Flexible substrate adaptive defect detection system based on machine vision

Abstract

The invention provides a flexible substrate adaptive defect detection system based on machine vision, and relates to the technical field of defect detection, and the system comprises an image obtaining module which is used for obtaining real-time image data of a flexible substrate in a target position based on preset image collection equipment, and carrying out image processing to obtain a first processing image set; the recognition processing module is used for performing feature extraction on each first processing image in the first processing image set to obtain a first feature image set; the defect detection module is used for classifying each feature in the first feature image set and performing defect detection based on a classification result; and the self-adaptive adjustment module is used for carrying out self-adaptive adjustment based on the defect detection result. By processing the image data of the target flexible substrate, the feature images are extracted and classified for defect detection, so that the defect detection of the target flexible substrate is more accurate, an adaptive adjustment strategy can be timely and accurately provided, and the defect detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of defect detection, and in particular to a flexible substrate adaptive defect detection system based on machine vision. Background Art

[0002] At present, as an important substrate for electronic parts, flexible substrates play a vital role in the manufacture of electronic products. However, during the production process, due to factors such as complex processes and material characteristics, a variety of defects may appear on FPCB, such as missing holes, mouse bites, open circuits, short circuits, burrs, fake copper, etc. These defects not only affect product quality, but may also cause performance degradation or even failure of electronic equipment.

[0003] Traditional defect detection methods, such as manual visual inspection, instrument online inspection, and functional testing, have problems such as low detection efficiency, insufficient accuracy, and high cost. With the development of machine vision technology, visual inspection technologies such as automated optical inspection (AOI) and machine vision inspection (MVI) have gradually become mainstream. These technologies can efficiently and accurately identify defects and improve product quality and production efficiency through image acquisition, processing, and analysis.

[0004] However, the adaptability of existing machine vision inspection systems to complex and changeable materials such as flexible substrates still needs to be improved.

[0005] Therefore, the present invention provides a flexible substrate adaptive defect detection system based on machine vision. Summary of the invention

[0006] The present invention provides a flexible substrate adaptive defect detection system based on machine vision, which is used to solve the problems of inaccurate defect detection and untimely detection in the prior art.

[0007] The present invention provides a flexible substrate adaptive defect detection system based on machine vision, comprising: An image acquisition module is used to acquire real-time image data of a target flexible substrate in a target position based on a preset image acquisition device, and perform image processing to obtain a first processed image set; Recognition processing module: used for extracting features from each first processed image in the first processed image set based on deep learning technology to obtain a first feature image set; Defect detection module: used for classifying each feature in the first feature image set and performing defect detection based on the classification result; Adaptive adjustment module: used to adaptively adjust the target flexible substrate based on the defect detection results.

[0008] The image acquisition module provided by the present invention comprises: An image acquisition unit is used to acquire real-time image data of a target flexible substrate in a target position at each detection moment of a current detection cycle based on a preset image acquisition device to obtain an initial image data set; Image processing unit: used for performing image fuzzy processing on the initial image data set to obtain a first fuzzy image, and performing grayscale transformation on the first fuzzy image after fuzzy processing to obtain a first processed image; Image synthesis unit: used for sorting the first processed images according to the corresponding detection moments to obtain an ordered set of first processed images.

[0009] The identification processing module provided by the present invention includes: An initial model unit: used for determining an initial feature extraction model of a target flexible substrate based on a substrate type of the target flexible substrate; Historical model unit: used to obtain the historical feature image of the target flexible substrate and the historical feature extraction model corresponding to each historical feature image, and determine the corresponding historical feature parameters in combination with the initial feature parameters contained in the initial feature extraction model, so as to comprehensively obtain the historical feature extraction set; A first model unit: used for performing parameter optimization based on the historical feature parameters of each historical feature extraction model in the historical feature extraction set, so as to obtain a first feature extraction model of the target flexible substrate; Feature extraction unit: used for performing feature extraction on each first processed image in the first processed image set based on the first feature extraction model to obtain a first feature image set.

