A battery appearance abnormality detection system and method based on visual recognition

Abstract

The present invention discloses a battery appearance anomaly detection system and method based on visual recognition, belonging to the technical field of battery anomaly detection. The present invention captures the appearance image of the battery by using a camera, collects the environmental information of the battery by a sensor, extracts the basic information of the battery in the database, constructs a three-dimensional appearance model of the battery, and generates a battery appearance comparison image; analyzes the historical appearance image and historical environmental information of the battery, constructs an image quality evaluation model and an image evaluation and environment association model; evaluates the image quality of the real-time appearance image of the battery, and adjusts the parameters of the lighting device in real time; compares the real-time appearance image of the battery with the battery appearance comparison image to determine the abnormal area; generates an anomaly detection result based on the comparison of the abnormal area with the database, and gives a prompt and alarm to the user. The accuracy and stability of battery appearance anomaly detection are improved, and the asset safety of users is effectively guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of battery anomaly detection, and particularly to a battery appearance anomaly detection system and method based on visual recognition. Background Art

[0002] With the popularization of electric vehicles, portable electronic devices, etc., batteries, as an efficient energy storage solution, have become increasingly important. However, the safety of batteries is directly related to the reliability of the products in use and user safety. Defects in the outer shell appearance may lead to a decline in battery performance and even cause safety problems.

[0003] Traditional manual detection methods are inefficient and there is a risk of missed detection; in existing machine vision systems, the real-time images of the batteries collected are not evaluated for image quality, and the parameters of image acquisition are not adjusted to improve image quality, which is not conducive to the accuracy of detection and there is a risk of misdetection.

[0004] Therefore, the present invention discloses a battery appearance anomaly detection system and method based on visual recognition to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide a battery appearance anomaly detection system and method based on visual recognition to solve the problems raised in the above background art.

[0006] To solve the above technical problems, the present invention provides the following technical solution: A battery appearance anomaly detection method based on visual recognition, the method includes the following steps:

[0007] S1: Use a camera to capture the appearance image of the battery, and a sensor to collect the environmental information of the battery. Upload the basic information, appearance image, and environmental information of the battery to the database, extract the basic information of the battery in the database, construct a three-dimensional appearance model of the battery, and generate a battery appearance comparison image according to the environmental information and the three-dimensional appearance model;

[0008] S2: Obtain the historical appearance images and historical environmental information of the battery, analyze the historical appearance images and historical environmental information of the battery, construct an image quality evaluation model based on the historical appearance images, and construct an image evaluation and environment correlation model according to the image quality evaluation results of the historical appearance images and the historical environmental information;

[0009] S3: Evaluate the image quality of the real-time appearance image of the battery based on the image quality evaluation model, and adjust the parameters of the lighting device in real time according to the image quality evaluation results and the image evaluation and environment correlation model;

[0010] S4: After adjusting the parameters of the lighting device, obtain the real-time appearance image of the battery again, compare the real-time appearance image of the battery with the comparison image of the battery appearance, and determine the abnormal area in the real-time appearance image of the battery;

[0011] S5: Based on the abnormal area comparison database, generate an abnormal detection result, and give a prompt and alarm to the user.

[0012] According to the above solution, the basic information of the battery in step S1 includes the identification information, type information, and specification information of the battery; the identification information corresponds to the battery one by one, the identification information includes the label information of the battery, and all information of the corresponding battery can be searched and extracted from the database according to the label information; the specification information includes the model information and standard appearance information of the battery; there are no defects and quality problems in the standard appearance information;

[0013] The environmental information includes light information, battery pose information, and the relative position between the battery and the camera; the light information includes the light value of the lighting device and the light value of the non-lighting device;

[0014] The light value of the non-lighting device is collected by a sensor for the natural light brightness when the lighting device is not turned on;

[0015] According to the specification information of the battery, create a three-dimensional appearance model of the battery using 3D modeling software; create a three-dimensional coordinate system based on the three-dimensional appearance model, and map the camera to the mapping point in the three-dimensional coordinate system based on the battery pose information and the relative position between the battery and the camera;

[0016] Creating a three-dimensional appearance model of the battery based on 3D modeling software can effectively present the appearance of a standard battery without problems.

[0017] Taking the mapping point of the camera in the three-dimensional coordinate system as the starting point, generate a shooting range ray according to the shooting range of the camera; generate a battery appearance comparison image based on the shooting range ray, camera shooting parameters, and three-dimensional appearance model, and perform image grayscale processing on the original color battery appearance comparison image.

[0018] Based on the three-dimensional appearance model, it can more accurately provide the basic information of the battery appearance comparison image; improve the stability and accuracy of the system's abnormal detection.

