An image processing and analysis system and method using AI and visual detection technology

Abstract

The present invention discloses an image processing and analysis system and method using AI and visual detection technology, belonging to the field of image processing technology. The system includes a data acquisition module, an intelligent detection module, an operation adjustment module and a visualization module; the data acquisition module is used to collect image detection records and equipment operation parameters of each assembly line; the intelligent detection module analyzes the relationship formula according to the image detection record and calculates the theoretical duration and ideal success rate, calculates the deviation coefficient of the assembly line together with the actual duration and success rate, and marks the abnormal assembly line according to the deviation coefficient; the operation adjustment module analyzes the detection algorithm under the abnormal assembly line and calculates the best quality score, obtains the relationship formula between light intensity and resolution according to the best quality score analysis, and adjusts the light intensity of the abnormal assembly line according to the relationship formula until the deviation coefficient is normal; the visualization module is used to display the equipment operation parameters of each assembly line, generate shooting records and processing records and store them as image detection records.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an image processing and analysis system and method using AI and visual detection technology. Background Art

[0002] As the automation level of the manufacturing industry continues to increase, the quality of parts and components can directly affect the quality and performance of the entire product, so the quality inspection of parts and components is crucial. Visual inspection technology can quickly and accurately detect surface defects, dimensions and other issues of parts and components, meeting the requirements of automated production.

[0003] At present, for the visual inspection of parts, it is usually used to use various detection algorithms to identify and detect images after taking pictures with a camera, and automatically determine whether they are qualified. This method has certain disadvantages. On the one hand, in order to facilitate later image processing and detection, the camera usually sets a fixed high-resolution mode to ensure that all images are clear and consistent. However, in the actual operation process, different algorithms are used for different parts due to different detection purposes, and not all algorithms have high resolution requirements for images. High-resolution images not only require longer shooting time, but also waste a lot of invalid time when using algorithm detection, and high-resolution images tend to occupy a lot of storage resources and cause waste. On the other hand, the light intensity of the environment will also affect the shooting time and image quality. When the light intensity is too low, the shooting time is often extended and the image quality is poor. Low-quality images will further affect the algorithm processing time, causing the entire detection process to be delayed. Therefore, at this stage, a more efficient and intelligent visual inspection technology solution is needed to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide an image processing and analysis system and method using AI and visual detection technology to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: an image processing and analysis method using AI and visual detection technology, the method comprising the following steps:

[0006] S100, collecting image inspection records of each component and equipment operating parameters during the visual inspection process of each production line.

[0007] S200. Use an artificial intelligence algorithm to obtain a relationship formula based on the fitting analysis of the image detection records, calculate the theoretical duration and ideal success rate through the relationship formula, calculate the deviation coefficient of the pipeline together with the duration and success rate in the actual process, and mark the abnormal pipeline according to the deviation coefficient.

[0008] S300, analyzing all detection algorithms under the abnormal pipeline to calculate the best quality score, fitting again to obtain the quality score fitting formula, substituting the best quality score into the quality score fitting formula to obtain the relationship formula between light intensity and resolution, referring to the relationship formula and setting the light intensity for adjustment until the deviation coefficient of the abnormal pipeline is normal.

[0009] S400, display the equipment operating parameters of each assembly line in real time through the visualization screen of the data center, generate shooting records and processing records according to the detection information of each component during the visual inspection process, and save them as image detection records.

[0010] In S100, the visual inspection process includes P inspection steps, image capture includes 1 inspection step, image processing includes P-1 inspection steps, and each image processing inspection step includes a different inspection algorithm. Image inspection records include shooting records and processing records; shooting records include light intensity, resolution, shooting time and quality score; processing records include quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure. Equipment operation parameters refer to the parameters of the inspection equipment, including light source brightness and camera parameters. Camera parameters refer to the resolution selected before shooting.

[0011] During the visual inspection process of the production line, the image capture step is performed first, and then the image processing step is performed. Image capture is performed only once, and image processing is performed several times. Image processing is performed in sequence according to the step order.

[0012] Shooting time refers to the time period from the start to the end of shooting. The length of shooting time is affected by the required image resolution, device performance, light intensity, and shooting environment.

[0013] In S200, the specific steps are as follows:

[0014] S201, establish an algorithm set for each assembly line, obtain the current illumination intensity GQ of each assembly line and the detection algorithms of all detection steps included, and put all these detection algorithms into the corresponding algorithm set. Use artificial intelligence algorithm to fit according to the image detection record, and analyze to obtain the relationship formula.

