AVI automatic image classification method

Abstract

The invention discloses an AVI automatic image classification method, and relates to the field of image recognition, and the method comprises the steps: inputting data annotation presetting which is a setting parameter for a classified image; processing the color image by adopting a training model, and extracting the coordinate, category and confidence score of a detection frame; a detection result meeting conditions is output and generated after model training; each result in the output detection results comprises coordinates of a bounding box, a category label and a confidence score, and the final result with the highest score is the defect position and category judged by the algorithm. According to the invention, initial inspection and reinspection parameters can be revised and managed in real time, AVF automatic determination is realized, and the operation is convenient.

Description

Technical Field

[0001] The invention relates to the field of image recognition, and in particular to an AVI automatic image classification method. Background Art

[0002] AVI automatic image classification method is an automatic determination software for COF defect images. It has some functional features such as initial inspection, re-inspection, and random inspection. It can revise and manage the initial inspection and re-inspection parameters in real time to realize AVF automatic determination. It is a rare determination software with easy operation.

[0003] However, the current AVI automatic image classification method cannot achieve defect identification for confidence setting and channel grayscale setting;

[0004] Therefore, an automatic AVI image classification method is urgently needed. Summary of the invention

[0005] The present invention provides an AVI automatic image classification method to solve the problems of the prior art, including:

[0006] Input data annotation preset, which is the setting parameters for the classification image;

[0007] Process the color image using the trained model to extract the coordinates, category, and confidence score of the detection box;

[0008] After the model is trained, the output generates detection results that meet the conditions;

[0009] Each of the output detection results includes the coordinates of a bounding box, a category label, and a confidence score. The result with the highest score is the defect location and category determined by the algorithm.

[0010] Among them, the detection frames are screened, highly overlapping frames are removed, and the frames with the highest confidence are retained.

[0011] Furthermore, it includes: before the training model is connected to the color image, the color image is first processed by a normalization operation to obtain processed image data.

[0012] Furthermore, the image data is processed by a backbone network, a neck network and a head network.

[0013] Furthermore, the image data is forward propagated in the neural network of the training model.

[0014] Furthermore, the screening of the detection frames, removing highly overlapping frames, and retaining the frames with the highest confidence, specifically includes: processing and extracting the coordinates, category and confidence score of each detection frame, and using a non-maximum suppression algorithm to screen the detection frames to obtain corresponding detection frames.

[0015] Furthermore, the image data is first judged in the neural network of the training model whether there are bad frames, specifically based on whether the non-maximum suppression algorithm score meets the OK&NG threshold. If the algorithm score is greater than the NG threshold, it is classified according to whether the OK threshold and the NG threshold are met or not:

[0016] NG classification, when the judgment algorithm score is greater than the NG threshold, it is classified as NG;

[0017] OK classification, when the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as OK classification;

[0018] If the algorithm score is less than the OK threshold, it is determined that there is no defective frame and is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again.

[0019] Furthermore, each of the output detection results includes the coordinates of a bounding box, a category label and a confidence score, and the result with the highest final score is the defect location and category determined by the algorithm, specifically including;

[0020] In the OK classification, according to the defect location and type, the results are analyzed, processed and judged based on the data annotation presets as follows:

[0021] s1. If the sum of the grayscale of the three channels of RGB in the center area of ​​the analysis is > 600, it is classified as NG. After analysis, the result is determined to be intercepted metal & shiny defects. Otherwise, execute s2.

[0022] s2. If after analyzing the target frame, the grayscale is intercepted according to the process inspection specification and the grayscale is ≤60, the target frame is judged to be NG classification, otherwise, s3 is executed;

[0023] s3. If the grayscale G of the center area of ​​1 / 4 of the target frame is ≤60, continue to judge the OK threshold score of the line classification algorithm. If yes, it is NG class, otherwise it is OK class, and intercept the picture that is not out of the frame in the center area.

