining camera calibration data and image processing algorithms, the detected defect's 2D image coordinates are converted into the part's 3D surface coordinates, enabling precise spatial location of the defect. The defect's location is then annotated on the image and 3D model, facilitating quick location and subsequent repairs.

[0028] Step 5: Result storage and alarm processing. Upload defect detection results and related image data to local and cloud data storage systems to support historical data management and quality tracking. At the same time, audio and visual alarms or remote notifications are triggered based on the defect type and severity, prompting operators to respond promptly to ensure production line quality and safety.

[0029] The beneficial effects of the present invention are as follows:

[0030] (1) The present invention provides a surface defect recognition system for industrial parts, which significantly improves the automation and intelligence level of part surface defect detection. By combining high-resolution industrial cameras with multi-angle, multi-light source combined lighting technology, the system can collect high-quality multi-view images in complex and changing industrial environments, effectively enhancing the ability to identify subtle defects such as small cracks and dents. At the same time, it integrates advanced image preprocessing technology and a deep learning-based defect detection model, and can automatically complete a series of difficult tasks such as denoising, enhancement, defect classification and positioning, greatly reducing the subjectivity and workload of manual inspection, significantly improving the accuracy and stability of inspection, and ensuring the consistency and reliability of product quality.

[0031] (2) The system also has comprehensive data management and intelligent alarm functions. Through data storage and cloud synchronization, real-time upload of test data, remote monitoring and historical data analysis are achieved, providing strong data support for production quality control. The abnormal alarm module can automatically trigger sound and light alarms and remote notifications according to the severity of the defect, and promptly remind operators to take measures to prevent serious defects from flowing into subsequent production links or the market, effectively ensuring production safety and normal operation of equipment. The system control unit supports seamless docking with other equipment on the production line, realizes linkage control, improves the overall automation level, and helps enterprises achieve intelligent manufacturing and efficient production. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0033] The above and other aspects of the present invention will now be described, by way of example only, with reference to the accompanying drawings, in which:

[0034] Figure 1 Schematic diagram of the system structure of the present invention;

[0035] Figure 2 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0036] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0037] like Figure 1-2 Figure 1 shows a surface defect recognition system for industrial parts. This system utilizes a multi-module collaborative design to achieve efficient and accurate identification of surface defects on industrial parts. The system's core components include a high-resolution industrial camera, an image preprocessing unit, a deep learning-based defect detection model, a defect localization module, a data storage and management unit, a system control unit, and an anomaly alarm module. Each module offers independent yet complementary functions, forming a complete automated defect detection solution.

[0038] First, the system is equipped with a high-resolution industrial camera for capturing surface images of industrial parts. This industrial camera not only provides high-definition imaging capabilities but also features an autofocus module and a light source compensation module. The autofocus function ensures that the camera can quickly and accurately adjust the focus when capturing parts of varying sizes and shapes, ensuring clear image detail. The light source compensation module automatically adjusts the camera's exposure parameters based on ambient lighting conditions, effectively preventing image blur or loss of detail caused by excessive or insufficient lighting. This ensures high-quality images are captured even in industrial environments with drastic changes in brightness, providing a solid foundation for subsequent analysis.

[0039] Secondly, the image preprocessing unit is responsible for optimizing the captured images, primarily including functions such as denoising, contrast enhancement, and edge detection. Through advanced image filtering algorithms, the system effectively removes background noise and stray light interference common in industrial environments, improving image clarity and signal-to-noise ratio. Contrast enhancement technology also emphasizes the grayscale difference between defective and normal areas in the image, making defect features more prominent. The edge detection function further highlights the morphological characteristics of defects by extracting object boundaries and texture contours in the image, ensuring that subsequent deep learning models can accurately capture defect details and enhance recognition effectiveness.

[0040] The deep learning-based defect detection model is the system's intelligent core. Designed using a convolutional neural network (CNN) architecture, it boasts powerful automatic feature extraction and classification capabilities. During the training phase, the model learns from a large amount of labeled industrial part surface image data, automatically identifying a variety of common and complex surface defects, including scratches, pits, cracks, and pores. This end-to-end automated recognition process significantly reduces manual intervention and improves detection speed and accuracy. The model output includes not only the defect type but also a confidence score, enabling operators to assess defect severity.

[0041] The defect localization module combines image processing algorithms with camera calibration information to accurately convert defect locations in 2D images into 3D surface coordinates of the part, achieving precise spatial localization of the defect. This localization information can assist in subsequent manual re-inspection or automated repair processes, improving overall production line quality management and operational efficiency.