[0015] S202, count the number of parts S detected by each production line within the duration SC, use the images corresponding to these parts as evaluation images, obtain the resolution of the evaluation images, and use the reference-free NIQE image quality evaluation algorithm to calculate the quality score of each evaluation image. Substitute the resolution and quality score of each evaluation image into the relationship formula to calculate the theoretical shooting time LPS of the evaluation image. i and theoretical processing time

[0016] The duration SC is set in advance by the staff. The larger the value, the wider the reference range and the weaker the sensitivity to the current fault.

[0017] The NIQE algorithm uses the features of natural images for modeling and evaluates image quality by calculating the degree of deviation of these features in the image to be evaluated. The NIQE algorithm outputs a score that represents the overall quality level of the image, with lower scores indicating higher image quality.

[0018] S203, analyze the test results of the evaluation image in each test step, and r The test result is the number of successful parts divided by S, which gives the test step BZ r Success rate of CG r . Get detection step BZ r Detection algorithm SF r , statistical detection algorithm SF r The corresponding number of processed records M, the number of processed records with successful detection results divided by M, to obtain the detection step BZ r Ideal success rate of LCG r . Get the shooting time PS in each part image detection record i and processing time Substitute it into the formula to calculate the deviation coefficient of each pipeline. Set the coefficient threshold and mark the pipeline with a deviation coefficient greater than the coefficient threshold as abnormal. The formula is as follows:

[0019]

[0020] Where PCX is the coefficient of deviation and f is a constant.

[0021] The deviation coefficient represents the comprehensive evaluation of the shooting time and overall quality of the images taken by the pipeline. When the shooting time is much longer than the normal time, it means that the light intensity or resolution setting is unreasonable. When the processing time or success rate deviates greatly from the normal value, it means that the overall image quality is not high and needs to be adjusted to reduce the shooting time and improve the image quality.

[0022] In S201, the relationship formula analysis steps are as follows:

[0023] S201-1. Retrieve all shooting records with a light intensity of GQ in the image detection record, obtain the resolution and shooting time in each shooting record, and use the resolution as an independent variable and the shooting time as a dependent variable. The independent variable is used as the input value of the linear regression model, and the output result is compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the relationship formula between the resolution and the shooting time is finally obtained.

[0024] S201-2. Retrieve all processing records with the same detection algorithm as in the algorithm set from the image detection records, and classify these processing records according to the different detection algorithms. Obtain the quality score and processing time in the same processing records, and use the quality score as the independent variable and the processing time as the dependent variable. The independent variable is used as the input value of the linear regression model, and the output result is compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the relationship formula between the quality score and the processing time under each detection algorithm is finally obtained.

[0025] In S300, the specific steps are as follows:

[0026] S301, obtain all detection algorithms in the algorithm set of the abnormal pipeline, and search the processing records corresponding to these detection algorithms in the image detection record. First, classify these processing records according to the different detection algorithms, and calculate the average processing time in each type of processing record. Secondly, the marking process takes less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score ZL ave .

[0027] S302, retrieve all shooting records in the image detection record, obtain the light intensity, resolution and quality score in each shooting record, use the light intensity and resolution as independent variables, and the quality score as the dependent variable. The independent variables are used as the input values ​​of the linear regression model, and the output results are compared with the dependent variables. The quality score fitting formula is finally obtained by continuously adjusting the parameters of the regression coefficient. ave Substituting into the quality score fitting formula, we can get the relationship formula between light intensity and resolution.

[0028] Light intensity refers to the light intensity of the image capture area in the production line. It is affected by the combined effect of natural light brightness and supplementary light brightness. It has a greater impact on the shooting time in environments with different light intensities. In darker environments, the camera needs a longer exposure time to capture enough light to make the image bright enough. Therefore, shooting scenes under low light intensity usually requires a longer exposure time, resulting in an increase in shooting time. In environments with higher light intensity, the camera can usually capture enough light faster and does not require a long exposure time. Therefore, in high light intensity environments, the shooting time may be shorter.

[0029] Resolution refers to the resolution requirement for captured images. Shooting higher resolution images requires more data to capture image details, so it takes longer to process and save image data during the shooting process, thereby increasing the shooting time.

[0030] By adjusting the light intensity and resolution, you can effectively adjust the shooting time and image quality score.

[0031] S303, set the illumination intensity GQ, substitute GQ into the relationship formula to calculate the resolution FBL, adjust the light source brightness in the device operating parameters so that the illumination intensity is GQ, adjust the camera parameters in the device operating parameters so that the resolution is FBL, and then enter step S202 again to determine whether the deviation coefficient is greater than the coefficient threshold. If the result is no, keep the status quo, and if the result is yes, reset the illumination intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold.