[0010] The first model unit provided according to the present invention comprises: A feature parameter extraction subunit: used to extract the historical feature parameters of each historical feature extraction model in the historical feature extraction set to obtain a first parameter set; Historical performance subunit: used to obtain historical substrate performance of the target flexible substrate corresponding to each historical feature extraction model in the historical feature extraction set; A first similarity subunit: used for acquiring the real-time substrate performance of the target flexible substrate, and comparing the real-time substrate performance with the historical substrate performance one by one, to obtain a first similarity between the real-time substrate performance and each historical substrate performance; A first performance subunit: used for obtaining a first similarity set of the target flexible substrate based on each first similarity, and matching each first similarity in the first similarity set with a historical characteristic parameter in the first parameter set, thereby comprehensively obtaining a first performance parameter set; A first reference subunit: used to extract the historical characteristic parameter with the highest first similarity in the first performance parameter set as the first reference characteristic parameter of the target flexible substrate; Environmental parameter subunit: used to obtain the real-time environmental parameters of the target flexible substrate, and simultaneously obtain the historical environmental parameters of the target flexible substrate to obtain a historical environmental parameter set; The first historical subunit is used to extract the historical environmental parameter with the highest parameter consistency with the real-time environmental parameter from the historical environmental parameter set as the first historical environmental parameter; A second reference subunit: used to obtain a historical characteristic parameter corresponding to the first historical environmental parameter as a second reference characteristic parameter of the target flexible substrate; Parameter optimization subunit: used to optimize the corresponding feature parameters of the initial feature extraction model based on the first reference feature parameters and the second reference feature parameters to obtain the first optimized feature parameters, thereby obtaining the first feature extraction model of the target flexible substrate.

[0011] The first similar subunit provided by the present invention comprises: Obtaining a first similarity between the real-time substrate performance and each historical substrate performance; ; Where S is the first similarity between the real-time substrate performance and the current historical substrate performance, is the performance conversion factor of the jth performance index of the current historical substrate performance, is the jth performance index of the real-time substrate performance, is the jth performance index of the current historical substrate performance, is the extracted transmission error of the jth performance index of the current historical substrate performance, is the error conversion factor of the jth performance index of the current historical substrate performance, e is the base of the exponential function, and m is the number of performance indexes in the real-time substrate performance, wherein the number of performance indexes in the real-time substrate performance is consistent with the number of performance indexes in the historical substrate performance.

[0012] The defect detection module provided by the present invention includes: A first classification unit: used for classifying each feature image in the first feature image set based on different feature types to obtain a first classified feature image set; Wherein, each first classification feature image subset in the first classification feature image set corresponds to the same feature type; A first detection unit: used for acquiring a corresponding defect detection model based on a feature type, and inputting a feature image in a first classification feature image subset into a corresponding defect detection model for defect detection to obtain a first detection result; The comprehensive detection unit is used to obtain a comprehensive defect detection result of the target flexible substrate based on each first detection result of the target flexible substrate.

[0013] The adaptive adjustment module provided by the present invention includes: Defect synthesis unit: used to obtain the historical defect detection results of the target flexible substrate in the historical detection process, and combine the real-time defect detection results to obtain the comprehensive defect detection results; Defect classification unit: used for classifying the comprehensive defect detection results based on different defect types, so as to obtain a classified defect detection set, wherein each classified defect detection subset of the classified defect detection set corresponds to the same defect type; An adaptive adjustment unit is used to adaptively adjust the target flexible substrate based on each classified defect detection subset.

[0014] The adaptive adjustment unit provided according to the present invention includes: A first strategy subunit: used for acquiring a corresponding first adaptive adjustment strategy from a defect-adjustment database based on a defect type corresponding to each classified defect detection subset; The second strategy subunit is used to obtain an initial adaptive adjustment strategy based on each first adaptive adjustment strategy; Comparison and optimization subunit: used to compare whether there are strategy conflicts and strategy duplication adjustment strategies in the initial adaptive adjustment strategy, so as to optimize the strategy and obtain the comprehensive adaptive adjustment strategy of the target flexible substrate, so as to adaptively adjust the defect detection model of the target flexible substrate.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a flexible substrate adaptive defect detection system based on machine vision, which processes the image data of the target flexible substrate to extract feature images, and classifies the feature images based on the image types, thereby performing defect detection on the classified feature images, which can make the defect detection of the target flexible substrate more accurate and better meet the defect detection requirements of the target flexible substrate, and can provide adaptive adjustment strategies in a timely and accurate manner, thereby improving the defect detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the 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 creative work.