[0042] The data storage and management unit is responsible for storing inspection results, raw images, and processed image data, supporting defect statistics and historical data analysis. This module also features data backup and retrieval capabilities, facilitating production quality tracking and continuous improvement. It also supports cloud-based data synchronization, enabling remote monitoring and big data analysis applications.

[0043] The system control unit serves as the overall coordination center, managing the coordinated operation of various functional modules. Through a human-machine interface, the system displays real-time inspection results, defect information, and statistical data, facilitating operator monitoring and management. Furthermore, the system features an abnormality alarm module. When a serious defect is detected, it automatically triggers audible and visual alarms and remote notifications, prompting maintenance personnel to intervene promptly to ensure production safety and product quality.

[0044] In summary, this industrial parts surface defect recognition system achieves fast, accurate and intelligent detection of part defects in industrial sites by integrating high-performance imaging equipment, intelligent image processing and deep learning algorithms. It has good adaptability and practicality and meets the needs of modern industrial automation quality control.

[0045] The industrial camera utilizes a multi-angle, multi-light source combination lighting design. By flexibly adjusting the illumination angle and intensity of different light sources, it can capture images of industrial parts from multiple perspectives and under multiple lighting conditions. This design not only effectively reduces shadows and reflections caused by a single light source, but also enhances the visibility of subtle surface structures, making surface defects such as dents, cracks, and scratches easier to detect. The detection of tiny cracks and shallow dents is particularly effective, greatly improving the accuracy and stability of the system's overall defect recognition, meeting the needs of high-precision inspection in complex industrial environments.

[0046] The image preprocessing unit integrates an advanced noise suppression algorithm based on adaptive filtering, effectively filtering and suppressing the complex and changing background noise found in industrial sites, significantly improving image clarity and signal-to-noise ratio. Furthermore, the unit employs histogram equalization technology to optimize the image's grayscale distribution, enhancing contrast and detail, particularly highlighting textures and edges in defect areas. These preprocessing steps ensure higher-quality image data input to the subsequent deep learning defect recognition model, thereby improving the accuracy and robustness of defect detection and adapting to diverse industrial inspection environments.

[0047] During the training phase, the deep learning defect detection model utilizes a rich dataset with multi-category annotations, covering a variety of common and complex surface defect types, including scratches, pits, cracks, and pores. By introducing data augmentation techniques such as rotation, scaling, and flipping, the model can learn a wider range of defect manifestations, enhancing its ability to recognize different defect characteristics. Furthermore, the application of transfer learning technology enables the model to leverage common features from the pre-trained network, rapidly adapting to variations in part materials and surface conditions, improving generalization and robustness. These combined techniques ensure that the model maintains high-precision and stable defect recognition performance in complex industrial environments.

[0048] The defect localization module, based on camera calibration technology, accurately acquires the internal and external parameters of industrial cameras. Combined with a spatial transformation algorithm, it converts the coordinates of defects detected in two-dimensional images into three-dimensional surface coordinates of the part. This conversion accurately reflects the specific location and form of the defect on the actual part surface, enabling high-precision spatial localization of the defect. This positioning information allows operators to quickly and accurately locate the defect area, greatly facilitating subsequent manual re-inspection and repair processes, improving production efficiency and repair quality. Furthermore, the module supports multi-view data fusion, further enhancing positioning accuracy and stability to meet the application requirements of complex industrial environments.

[0049] The data storage and management unit boasts powerful cloud-based synchronization capabilities, enabling real-time upload of all data generated during the inspection process, including original images, processing results, and defect analysis reports, to a cloud server. Through the cloud platform, managers can remotely monitor the production line's operating status and quality inspections, promptly identifying and addressing anomalies. Furthermore, the system supports archiving and comparative analysis of historical inspection data, helping companies track quality trends and identify potential production issues. Furthermore, combined with big data mining technology, it enables in-depth analysis of massive amounts of inspection data, optimizing production processes and quality control strategies, and thereby comprehensively improving the quality management and operational efficiency of industrial parts production.

[0050] The system control unit is equipped with multiple built-in industry-standard communication interfaces, such as Ethernet, Modbus, and Profibus, enabling seamless connection and data exchange with other automated equipment on the production line. Through these interfaces, the system can share detection data and device status information in real time, enabling coordinated control and operation across multiple devices, ensuring a highly automated and intelligent production process. This design not only improves the system's compatibility and scalability, but also effectively reduces manual intervention and operational errors, significantly improving the operational efficiency and production quality of the entire production line, and providing stable and reliable intelligent detection support for modern industrial manufacturing.