[0032] An image processing and analysis system using AI and visual inspection technologies, the system comprising a data acquisition module, an intelligent inspection module, an operation adjustment module and a visualization module.

[0033] The data acquisition module is used to collect the image detection records and equipment operation parameters of each assembly line. The intelligent detection module analyzes the relationship formula based on the image detection records and calculates the theoretical duration and ideal success rate. It calculates the deviation coefficient of the assembly line together with the actual duration and success rate, and marks the abnormal assembly line according to the deviation coefficient. The operation adjustment module analyzes the detection algorithm under the abnormal assembly line and calculates the best quality score. According to the analysis of the best quality score, the relationship formula between light intensity and resolution is obtained. According to the relationship formula, the light intensity of the abnormal assembly line is adjusted until the deviation coefficient is normal. The visualization module is used to display the equipment operation parameters of each assembly line, generate shooting records and processing records and save them as image detection records.

[0034] The data acquisition module includes a detection record acquisition unit, an equipment parameter acquisition unit and a detection step acquisition unit.

[0035] The inspection record acquisition unit is used to collect image inspection records of each component during the visual inspection process, including shooting records and processing records. Shooting records include light intensity, resolution, shooting time and quality score. Processing records include quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure.

[0036] The device parameter acquisition unit is used to collect the parameters of the detection device, including the brightness of the light source and the camera parameters. The camera parameters refer to the resolution selected before shooting.

[0037] The detection step acquisition unit is used to acquire the detection steps of the visual detection process, including 1 image shooting detection step and P-1 image processing detection steps. Each image processing detection step includes a different detection algorithm.

[0038] The intelligent detection module includes a relationship analysis unit and a deviation judgment unit.

[0039] The relationship analysis unit is used to fit the relationship formula. Each assembly line establishes an algorithm set, obtains the current light intensity GQ of the assembly line and the detection algorithms of all detection steps included, and puts all these detection algorithms into the corresponding algorithm set. The artificial intelligence algorithm is used to fit the image detection records and analyze the relationship formula.

[0040] The deviation judgment unit is used to calculate the deviation coefficient of the pipeline.

[0041] First, count the number of parts S inspected by each production line within the duration SC, use the images corresponding to these parts as evaluation images, and calculate the quality score based on the resolution of the evaluation images. Substitute them into the relationship formula to calculate the theoretical shooting time LPS of the evaluation image. i and theoretical processing time

[0042] Secondly, the pipeline offline detection step BZ r The test result is the number of successful parts divided by S, which gives the test step BZ r Success rate of CG r . Get detection step BZ r Detection algorithm SF r , statistical detection algorithm SF r The corresponding number of processed records M, the number of processed records with successful detection results divided by M, to obtain the detection step BZ r Ideal success rate of LCG r .

[0043] Finally, obtain the shooting time PS in the image detection record of each part i and processing time Substituting into the formula: In , the deviation coefficient PCX of each pipeline is calculated respectively. The coefficient threshold is set, and the pipelines with deviation coefficients greater than the coefficient threshold are marked as abnormal. Where f is a constant.

[0044] The operation adjustment module includes an image processing unit and a parameter adjustment unit.

[0045] The image processing unit is used to calculate the best quality score.

[0046] First, the processing records corresponding to all detection algorithms in the algorithm set are retrieved in the image detection records, these processing records are classified according to the different detection algorithms, and the average processing time in each type of processing record is calculated. Secondly, the marking process takes less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score ZL ave .

[0047] The parameter adjustment unit is used to adjust the illumination intensity and resolution.

[0048] First, retrieve all shooting records in the image detection record, take the light intensity and resolution in each shooting record as independent variables, and the quality score as the dependent variable. The independent variables are input into the linear regression model, and the output results are compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the quality score fitting formula is finally obtained. ave Substituting into the quality score fitting formula, we can get the relationship formula between light intensity and resolution.

[0049] Then set the light intensity GQ, calculate the resolution FBL through the relationship formula, adjust the light source brightness in the equipment operating parameters so that the light intensity is GQ and the resolution is FBL. If the deviation coefficient is still greater than the coefficient threshold, reset the light intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold.