[0017] Figure 1It is a structural diagram of a flexible substrate adaptive defect detection system based on machine vision provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Embodiment 1: The embodiment of the present invention provides a flexible substrate adaptive defect detection system based on machine vision, such as Figure 1 As shown, including: An image acquisition module is used to acquire real-time image data of a target flexible substrate in a target position based on a preset image acquisition device, and perform image processing to obtain a first processed image set; Recognition processing module: used for extracting features from each first processed image in the first processed image set based on deep learning technology to obtain a first feature image set; Defect detection module: used for classifying each feature in the first feature image set and performing defect detection based on the classification result; Adaptive adjustment module: used to adaptively adjust the target flexible substrate based on the defect detection results.

[0020] In this embodiment, the preset image acquisition device includes a video camera, a camera or other devices capable of acquiring images, which are usually configured to capture high-quality images under specific conditions.

[0021] In this embodiment, the target position refers to the relative position of a portion of the target flexible substrate that needs to be inspected for defects.

[0022] In this embodiment, the real-time image data refers to image information captured by a preset image acquisition device at a specific time point or over a continuous period of time.

[0023] In this embodiment, image processing is a process of analyzing, converting or operating image data to extract useful information or improve the visual effect of the image, for example, including steps such as image enhancement, filtering, and edge detection.

[0024] In this embodiment, the first processed image set is an image set obtained after preliminary image processing, and these images have been transformed or processed in some way to facilitate subsequent feature extraction and analysis.

[0025] In this embodiment, deep learning technology is a machine learning method that simulates the structure and function of the human brain neural network to achieve complex data analysis and pattern recognition. In the field of image processing, deep learning technology is often used for tasks such as feature extraction and image classification.

[0026] In this embodiment, the first feature image set is an image set composed of feature images extracted from the first processed image set using deep learning technology.

[0027] In this embodiment, feature classification is a process of classifying the extracted feature images according to certain rules or standards.

[0028] In this embodiment, defect detection is a detection process of identifying whether there are defects or abnormalities in the flexible substrate based on the extracted features and classification results.

[0029] In this embodiment, the adaptive adjustment refers to a process of making necessary adjustments or repairs to the target flexible substrate according to the defect detection result and a process of adjusting and optimizing the defect detection model of the target flexible substrate.

[0030] The beneficial effect of the above technical solution is: by processing the image data of the target flexible substrate, the feature image is extracted, and the feature image is classified based on the image type, so as to perform defect detection on the classified feature image, so that the defect detection of the target flexible substrate can be more accurate and better able to meet the defect detection needs of the target flexible substrate, and adaptive adjustment strategies can be provided in a timely and accurate manner, thereby improving the defect detection efficiency.

[0031] Embodiment 2: Based on the first embodiment, the image acquisition module includes: An image acquisition unit is used to acquire real-time image data of a target flexible substrate in a target position at each detection moment of a current detection cycle based on a preset image acquisition device to obtain an initial image data set; Image processing unit: used for performing image fuzzy processing on the initial image data set to obtain a first fuzzy image, and performing grayscale transformation on the first fuzzy image after fuzzy processing to obtain a first processed image; Image synthesis unit: used for sorting the first processed images according to the corresponding detection moments to obtain an ordered set of first processed images.

[0032] In this embodiment, the preset image acquisition device includes a video camera, a camera or other devices capable of acquiring images, which are usually configured to capture high-quality images under specific conditions.

[0033] In this embodiment, the target position refers to the relative position of a portion of the target flexible substrate that needs to be inspected for defects.

[0034] In this embodiment, the current detection cycle refers to a whole detection cycle for performing defect detection on the target flexible substrate.

[0035] In this embodiment, the real-time image data refers to image information captured by a preset image acquisition device at a specific time point or over a continuous period of time.

[0036] In this embodiment, the initial image data set refers to an image data set obtained by integrating the real-time image data at each detection moment in the current detection cycle.

[0037] In this embodiment, image blur processing and grayscale transformation are both performed on image data, thereby improving image detection efficiency and facilitating the feature extraction process.

[0038] In this embodiment, the first processed image set is an image set obtained after preliminary image processing. These images have been fuzzy processed and grayscale transformed to facilitate subsequent feature extraction and analysis.

[0039] The beneficial effect of the above technical solution is that by processing the real-time image data of the target flexible substrate, the processed image can be made more accurate, the feature extraction efficiency can be improved, and thus the defect detection of the target flexible substrate can be made more accurate.