[0024] Furthermore, if the algorithm score is less than the OK threshold, it is determined that there is no bad frame, and the frame is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again, specifically including:

[0025] Rechecking again based on the algorithm score of the foreign body classification model, the algorithm score is a non-maximum suppression algorithm score;

[0026] When the algorithm score is again judged to be greater than the NG threshold, it is classified as NG;

[0027] When the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as the OK classification;

[0028] For the OK classification, the detection area is expanded to determine the gray scale of foreign matter. If the RGB of the entire image is > 700, it is classified as NG, and the sampling holes and NG hole defects are intercepted. If the RGB of the entire image is < 700, it is classified as OK.

[0029] The invention provides an AVI automatic image classification method, which can revise and manage the initial inspection and re-inspection parameters in real time and realize automatic AVF determination. The method is a rare determination software and is easy to operate. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0031] Figure 1 A flow chart of an AVI automatic image classification method provided by an exemplary embodiment of the present invention. DETAILED DESCRIPTION

[0032] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0033] The present invention provides an AVI automatic image classification method, aiming to solve the above technical problems in the prior art.

[0034] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present invention will be described below in conjunction with the accompanying drawings.

[0035] Embodiment 1: The present invention provides an exemplary embodiment of an AVI automatic image classification method, such as Figure 1 As shown, including:

[0036] a1. Input data annotation preset, which is the setting parameters for the classification image;

[0037] a2. Process the color image using the training model to extract the coordinates, category, and confidence score of the detection box;

[0038] The method includes: before the training model is connected to the color image, the color image is first processed by a normalization operation to obtain processed image data. The image data is processed by a backbone network, a neck network and a head network. The image data is forward propagated in the neural network of the training model.

[0039] a3. After the model is trained, the output generates a test result that meets the conditions;

[0040] a4. Each of the output detection results includes the coordinates of a bounding box, a category label, and a confidence score. The result with the highest score is the defect location and category determined by the algorithm.

[0041] Among them, the detection frames are screened, highly overlapping frames are removed, and the frames with the highest confidence are retained. The screening of the detection frames, removing highly overlapping frames, and retaining the frames with the highest confidence specifically includes: processing and extracting the coordinates, category and confidence score of each detection frame, and using a non-maximum suppression algorithm to screen the detection frames to obtain corresponding detection frames.

[0042] Preferably, the image data is first judged in the neural network of the training model whether there is a bad frame, specifically based on whether the non-maximum suppression algorithm score meets the OK&NG threshold. If the algorithm score is greater than the NG threshold, it is classified according to whether the OK threshold and the NG threshold are met or not:

[0043] NG classification, when the judgment algorithm score is greater than the NG threshold, it is classified as NG;

[0044] OK classification, when the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as OK classification;

[0045] If the algorithm score is less than the OK threshold, it is determined that there is no defective frame and is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again.

[0046] Preferably, each of the output detection results includes the coordinates of a bounding box, a category label and a confidence score, and the result with the highest final score is the defect location and category determined by the algorithm, specifically including:

[0047] In the OK classification, according to the defect location and type, the results are analyzed, processed and judged based on the data annotation presets as follows:

[0048] s1. If the sum of the grayscale of the three channels of RGB in the center area of ​​the analysis is > 600, it is classified as NG. After analysis, the result is determined to be intercepted metal & shiny defects. Otherwise, execute s2.

[0049] s2. If after analyzing the target frame, the grayscale is intercepted according to the process inspection specification and the grayscale is ≤60, the target frame is judged to be NG classification, otherwise, s3 is executed;

[0050] s3. If the grayscale G of the center area of ​​1 / 4 of the target frame is ≤60, continue to judge the OK threshold score of the line classification algorithm. If yes, it is NG class, otherwise it is OK class, and intercept the picture that is not out of the frame in the center area.

[0051] Preferably, if the algorithm score is less than the OK threshold, it is determined that there is no bad frame, and the frame is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again, specifically including:

[0052] Rechecking again based on the algorithm score of the foreign body classification model, the algorithm score is a non-maximum suppression algorithm score;

[0053] When the algorithm score is again judged to be greater than the NG threshold, it is classified as NG;

[0054] When the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as the OK classification;

[0055] For the OK classification, the detection area is expanded to determine the gray scale of foreign matter. If the RGB of the entire image is > 700, it is classified as NG, and the sampling holes and NG hole defects are intercepted. If the RGB of the entire image is < 700, it is classified as OK.