[0050] The visualization module displays the equipment operating parameters of each assembly line in real time through the data center visualization screen, generates shooting records and processing records based on the detection information of each component, and saves them as image detection records.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0052] 1. Accurate abnormality judgment: In this application, a relationship formula is fitted according to the image detection record, and the theoretical duration and ideal success rate of each assembly line are calculated by the relationship formula. The deviation coefficient is calculated by comparing with the actual duration and success rate. The abnormal assembly line is found through the deviation coefficient, which is more accurate than the abnormality judgment method of setting a fixed threshold in the prior art.

[0053] 2. Intelligent detection adjustment: This application calculates the best quality score based on the detection algorithm contained in the abnormal pipeline combined with the processing records. After substituting the best quality score into the fitting formula, the relationship between light intensity and resolution is analyzed. The light intensity and resolution are reasonably set according to the relationship between the two. Compared with the mechanical improvement of one value in traditional technology, it is more intelligent.

[0054] In summary, compared with traditional technologies, the present invention has the advantages of accurate abnormality judgment and intelligent detection and adjustment, and can improve the visual inspection efficiency of parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0056] Figure 1 It is a flow chart of an image processing and analysis method using AI and visual detection technology of the present invention;

[0057] Figure 2 It is a structural schematic diagram of an image processing and analysis system using AI and visual detection technology in the present invention. DETAILED DESCRIPTION

[0058] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0059] See also Figure 1 The present invention provides an image processing and analysis method using AI and visual detection technology, the method comprising the following steps:

[0060] S100, collecting image inspection records of each component and equipment operating parameters during the visual inspection process of each production line.

[0061] S200. Use an artificial intelligence algorithm to obtain a relationship formula based on the fitting analysis of the image detection records, calculate the theoretical duration and ideal success rate through the relationship formula, calculate the deviation coefficient of the pipeline together with the duration and success rate in the actual process, and mark the abnormal pipeline according to the deviation coefficient.

[0062] S300, analyzing all detection algorithms under the abnormal pipeline to calculate the best quality score, fitting again to obtain the quality score fitting formula, substituting the best quality score into the quality score fitting formula to obtain the relationship formula between light intensity and resolution, referring to the relationship formula and setting the light intensity for adjustment until the deviation coefficient of the abnormal pipeline is normal.

[0063] S400, display the equipment operating parameters of each assembly line in real time through the visualization screen of the data center, generate shooting records and processing records according to the detection information of each component during the visual inspection process, and save them as image detection records.

[0064] In S100, the visual inspection process includes P inspection steps, image capture includes 1 inspection step, image processing includes P-1 inspection steps, and each image processing inspection step includes a different inspection algorithm. Image inspection records include shooting records and processing records; shooting records include light intensity, resolution, shooting time and quality score; processing records include quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure. Equipment operation parameters refer to the parameters of the inspection equipment, including light source brightness and camera parameters. Camera parameters refer to the resolution selected before shooting.

[0065] During the visual inspection process of the production line, the image capture step is performed first, and then the image processing step is performed. Image capture is performed only once, and image processing is performed several times. Image processing is performed in sequence according to the step order.

[0066] Shooting time refers to the time period from the start to the end of shooting. The length of shooting time is affected by the required image resolution, device performance, light intensity, and shooting environment.

[0067] In S200, the specific steps are as follows:

[0068] S201, establish an algorithm set for each assembly line, obtain the current illumination intensity GQ of each assembly line and the detection algorithms of all detection steps included, and put all these detection algorithms into the corresponding algorithm set. Use artificial intelligence algorithm to fit according to the image detection record, and analyze to obtain the relationship formula.

[0069] S202, count the number of parts S detected by each production line within the duration SC, use the images corresponding to these parts as evaluation images, obtain the resolution of the evaluation images, and use the reference-free NIQE image quality evaluation algorithm to calculate the quality score of each evaluation image. Substitute the resolution and quality score of each evaluation image into the relationship formula to calculate the theoretical shooting time LPS of the evaluation image. i and theoretical processing time

[0070] The duration SC is set in advance by the staff. The larger the value, the wider the reference range and the weaker the sensitivity to the current fault.

[0071] The NIQE algorithm uses the features of natural images for modeling and evaluates image quality by calculating the degree of deviation of these features in the image to be evaluated. The NIQE algorithm outputs a score that represents the overall quality level of the image, with lower scores indicating higher image quality.