[0040] Embodiment 3: Based on the second embodiment, the identification processing module includes: An initial model unit: used for determining an initial feature extraction model of a target flexible substrate based on a substrate type of the target flexible substrate; Historical model unit: used to obtain the historical feature image of the target flexible substrate and the historical feature extraction model corresponding to each historical feature image, and determine the corresponding historical feature parameters in combination with the initial feature parameters contained in the initial feature extraction model, so as to comprehensively obtain the historical feature extraction set; A first model unit: used for performing parameter optimization based on the historical feature parameters of each historical feature extraction model in the historical feature extraction set, so as to obtain a first feature extraction model of the target flexible substrate; Feature extraction unit: used for performing feature extraction on each first processed image in the first processed image set based on the first feature extraction model to obtain a first feature image set.

[0041] In this embodiment, the substrate type is determined comprehensively according to the substrate material, the number of circuit layers, and the bonding method between the substrate and the copper foil.

[0042] In this embodiment, the initial feature extraction model is a feature extraction model determined based on the substrate type of the target flexible substrate and consistent with the current substrate type.

[0043] In this embodiment, the historical characteristic image refers to a characteristic image of the target flexible substrate during a historical detection process.

[0044] In this embodiment, the historical feature extraction model is a feature extraction model corresponding to the target flexible substrate during the historical defect detection process.

[0045] In this embodiment, the historical feature parameters refer to the feature parameters included in each historical feature extraction model.

[0046] In this embodiment, the historical feature extraction set refers to a set including each historical feature parameter and a corresponding historical feature extraction model.

[0047] In this embodiment, the first feature extraction model refers to a feature model obtained by optimizing the corresponding feature parameters of the initial feature extraction model based on the historical feature parameters in the historical feature extraction set.

[0048] In this embodiment, the first feature image set refers to a set of feature images obtained by performing feature extraction on each first processed image in the first processed image set based on the first feature extraction model.

[0049] The beneficial effect of the above technical solution is: by determining the feature extraction model based on the substrate type and optimizing the model, the feature extraction model of the target flexible substrate can better meet the real-time detection requirements, improve the feature extraction efficiency, and thus make the defect detection of the target flexible substrate more accurate.

[0050] Embodiment 4: Based on Example 3, the first model unit includes: A feature parameter extraction subunit: used to extract the historical feature parameters of each historical feature extraction model in the historical feature extraction set to obtain a first parameter set; Historical performance subunit: used to obtain historical substrate performance of the target flexible substrate corresponding to each historical feature extraction model in the historical feature extraction set; A first similarity subunit: used for acquiring the real-time substrate performance of the target flexible substrate, and comparing the real-time substrate performance with the historical substrate performance one by one, to obtain a first similarity between the real-time substrate performance and each historical substrate performance; A first performance subunit: used for obtaining a first similarity set of the target flexible substrate based on each first similarity, and matching each first similarity in the first similarity set with a historical characteristic parameter in the first parameter set, thereby comprehensively obtaining a first performance parameter set; A first reference subunit: used to extract the historical characteristic parameter with the highest first similarity in the first performance parameter set as the first reference characteristic parameter of the target flexible substrate; Environmental parameter subunit: used to obtain the real-time environmental parameters of the target flexible substrate, and simultaneously obtain the historical environmental parameters of the target flexible substrate to obtain a historical environmental parameter set; The first historical subunit is used to extract the historical environmental parameter with the highest parameter consistency with the real-time environmental parameter from the historical environmental parameter set as the first historical environmental parameter; A second reference subunit: used to obtain a historical characteristic parameter corresponding to the first historical environmental parameter as a second reference characteristic parameter of the target flexible substrate; Parameter optimization subunit: used to optimize the corresponding feature parameters of the initial feature extraction model based on the first reference feature parameters and the second reference feature parameters to obtain the first optimized feature parameters, thereby obtaining the first feature extraction model of the target flexible substrate.

[0051] In this embodiment, the historical substrate performance refers to the historical performance data of the target flexible substrate corresponding to the historical feature extraction model, and the historical substrate performance refers to data on the electrical performance, mechanical performance, thermal performance, etc. of the flexible substrate.

[0052] In this embodiment, the real-time substrate performance refers to the performance data of the target flexible substrate at the current moment.

[0053] In this embodiment, the first similarity refers to the similarity between the real-time substrate performance and each historical substrate performance.

[0054] In this embodiment, the first similarity set includes a set of first similarities between the real-time substrate performance and each historical substrate performance.

[0055] In this embodiment, the first performance parameter set is a set obtained by matching each first similarity in the first similarity set with a historical feature parameter in the first parameter set.