[0051] The human-machine interaction interface is designed to be intuitive and friendly, and can display a real-time visual report of surface defects of industrial parts, including key information such as defect type, location and severity, to help operators quickly understand the test results. At the same time, the interface also dynamically displays real-time detection curves to reflect quality fluctuations during the production process, facilitating the timely detection of anomalies. In addition, the system supports historical trend analysis functions, which can collect and compare past test data to assist in quality management and decision-making. In order to adapt to the needs of different regions and operators, the human-machine interface provides multi-language switching functions and allows users to customize alarm threshold settings to achieve personalized configuration and improve the convenience and flexibility of operation.

[0052] The abnormality alarm module integrates an audible and visual alarm device and remote notification function, and can automatically generate graded alarms based on the type and severity of the detected defect. For minor defects, the system triggers a low-intensity prompt, while for serious or critical defects, it activates a high-intensity audible and visual alarm to draw the attention of on-site personnel. At the same time, the module supports multiple remote notification methods such as SMS, email, and mobile application push, sending abnormality information to relevant maintenance personnel and managers in real time, ensuring that they are immediately aware of production abnormalities and can quickly take countermeasures. This function effectively improves the safety management level of the production site, reduces potential risks, and ensures the continuity of the production process and the stability of product quality.

[0053] A method for identifying surface defects of industrial parts, which is used in the above-mentioned industrial parts surface defect identification system; the steps of the method are as follows:

[0054] Step 1: Capture images of industrial part surfaces. Using a high-resolution industrial camera, the surface of the inspected part is photographed from multiple angles and with multiple light sources. Focus and light intensity are automatically adjusted to accommodate different part shapes and materials, resulting in clear, detailed raw image data, providing a high-quality image foundation for subsequent processing.

[0055] Step 2: Image preprocessing. The collected original image is denoised using an adaptive filtering algorithm to suppress noise introduced by the environment and equipment. Image enhancement techniques such as histogram equalization are used to enhance image contrast and edge information, ensuring that defect features are fully visible in the image, preparing for input into the defect detection model.

[0056] Step 3: Defect Detection and Classification. The preprocessed image is fed into a deep learning model built on a convolutional neural network. The model automatically identifies various surface defect types, including scratches, pits, cracks, and pores. It accurately classifies defects through multi-category output and simultaneously outputs a defect confidence score to improve detection reliability.

[0057] Step 4: Defect Location and Annotation. Combining camera calibration data and image processing algorithms, the detected defect's 2D image coordinates are converted into the part's 3D surface coordinates, enabling precise spatial location of the defect. The defect's location is then annotated on the image and 3D model, facilitating quick location and subsequent repairs.

[0058] Step 5: Result storage and alarm processing. Upload defect detection results and related image data to local and cloud data storage systems to support historical data management and quality tracking. At the same time, audio and visual alarms or remote notifications are triggered based on the defect type and severity, prompting operators to respond promptly to ensure production line quality and safety.

[0059] This method first uses a high-resolution industrial camera to capture the surface of the inspected part from multiple angles and with multiple light sources. The camera automatically adjusts the focus and illumination intensity to accommodate the shape and material of the part, generating clear, detailed raw image data. This provides a high-quality image foundation for subsequent processing. The captured raw images are then denoised using an adaptive filtering algorithm to effectively suppress environmental and equipment noise. Image enhancement techniques such as histogram equalization are also used to enhance image contrast and edge information, ensuring that defect features are fully visible in the image and preparing the input for the defect detection model. The preprocessed images are then fed into a deep learning model built using a convolutional neural network. The model automatically identifies various surface defect types, including scratches, pits, cracks, and pores, and accurately classifies them through multi-category output. It also outputs a defect confidence score to enhance detection reliability. Furthermore, by combining camera calibration data with image processing algorithms, the detected defect's 2D image coordinates are converted to 3D surface coordinates on the part, enabling precise spatial localization of the defect. The defect location is then annotated on the image and 3D model, facilitating rapid location and subsequent repair by operators. Finally, the defect detection results and related image data are uploaded to local and cloud data storage systems to support historical data management and quality tracking. At the same time, audio and visual alarms or remote notifications are triggered according to the type and severity of the defect, prompting operators to respond in time, thereby ensuring the quality and safety of the production line.

[0060] Various modifications to the present disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the scope of the present disclosure. Therefore, the present disclosure is not limited to the examples and designs described herein, but should be given the widest scope consistent with the principles and novel features disclosed herein. Although one or more exemplary embodiments of the present disclosure have been described with reference to the accompanying drawings, it will be understood by those skilled in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the present disclosure as defined in the appended claims.