[0072] S203, analyze the test results of the evaluation image in each test step, and r The test result is the number of successful parts divided by S, which gives the test step BZ r Success rate of CG r . Get detection step BZ r Detection algorithm SF r , statistical detection algorithm SF r The corresponding number of processed records M, the number of processed records with successful detection results divided by M, to obtain the detection step BZ r Ideal success rate of LCG r . Get the shooting time PS in each part image detection record i and processing time Substitute it into the formula to calculate the deviation coefficient of each pipeline. Set the coefficient threshold and mark the pipeline with a deviation coefficient greater than the coefficient threshold as abnormal. The formula is as follows:

[0073]

[0074] Where PCX is the coefficient of deviation and f is a constant.

[0075] The deviation coefficient represents the comprehensive evaluation of the shooting time and overall quality of the images taken by the pipeline. When the shooting time is much longer than the normal time, it means that the light intensity or resolution setting is unreasonable. When the processing time or success rate deviates greatly from the normal value, it means that the overall image quality is not high and needs to be adjusted to reduce the shooting time and improve the image quality.

[0076] In S201, the relationship formula analysis steps are as follows:

[0077] S201-1. Retrieve all shooting records with a light intensity of GQ in the image detection record, obtain the resolution and shooting time in each shooting record, and use the resolution as an independent variable and the shooting time as a dependent variable. The independent variable is used as the input value of the linear regression model, and the output result is compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the relationship formula between the resolution and the shooting time is finally obtained.

[0078] S201-2. Retrieve all processing records with the same detection algorithm as in the algorithm set from the image detection records, and classify these processing records according to the different detection algorithms. Obtain the quality score and processing time in the same processing records, and use the quality score as the independent variable and the processing time as the dependent variable. The independent variable is used as the input value of the linear regression model, and the output result is compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the relationship formula between the quality score and the processing time under each detection algorithm is finally obtained.

[0079] In S300, the specific steps are as follows:

[0080] S301, obtain all detection algorithms in the algorithm set of the abnormal pipeline, and search the processing records corresponding to these detection algorithms in the image detection record. First, classify these processing records according to the different detection algorithms, and calculate the average processing time in each type of processing record. Secondly, the marking process takes less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score ZL ave .

[0081] S302, retrieve all shooting records in the image detection record, obtain the light intensity, resolution and quality score in each shooting record, use the light intensity and resolution as independent variables, and the quality score as the dependent variable. The independent variables are used as the input values ​​of the linear regression model, and the output results are compared with the dependent variables. The quality score fitting formula is finally obtained by continuously adjusting the parameters of the regression coefficient. ave Substituting into the quality score fitting formula, we can get the relationship formula between light intensity and resolution.

[0082] Light intensity refers to the light intensity of the image capture area in the production line. It is affected by the combined effect of natural light brightness and supplementary light brightness. It has a greater impact on the shooting time in environments with different light intensities. In darker environments, the camera needs a longer exposure time to capture enough light to make the image bright enough. Therefore, shooting scenes under low light intensity usually requires a longer exposure time, resulting in an increase in shooting time. In environments with higher light intensity, the camera can usually capture enough light faster and does not require a long exposure time. Therefore, in high light intensity environments, the shooting time may be shorter.

[0083] Resolution refers to the resolution requirement for captured images. Shooting higher resolution images requires more data to capture image details, so it takes longer to process and save image data during the shooting process, thereby increasing the shooting time.

[0084] By adjusting the light intensity and resolution, you can effectively adjust the shooting time and image quality score.

[0085] S303, set the illumination intensity GQ, substitute GQ into the relationship formula to calculate the resolution FBL, adjust the light source brightness in the device operating parameters so that the illumination intensity is GQ, adjust the camera parameters in the device operating parameters so that the resolution is FBL, and then enter step S202 again to determine whether the deviation coefficient is greater than the coefficient threshold. If the result is no, keep the status quo, and if the result is yes, reset the illumination intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold.

[0086] See also Figure 2 The present invention provides an image processing and analysis system using AI and visual detection technology, the system comprising a data acquisition module, an intelligent detection module, an operation adjustment module and a visualization module.

[0087] The data acquisition module is used to collect the image detection records and equipment operation parameters of each assembly line. The intelligent detection module analyzes the relationship formula based on the image detection records and calculates the theoretical duration and ideal success rate. It calculates the deviation coefficient of the assembly line together with the actual duration and success rate, and marks the abnormal assembly line according to the deviation coefficient. The operation adjustment module analyzes the detection algorithm under the abnormal assembly line and calculates the best quality score. According to the analysis of the best quality score, the relationship formula between light intensity and resolution is obtained. According to the relationship formula, the light intensity of the abnormal assembly line is adjusted until the deviation coefficient is normal. The visualization module is used to display the equipment operation parameters of each assembly line, generate shooting records and processing records and save them as image detection records.