[0056] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0057] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0058] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0059] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

[0060] Those skilled in the art will appreciate that the embodiments of the present invention may be provided as methods or systems. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects.

[0061] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0062] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.

[0063] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. The description and examples are to be considered exemplary only, and the true scope and spirit of the present invention is indicated by the claims above.

[0064] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. An AVI automatic image classification method, characterized in that: include: Input data annotation preset, which is the setting parameters for the classification image; Process the color image using the trained model to extract the coordinates, category, and confidence score of the detection box; After the model is trained, the output generates detection results that meet the conditions; Each of the output detection results includes the coordinates of a bounding box, a category label, and a confidence score. The result with the highest score is the defect location and category determined by the algorithm. Among them, the detection frames are screened, highly overlapping frames are removed, and the frames with the highest confidence are retained.

2. A method for automatic AVI image classification according to claim 1, characterized in that: include: Before the training model is connected to the color image, the color image is first processed by a normalization operation to obtain the processed image data.

3. A method for automatic AVI image classification according to claim 2, characterized in that: The image data is processed by a backbone network, a neck network and a head network.

4. A method for automatic AVI image classification according to claim 3, characterized in that: The image data is forward propagated in the neural network of the training model.

5. A method for automatic AVI image classification according to claim 4, characterized in that: The screening of the detection frames, removing highly overlapping frames, and retaining the frames with the highest confidence, specifically includes: processing and extracting the coordinates, category, and confidence score of each detection frame, and using a non-maximum suppression algorithm to screen the detection frames to obtain corresponding detection frames.

6. A method for automatic AVI image classification according to claim 5, characterized in that: In the neural network of the training model, the image data is first judged whether there are bad frames, specifically based on whether the non-maximum suppression algorithm score meets the OK&NG threshold. If the algorithm score is greater than the NG threshold, it is classified according to whether the OK threshold and the NG threshold are met or not: NG classification, when the judgment algorithm score is greater than the NG threshold, it is classified as NG; OK classification, when the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as OK classification; If the algorithm score is less than the OK threshold, it is determined that there is no defective frame and is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again.

7. A method for automatic AVI image classification according to claim 6, characterized in that: Each of the output detection results includes the coordinates of a bounding box, a category label and a confidence score. The result with the highest final score is the defect location and category determined by the algorithm, specifically including: In the OK classification, according to the defect location and type, the results are analyzed, processed and judged based on the data annotation presets as follows: s1. If the sum of the grayscale of the three channels of RGB in the center area of ​​the analysis is > 600, it is classified as NG. After analysis, the result is determined to be intercepted metal & shiny defects. Otherwise, execute s2. s2. If after analyzing the target frame, the grayscale is intercepted according to the process inspection specification and the grayscale is ≤60, the target frame is judged to be NG classification, otherwise, s3 is executed; s3. If the grayscale G of the center area of ​​1 / 4 of the target frame is ≤60, continue to judge the OK threshold score of the line classification algorithm. If yes, it is NG class, otherwise it is OK class, and intercept the picture that is not out of the frame in the center area.

8. A method for automatic AVI image classification according to claim 7, characterized in that: If the algorithm score is less than the OK threshold, it is determined that there is no bad frame and is output to the foreign body classification model. The foreign body classification model is used to determine the foreign body and perform OK classification and NG classification again, specifically including: Rechecking again based on the algorithm score of the foreign body classification model, the algorithm score is a non-maximum suppression algorithm score; When the algorithm score is again judged to be greater than the NG threshold, it is classified as NG; When the judgment algorithm score is less than the NG threshold and greater than the OK threshold, it is classified as the OK classification; For the OK classification, the detection area is expanded to determine the gray scale of foreign matter. If the RGB of the entire image is > 700, it is classified as NG, and the sampling holes and NG hole defects are intercepted. If the RGB of the entire image is < 700, it is classified as OK.

Images