Claims

1. A surface defect recognition system for industrial parts, characterized in that: The system comprises: High-resolution industrial camera: used to capture images of the surface of industrial parts. The industrial camera is equipped with automatic focus and light source compensation modules to ensure clear images in different lighting environments. Image preprocessing unit: includes denoising, contrast enhancement and edge detection functions to ensure the prominence of defect features in the original image; Deep learning-based defect detection model: This model, designed using a convolutional neural network structure, can automatically identify and classify various surface defects including but not limited to scratches, pits, cracks, and pores; Defect location module: combines image processing algorithms and calibration information to accurately mark the specific location of defects on the part surface; Data storage and management unit: used to save inspection results, original and processed image data, and support defect statistics and historical data analysis; System control unit: coordinates the collaborative work of various modules and displays the test results and related information through the human-computer interaction interface; Abnormal alarm module: When a serious defect is detected, the alarm is automatically triggered to prompt the operator to deal with it in time.

2. The industrial parts surface defect recognition system according to claim 1, characterized in that: The industrial camera adopts a multi-angle and multi-light source combined lighting design. By changing the lighting angle and intensity, it obtains multi-view images to improve the accuracy of defect recognition, especially significantly improving the detection effect of dents and tiny cracks.

3. The industrial parts surface defect recognition system according to claim 1, characterized in that: The image preprocessing unit integrates a noise suppression algorithm based on adaptive filtering, which can effectively filter out complex background noise in industrial sites. At the same time, it uses a histogram equalization method to enhance image details and ensure the input quality of the subsequent defect recognition model.

4. The industrial parts surface defect recognition system according to claim 1, characterized in that: The deep learning defect detection model uses a multi-category labeled dataset during the training process, and through data enhancement and transfer learning techniques, improves the model's generalization ability for various defect types, ensuring accurate identification of parts with different materials and surface conditions.

5. The industrial parts surface defect recognition system according to claim 1, characterized in that: The defect location module converts the two-dimensional image coordinates into the three-dimensional surface coordinates of the part through camera calibration and space transformation algorithm, thereby achieving accurate location of defects and facilitating subsequent manual re-inspection and repair processes.

6. The industrial parts surface defect recognition system according to claim 1, characterized in that: The data storage and management unit supports cloud synchronization function, and can upload detection data to the cloud server in real time, realize remote monitoring, historical data comparison analysis and big data mining, and improve the level of production quality control.

7. The industrial parts surface defect recognition system according to claim 1, characterized in that: The system control unit has a built-in industrial standard communication interface, which supports seamless connection with other equipment on the production line, realizes data sharing and linkage control, and improves automated production efficiency.

8. The industrial parts surface defect recognition system according to claim 1, characterized in that: The human-computer interaction interface provides visual defect reports, real-time detection curves and historical trend analysis, and supports multi-language switching and customized alarm threshold settings, making it convenient for different operators to use.

9. The industrial parts surface defect recognition system according to claim 1, characterized in that: The abnormal alarm module includes sound and light alarm and remote notification functions, which can automatically generate graded alarms according to the severity of the defects and promptly notify maintenance personnel via text messages, emails, etc. to ensure production safety.

10. A method for identifying surface defects of industrial parts, the method being used in conjunction with an industrial part surface defect identification system according to any one of claims 1 to 9; characterized in that: The steps of the method are as follows: Step 1: Surface image acquisition for industrial parts. High-resolution industrial cameras are used to capture images of the surface of the part being inspected from multiple angles and using multiple light sources. Focus and light intensity are automatically adjusted to accommodate different part shapes and materials, obtaining clear, detailed raw image data to provide a high-quality image foundation for subsequent processing. Step 2: Image preprocessing: De-noising the original image. Adaptive filtering algorithms are used to suppress noise from the environment and equipment. Image enhancement techniques such as histogram equalization are used to improve image contrast and edge information, ensuring that defect features are fully displayed in the image, preparing for input into the defect detection model. Step 3: Defect detection and classification: The preprocessed image is fed into a deep learning model based on a convolutional neural network. The model automatically identifies various surface defect types, including scratches, pits, cracks, and pores. It accurately classifies defects through multi-category output and simultaneously outputs a defect confidence score to improve detection reliability. Step 4: Defect location and annotation. Combining camera calibration data and image processing algorithms, the detected defect's 2D image coordinates are converted into the part's 3D surface coordinates, achieving precise spatial location of the defect. The defect's location is then annotated on the image and 3D model, facilitating quick location and subsequent repairs by operators. Step 5: Result storage and alarm processing: Upload the defect detection results and related image data to local and cloud data storage systems to support historical data management and quality tracking; at the same time, trigger audible and visual alarms or remote notifications based on the defect type and severity, prompting operators to respond in a timely manner to ensure production line quality and safety.

Images