[0088] The data acquisition module includes a detection record acquisition unit, an equipment parameter acquisition unit and a detection step acquisition unit.

[0089] The inspection record acquisition unit is used to collect image inspection records of each component during the visual inspection process, including shooting records and processing records. Shooting records include light intensity, resolution, shooting time and quality score. Processing records include quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure.

[0090] The device parameter acquisition unit is used to collect the parameters of the detection device, including the brightness of the light source and the camera parameters. The camera parameters refer to the resolution selected before shooting.

[0091] The detection step acquisition unit is used to acquire the detection steps of the visual detection process, including 1 image shooting detection step and P-1 image processing detection steps. Each image processing detection step includes a different detection algorithm.

[0092] The intelligent detection module includes a relationship analysis unit and a deviation judgment unit.

[0093] The relationship analysis unit is used to fit the relationship formula. Each assembly line establishes an algorithm set, obtains the current light intensity GQ of the assembly line and the detection algorithms of all detection steps included, and puts all these detection algorithms into the corresponding algorithm set. The artificial intelligence algorithm is used to fit the image detection records and analyze the relationship formula.

[0094] The deviation judgment unit is used to calculate the deviation coefficient of the pipeline.

[0095] First, count the number of parts S inspected by each production line within the duration SC, use the images corresponding to these parts as evaluation images, and calculate the quality score based on the resolution of the evaluation images. Substitute them into the relationship formula to calculate the theoretical shooting time LPS of the evaluation image. i and theoretical processing time

[0096] Secondly, the pipeline offline detection step BZ r The test result is the number of successful parts divided by S, which gives the test step BZ r Success rate of CG r . Get detection step BZ r Detection algorithm SF r , statistical detection algorithm SF r The corresponding number of processed records M, the number of processed records with successful detection results divided by M, to obtain the detection step BZ r Ideal success rate of LCG r .

[0097] Finally, obtain the shooting time PS in the image detection record of each part i and processing time Substituting into the formula: In , the deviation coefficient PCX of each pipeline is calculated respectively. The coefficient threshold is set, and the pipelines with deviation coefficients greater than the coefficient threshold are marked as abnormal. Where f is a constant.

[0098] The operation adjustment module includes an image processing unit and a parameter adjustment unit.

[0099] The image processing unit is used to calculate the best quality score.

[0100] First, the processing records corresponding to all detection algorithms in the algorithm set are retrieved in the image detection records, these processing records are classified according to the different detection algorithms, and the average processing time in each type of processing record is calculated. Secondly, the marking process takes less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score ZL ave .

[0101] The parameter adjustment unit is used to adjust the illumination intensity and resolution.

[0102] First, retrieve all shooting records in the image detection record, take the light intensity and resolution in each shooting record as independent variables, and the quality score as the dependent variable. The independent variables are input into the linear regression model, and the output results are compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the quality score fitting formula is finally obtained. ave Substituting into the quality score fitting formula, we can get the relationship formula between light intensity and resolution.

[0103] Then set the light intensity GQ, calculate the resolution FBL through the relationship formula, adjust the light source brightness in the equipment operating parameters so that the light intensity is GQ and the resolution is FBL. If the deviation coefficient is still greater than the coefficient threshold, reset the light intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold.

[0104] The visualization module displays the equipment operating parameters of each assembly line in real time through the data center visualization screen, generates shooting records and processing records based on the detection information of each component, and saves them as image detection records.

[0105] Embodiment 1:

[0106] Assume that the number of parts inspected by assembly line A1 in a period of time is 1, and assembly line A1 has three inspection steps. The details of each inspection step are as follows:

[0107] Detection step 1 (image capture):

[0108] Theoretical shooting time: 0.28s; shooting time 0.56s;

[0109] Detection step 2 (image processing):

[0110] Theoretical processing time: 0.9s; processing time 1.2s; ideal success rate 96%; success rate 86%;

[0111] Detection step 3 (image processing):

[0112] Theoretical processing time: 0.6s; processing time 0.8s; ideal success rate 98%; success rate 90%;

[0113] Assuming the constant f is 2, substitute the formula to calculate the deviation coefficient of pipeline A1:

[0114]

[0115] Then the deviation coefficient of pipeline A1 is 1.274.

[0116] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device.