[0056] In this embodiment, the first reference characteristic parameter is a historical characteristic parameter with the highest first similarity in the first performance parameter set.

[0057] In this embodiment, the real-time environmental parameters refer to parameters of the target flexible substrate in the current environment, such as temperature, humidity, pressure, etc.

[0058] In this embodiment, the historical environmental parameter set includes a set of environmental parameters of the target flexible substrate under different historical working environments.

[0059] In this embodiment, the first historical environment parameter refers to a historical environment parameter in the historical environment parameter set that has the highest degree of consistency with the real-time environment parameter.

[0060] In this embodiment, the second reference characteristic parameter is a historical characteristic parameter corresponding to the first historical environment parameter.

[0061] In this embodiment, the first optimized feature parameter is a feature parameter obtained by optimizing the corresponding feature parameter of the initial feature extraction model.

[0062] The beneficial effect of the above technical solution is: by determining the feature extraction model based on the substrate type and optimizing the model in combination with environmental parameters and historical performance parameters, the feature extraction model of the target flexible substrate can better meet the real-time detection requirements, thereby making the defect detection of the target flexible substrate more accurate.

[0063] Embodiment 5: Based on Example 4, the first similar subunit includes: Obtaining a first similarity between the real-time substrate performance and each historical substrate performance; ; Where S is the first similarity between the real-time substrate performance and the current historical substrate performance, is the performance conversion factor of the jth performance index of the current historical substrate performance, is the jth performance index of the real-time substrate performance, is the jth performance index of the current historical substrate performance, is the extracted transmission error of the jth performance index of the current historical substrate performance, is the error conversion factor of the jth performance index of the current historical substrate performance, e is the base of the exponential function, and m is the number of performance indexes in the real-time substrate performance, wherein the number of performance indexes in the real-time substrate performance is consistent with the number of performance indexes in the historical substrate performance.

[0064] The beneficial effect of the above technical solution is: by determining the similarity between the real-time substrate performance and the historical substrate performance, and then combining the historical environmental parameters for model optimization, the feature extraction model of the target flexible substrate can better meet the real-time detection requirements, thereby making the defect detection of the target flexible substrate more accurate.

[0065] Embodiment 6: Based on the third embodiment, the defect detection module includes: A first classification unit: used for classifying each feature image in the first feature image set based on different feature types to obtain a first classified feature image set; Wherein, each first classification feature image subset in the first classification feature image set corresponds to the same feature type; A first detection unit: used for acquiring a corresponding defect detection model based on a feature type, and inputting a feature image in a first classification feature image subset into a corresponding defect detection model for defect detection to obtain a first detection result; The comprehensive detection unit is used to obtain a comprehensive defect detection result of the target flexible substrate based on each first detection result of the target flexible substrate.

[0066] In this embodiment, feature classification is a process of classifying the extracted feature images according to certain rules or standards.

[0067] In this embodiment, each first classification feature image subset in the first classification feature image set corresponds to the same feature type.

[0068] In this embodiment, the defect detection model refers to an algorithm or system specifically used to detect various defects on the flexible substrate, which is usually based on image processing, machine learning or deep learning technologies, and can automatically identify and locate defects on the flexible substrate.

[0069] In this embodiment, the first detection result is obtained by inputting the feature images in the first classification feature image subset into the corresponding defect detection model to perform defect detection.

[0070] In this embodiment, the comprehensive defect detection is a detection process of identifying whether there are defects or abnormalities in the flexible substrate based on the extracted features and classification results.

[0071] The beneficial effect of the above technical solution is that by classifying the image type based on the feature image, and then performing defect detection on the classified feature image, the defect detection of the target flexible substrate can be made more accurate, thereby improving the defect detection efficiency.

[0072] Embodiment 7: Based on Example 6, the adaptive adjustment module includes: Defect synthesis unit: used to obtain the historical defect detection results of the target flexible substrate in the historical detection process, and combine the real-time defect detection results to obtain the comprehensive defect detection results; Defect classification unit: used for classifying the comprehensive defect detection results based on different defect types, so as to obtain a classified defect detection set, wherein each classified defect detection subset of the classified defect detection set corresponds to the same defect type; An adaptive adjustment unit is used to adaptively adjust the target flexible substrate based on each classified defect detection subset.

[0073] In this embodiment, the comprehensive defect detection result is obtained based on the historical defect detection result and the real-time defect detection result of the target flexible substrate.

[0074] In this embodiment, the classified defect detection set is formed based on the classification results after classifying the comprehensive defect detection results based on different defect types, wherein the defect types include scratch defects, surface unevenness defects, color defects, and line unevenness defects, wherein the defect classification dimensions are different and the defect types are also different.

[0075] In this embodiment, the adaptive adjustment refers to a process of making necessary adjustments or repairs to the target flexible substrate according to the defect detection result and a process of adjusting and optimizing the defect detection model of the target flexible substrate.

[0076] The beneficial effect of the above technical solution is that by adaptively adjusting the target flexible substrate, defect detection of the target flexible substrate can be made more accurate, and an adaptive adjustment strategy can be provided timely and accurately, thereby improving defect detection efficiency.

[0077] Embodiment 8: Based on Embodiment 7, the adaptive adjustment unit includes: A first strategy subunit: used for acquiring a corresponding first adaptive adjustment strategy from a defect-adjustment database based on a defect type corresponding to each classified defect detection subset; The second strategy subunit is used to obtain an initial adaptive adjustment strategy based on each first adaptive adjustment strategy; Comparison and optimization subunit: used to compare whether there are strategy conflicts and strategy duplication adjustment strategies in the initial adaptive adjustment strategy, so as to optimize the strategy and obtain the comprehensive adaptive adjustment strategy of the target flexible substrate, so as to adaptively adjust the defect detection model of the target flexible substrate.

[0078] In this embodiment, the first adaptive adjustment strategy is to obtain a corresponding defect adjustment strategy from a defect-adjustment database based on the defect type corresponding to each classified defect detection subset.

[0079] In this embodiment, the initial adaptive adjustment strategy is obtained by integrating each first adaptive adjustment strategy of the target flexible substrate.

[0080] In this embodiment, the comprehensive adaptive adjustment strategy deletes the policy conflicts and policy duplications existing in the initial adaptive adjustment strategy, thereby determining.

[0081] In this embodiment, the adaptive adjustment refers to a process of making necessary adjustments or repairs to the target flexible substrate according to the defect detection result and a process of adjusting and optimizing the defect detection model of the target flexible substrate.

[0082] The beneficial effect of the above technical solution is that by adaptively adjusting the target flexible substrate, defect detection of the target flexible substrate can be made more accurate, and an adaptive adjustment strategy can be provided timely and accurately, thereby improving defect detection efficiency.

[0083] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A flexible substrate adaptive defect detection system based on machine vision, characterized in that: include: An image acquisition module is used to acquire real-time image data of a target flexible substrate in a target position based on a preset image acquisition device, and perform image processing to obtain a first processed image set; Recognition processing module: used for extracting features from each first processed image in the first processed image set based on deep learning technology to obtain a first feature image set; Defect detection module: used for classifying each feature in the first feature image set and performing defect detection based on the classification result; Adaptive adjustment module: used to adaptively adjust the target flexible substrate based on the defect detection results.

2. The flexible substrate adaptive defect detection system based on machine vision according to claim 1, characterized in that: Image acquisition module, including: An image acquisition unit is used to acquire real-time image data of a target flexible substrate in a target position at each detection moment of a current detection cycle based on a preset image acquisition device to obtain an initial image data set; Image processing unit: used for performing image fuzzy processing on the initial image data set to obtain a first fuzzy image, and performing grayscale transformation on the first fuzzy image after fuzzy processing to obtain a first processed image; Image synthesis unit: used for sorting the first processed images according to the corresponding detection moments to obtain an ordered set of first processed images.

3. The flexible substrate adaptive defect detection system based on machine vision according to claim 2, characterized in that: Identification processing module, including: An initial model unit: used for determining an initial feature extraction model of a target flexible substrate based on a substrate type of the target flexible substrate; Historical model unit: used to obtain the historical feature image of the target flexible substrate and the historical feature extraction model corresponding to each historical feature image, and determine the corresponding historical feature parameters in combination with the initial feature parameters contained in the initial feature extraction model, so as to comprehensively obtain the historical feature extraction set; A first model unit: used for performing parameter optimization based on the historical feature parameters of each historical feature extraction model in the historical feature extraction set, so as to obtain a first feature extraction model of the target flexible substrate; Feature extraction unit: used for performing feature extraction on each first processed image in the first processed image set based on the first feature extraction model to obtain a first feature image set.