[0117] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An image processing and analysis method using AI and visual inspection technology, characterized in that: The method comprises the following steps: S100, collecting image inspection records of each component and equipment operating parameters during the visual inspection process of each production line; S200, using an artificial intelligence algorithm to obtain a relationship formula based on the image detection record fitting analysis, calculating the theoretical duration and the ideal success rate through the relationship formula, calculating the deviation coefficient of the pipeline together with the duration and success rate in the actual process, and marking the abnormal pipeline according to the deviation coefficient; S300, analyzing all detection algorithms under the abnormal pipeline to calculate the best quality score, fitting again to obtain a quality score fitting formula, substituting the best quality score into the quality score fitting formula to obtain a relationship formula between light intensity and resolution, referring to the relationship formula and setting the light intensity for adjustment until the deviation coefficient of the abnormal pipeline is normal; S400, displaying the equipment operating parameters of each assembly line in real time through the visualization screen of the data center, generating shooting records and processing records according to the detection information of each component in the visual inspection process, and saving them as image detection records; In S100, the visual inspection process includes P inspection steps, image capture includes 1 inspection step, image processing includes P-1 inspection steps, and each image processing inspection step includes a different inspection algorithm; the image inspection record includes a shooting record and a processing record; the shooting record includes light intensity, resolution, shooting time and quality score; the processing record includes quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure; the equipment operation parameters refer to the parameters of the inspection equipment, including light source brightness and camera parameters, and the camera parameters refer to the resolution selected before shooting; In S200, the specific steps are as follows: S201. Establish an algorithm set for each pipeline to obtain the current light intensity of each pipeline. And the detection algorithms of all the detection steps included, all these detection algorithms are put into the corresponding algorithm set; artificial intelligence algorithm is used to fit according to the image detection records, and the relationship formula is obtained by analysis; S202, Statistical duration The number of parts tested on each production line , take the images corresponding to these parts as evaluation images, obtain the resolution of the evaluation images, and use the reference-free NIQE image quality evaluation algorithm to calculate the quality score of each evaluation image; substitute the resolution and quality score of each evaluation image into the relationship formula to calculate the theoretical shooting time of the evaluation image. and theoretical processing time ; S203, analyzing the test results of the evaluation image in each test step, and The inspection result is the number of successful parts divided by , get the detection steps Success rate ; Get the detection steps Detection algorithm , statistical detection algorithm The corresponding number of processing records , divide the number of processed records with successful detection results by , get the detection steps The ideal success rate ; Get the shooting time of each part image detection record and processing time , substitute it into the formula to calculate the deviation coefficient of each pipeline respectively; set the coefficient threshold, and mark the pipeline with a deviation coefficient greater than the coefficient threshold as abnormal; the formula is as follows: ， In the formula, is the coefficient of variation, is a constant; In S201, the relationship formula analysis steps are as follows: S201-1. Retrieve the light intensity in the image detection record. All shooting records are obtained, and the resolution and shooting time in each shooting record are obtained. The resolution is used as the independent variable and the shooting time is used as the dependent variable. The independent variable is used as the input value of the linear regression model, and the output result is compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the relationship formula between the resolution and the shooting time is finally obtained. S201-2, searching all processing records of the same detection algorithm as that in the algorithm set in the image detection record, and classifying these processing records according to the different detection algorithms; obtaining the quality score and processing time in the same processing records, taking the quality score as the independent variable and the processing time as the dependent variable; taking the independent variable as the input value of the linear regression model, comparing the output result with the dependent variable, and finally obtaining the relationship formula between the quality score and the processing time under each detection algorithm by continuously adjusting the parameters of the regression coefficient; In S300, the specific steps are as follows: S301, obtain all detection algorithms in the algorithm set of the abnormal pipeline, and retrieve the processing records corresponding to these detection algorithms in the image detection record; first, classify these processing records according to the different detection algorithms, and calculate the average processing time in each type of processing record ; Secondly, the marking processing time is less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. ; Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score. ; S302, searching all shooting records in the image detection record, obtaining the light intensity, resolution and quality score in each shooting record, taking the light intensity and resolution as independent variables and the quality score as dependent variable; taking the independent variables as input values ​​of the linear regression model, comparing the output results with the dependent variables, and finally obtaining the quality score fitting formula by continuously adjusting the parameters of the regression coefficient; Substituting into the quality score fitting formula, we get the relationship formula between light intensity and resolution; S303, set the illumination intensity GQ, substitute GQ into the relationship formula to calculate the resolution FBL, adjust the light source brightness in the device operating parameters so that the illumination intensity is GQ, adjust the camera parameters in the device operating parameters so that the resolution is FBL, and then enter step S202 again to determine whether the deviation coefficient is greater than the coefficient threshold; if the result is no, keep the status quo; if the result is yes, reset the illumination intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold.