4. The flexible substrate adaptive defect detection system based on machine vision according to claim 3 is characterized in that: The first model unit includes: A feature parameter extraction subunit: used to extract the historical feature parameters of each historical feature extraction model in the historical feature extraction set to obtain a first parameter set; Historical performance subunit: used to obtain historical substrate performance of the target flexible substrate corresponding to each historical feature extraction model in the historical feature extraction set; A first similarity subunit: used for acquiring the real-time substrate performance of the target flexible substrate, and comparing the real-time substrate performance with the historical substrate performance one by one, to obtain a first similarity between the real-time substrate performance and each historical substrate performance; A first performance subunit: used for obtaining a first similarity set of the target flexible substrate based on each first similarity, and matching each first similarity in the first similarity set with a historical characteristic parameter in the first parameter set, thereby comprehensively obtaining a first performance parameter set; A first reference subunit: used to extract the historical characteristic parameter with the highest first similarity in the first performance parameter set as the first reference characteristic parameter of the target flexible substrate; Environmental parameter subunit: used to obtain the real-time environmental parameters of the target flexible substrate, and simultaneously obtain the historical environmental parameters of the target flexible substrate to obtain a historical environmental parameter set; The first historical subunit is used to extract the historical environmental parameter with the highest parameter consistency with the real-time environmental parameter from the historical environmental parameter set as the first historical environmental parameter; A second reference subunit: used to obtain a historical characteristic parameter corresponding to the first historical environmental parameter as a second reference characteristic parameter of the target flexible substrate; Parameter optimization subunit: used to optimize the corresponding feature parameters of the initial feature extraction model based on the first reference feature parameters and the second reference feature parameters to obtain the first optimized feature parameters, thereby obtaining the first feature extraction model of the target flexible substrate.

5. The flexible substrate adaptive defect detection system based on machine vision according to claim 4, characterized in that: The first similar subunit comprises: Obtaining a first similarity between the real-time substrate performance and each historical substrate performance; ; Where S is the first similarity between the real-time substrate performance and the current historical substrate performance, is the performance conversion factor of the jth performance index of the current historical substrate performance, is the jth performance index of the real-time substrate performance, is the jth performance index of the current historical substrate performance, is the extracted transmission error of the jth performance index of the current historical substrate performance, is the error conversion factor of the jth performance index of the current historical substrate performance, e is the base of the exponential function, and m is the number of performance indexes in the real-time substrate performance, wherein the number of performance indexes in the real-time substrate performance is consistent with the number of performance indexes in the historical substrate performance.

6. The flexible substrate adaptive defect detection system based on machine vision according to claim 3, characterized in that: Defect detection module, including: A first classification unit: used for classifying each feature image in the first feature image set based on different feature types to obtain a first classified feature image set; Wherein, each first classification feature image subset in the first classification feature image set corresponds to the same feature type; A first detection unit: used for acquiring a corresponding defect detection model based on a feature type, and inputting a feature image in a first classification feature image subset into a corresponding defect detection model for defect detection to obtain a first detection result; The comprehensive detection unit is used to obtain a comprehensive defect detection result of the target flexible substrate based on each first detection result of the target flexible substrate.

7. The flexible substrate adaptive defect detection system based on machine vision according to claim 6, characterized in that: Adaptive adjustment module, including: Defect synthesis unit: used to obtain the historical defect detection results of the target flexible substrate in the historical detection process, and combine the real-time defect detection results to obtain the comprehensive defect detection results; Defect classification unit: used for classifying the comprehensive defect detection results based on different defect types, so as to obtain a classified defect detection set, wherein each classified defect detection subset of the classified defect detection set corresponds to the same defect type; An adaptive adjustment unit is used to adaptively adjust the target flexible substrate based on each classified defect detection subset.

8. The flexible substrate adaptive defect detection system based on machine vision according to claim 7, characterized in that: Adaptive adjustment unit, comprising: A first strategy subunit: used for acquiring a corresponding first adaptive adjustment strategy from a defect-adjustment database based on a defect type corresponding to each classified defect detection subset; The second strategy subunit is used to obtain an initial adaptive adjustment strategy based on each first adaptive adjustment strategy; Comparison and optimization subunit: used to compare whether there are strategy conflicts and strategy duplication adjustment strategies in the initial adaptive adjustment strategy, so as to optimize the strategy and obtain the comprehensive adaptive adjustment strategy of the target flexible substrate, so as to adaptively adjust the defect detection model of the target flexible substrate.

Images