2. An image processing and analysis system using AI and visual inspection technology, characterized in that: The system includes a data acquisition module, an intelligent detection module, an operation adjustment module and a visualization module; The data acquisition module is used to collect image detection records and equipment operating parameters of each production line; The intelligent detection module analyzes the relationship formula based on the image detection records and calculates the theoretical duration and ideal success rate. It calculates the deviation coefficient of the pipeline together with the actual duration and success rate, and marks the abnormal pipeline according to the deviation coefficient. The operation adjustment module analyzes the detection algorithm under the abnormal pipeline and calculates the best quality score. The relationship formula between light intensity and resolution is obtained based on the analysis of the best quality score. The light intensity of the abnormal pipeline is adjusted according to the relationship formula until the deviation coefficient is normal. The visualization module is used to display the equipment operating parameters of each assembly line, generate shooting records and processing records and save them as image detection records; The data acquisition module includes a detection record acquisition unit, an equipment parameter acquisition unit and a detection step acquisition unit; The inspection record acquisition unit is used to collect image inspection records of each component during the visual inspection process, including shooting records and processing records; shooting records include light intensity, resolution, shooting time and quality score; processing records include quality score, inspection algorithm, processing time and inspection result, and the inspection result is success or failure; The device parameter acquisition unit is used to acquire the parameters of the detection device, including the light source brightness and camera parameters; the camera parameters refer to the resolution selected before shooting; The detection step acquisition unit is used to acquire the detection steps of the visual detection process, including 1 image shooting detection step and P-1 image processing detection steps; each image processing detection step includes a different detection algorithm; The intelligent detection module includes a relationship analysis unit and a deviation judgment unit; The relationship analysis unit is used to fit the relationship formula; each pipeline establishes an algorithm set to obtain the current light intensity of the pipeline And the detection algorithms of all the detection steps included, all these detection algorithms are put into the corresponding algorithm set; artificial intelligence algorithm is used to fit according to the image detection records, and the relationship formula is obtained by analysis; The deviation judgment unit is used to calculate the deviation coefficient of the pipeline; First, count the duration The number of parts tested on each production line , take the images corresponding to these parts as the evaluation images, and calculate the quality score according to the resolution of the evaluation images; substitute them into the relationship formula to calculate the theoretical shooting time of the evaluation images and theoretical processing time ; Secondly, the offline detection step The inspection result is the number of successful parts divided by , get the detection steps Success rate ; Get the detection steps Detection algorithm , statistical detection algorithm The corresponding number of processing records , divide the number of processed records with successful detection results by , get the detection steps The ideal success rate ; Finally, get the shooting time in the image detection record of each part and processing time , substitute into the formula: In the above example, the deviation coefficient of each pipeline is calculated separately. ; Set the coefficient threshold and mark the pipeline whose deviation coefficient is greater than the coefficient threshold as abnormal; is a constant; The operation adjustment module includes an image processing unit and a parameter adjustment unit; The image processing unit is used to calculate the best quality score; First, the processing records corresponding to all detection algorithms in the algorithm set are retrieved in the image detection records, these processing records are classified according to the different detection algorithms, and the average processing time in each type of processing record is calculated. ; Secondly, the marking processing time is less than The processing time of the marked processing records is calculated again to obtain the optimal processing time. ; Finally, Substitute the corresponding detection algorithm into the relationship formula to calculate the quality score, and calculate the average of the quality scores corresponding to all detection algorithms in the algorithm set to obtain the best quality score. ; The parameter adjustment unit is used to adjust the illumination intensity and resolution; First, retrieve all shooting records in the image detection record, take the light intensity and resolution in each shooting record as independent variables, and the quality score as the dependent variable. The independent variables are input into the linear regression model, and the output results are compared with the dependent variable. By continuously adjusting the parameters of the regression coefficient, the quality score fitting formula is finally obtained. Substituting into the quality score fitting formula, we get the relationship formula between light intensity and resolution; Then set the illumination intensity GQ, calculate the resolution FBL through the relationship formula, adjust the light source brightness in the equipment operation parameters so that the illumination intensity is GQ and the resolution is FBL. If the deviation coefficient is still greater than the coefficient threshold, reset the illumination intensity and calculate the resolution until it is determined that the deviation coefficient is not greater than the coefficient threshold. The visualization module displays the equipment operating parameters of each assembly line in real time through the data center visualization screen, generates shooting records and processing records based on the detection information of each component, and saves them as image detection records.

Images