[0019] According to the above solution, in step S2, perform image grayscale processing on the historical appearance images collected by one camera according to the same rules, and combine them to generate a historical appearance image set, denoted as HAI = {HAI i |i ∈ [1, N]}, where HAI i represents the historical appearance image with the serial number i in the historical appearance image set, and N represents the total number of historical appearance images in the historical appearance image set;

[0020] For the historical appearance image HAI i , calculate the evaluation image quality parameters of the historical appearance image respectively, and normalize each evaluation image quality parameter; the evaluation image quality parameters include the Laplace transform value SH i , the standard deviation CO of the histogram i , the average value BR of the pixels i and the peak signal-to-noise ratio NO i ;

[0021] Construct an image quality evaluation model based on the evaluation image quality parameters of the historical appearance image:

[0022] IQV i =A 1 ×SH i +A 2 ×CO i +A 3 ×BR i +A 4 ×NO i ;

[0023] Where IQV i represents the image quality value of the historical appearance image HAI i , A 1 , A 2 , A 3 and A 4 are the weight values of the Laplace transform value, the standard deviation of the histogram, the average value of the pixels, and the peak signal-to-noise ratio of the historical appearance image respectively, where A 1 +A 2 +A 3 +A 4 =1;

[0024] The Laplace transform value can effectively evaluate the clarity of the image, the standard deviation of the histogram can effectively evaluate the contrast of the image, the average value of the pixels can effectively evaluate the brightness of the image, and the peak signal-to-noise ratio can effectively evaluate the noise level of the image; evaluating the overall image quality of the historical appearance image based on the above multiple parameters can effectively improve the accuracy of the evaluation;

[0025] Evaluate the image quality of the historical appearance image HAI based on the image quality evaluation model i ; According to the image quality value IQV i of the historical appearance image HAI i , the lighting device brightness value LFB i and the non-lighting device brightness value NLFB i Construct an image evaluation and environment association set, denoted as IEE ={(LFBi , NLFB i , IQV i ) | i ∈ [1, N]}; Construct an image evaluation and environment association model based on the image evaluation and environment association set: z = α 1 ×x + α 2 ×y + α 3 ; where α 1 , α 2 and α 3 represent fitting coefficients, x represents the independent variable of the illumination device brightness value, y represents the independent variable of the non-illumination device brightness value, z represents the dependent variable of the image quality value. Use the least squares method to calculate and solve α 1 , α 2 and α 3 in the image evaluation and environment association model.

[0026] Constructing an image evaluation and environment association model using the image quality value, illumination device brightness value, and non-illumination device brightness value can effectively associate image evaluation and brightness information, providing a numerical basis for subsequent adjustment of illumination device parameters and effectively improving the automation of the system;

[0027] According to the above solution, in step S3, calculate the evaluation image quality parameters of the real-time appearance image of the battery respectively, and normalize each evaluation image quality parameter; substitute the normalized evaluation image quality parameters into the image quality evaluation model to evaluate the image quality of the real-time appearance image of the battery;

[0028] Set an image quality value threshold for the image quality value. If the image quality value of the real-time appearance image of the battery is greater than or equal to the image quality value threshold, directly jump to step S4;

[0029] If the image quality value of the real-time appearance image of the battery is less than the image quality value threshold, substitute the image quality value threshold and the non-illumination device brightness value into the fitted image evaluation and environment association model, calculate the illumination device brightness value, and adjust the illumination device parameters to reach the illumination device brightness value.

[0030] Adjusting the illumination device parameters for the real-time appearance image of the battery with poor image quality improves the image quality of the real-time appearance image of the battery and enhances the accuracy and stability of battery appearance abnormality detection;

[0031] According to the above solution, in step S4, after adjusting the illumination device parameters, obtain the real-time appearance image of the battery again. When the real-time appearance image of the battery and the battery appearance comparison image are collected, the battery pose information, the relative position between the battery and the camera, and the focal length of the camera are the same;

[0032] By setting the battery pose information, the relative position between the battery and the camera, and the focal length of the camera, images with the same camera field of view can be obtained, improving the subsequent comparison efficiency;

[0033] Compare the real-time appearance image of the battery that has been subjected to image grayscale processing separately with the battery appearance comparison image, including:

[0034] Split the real-time appearance image of the battery and the battery appearance comparison image into pixel blocks of size P×Q;

[0035] After splitting the real-time appearance image of the battery and the battery appearance comparison image, calculate the similarity. When calculating, use a distributed system to calculate the pixel degrees of multiple groups of pixel blocks together, improving the calculation efficiency of the system;

[0036] Perform pixel value similarity PVS calculation on the pixel blocks of the real-time appearance image of the battery and the corresponding pixel blocks of the battery appearance comparison image. The specific calculation formula is as follows:

[0037]

[0038] Where RPV (p，q) represents the pixel value when the pixel block coordinates in the real-time appearance image of the battery are (p, q), and CPV (p，q) represents the pixel value when the pixel block coordinates in the battery appearance comparison image are (p, q). Determine the abnormal area in the real-time appearance image of the battery, where p ∈ [1, P] and q ∈ [1, Q];

[0039] Set a similarity threshold for the pixel value similarity of the pixel blocks. Extract all pixel blocks with a pixel value similarity greater than the similarity threshold, and arrange the pixel blocks in the order of the original real-time appearance image of the battery, which is recorded as the abnormal area of the real-time appearance image of the battery.

[0040] According to the above solution, in step S5, based on the abnormal area, compare the defect database, and the database includes images of common abnormal phenomena of the battery;

[0041] Generate the specific abnormal type of the abnormal area. According to the specific abnormal type, give different levels of prompts and alarms for the system settings to the user.

[0042] Give different levels of prompts and alarms for the system settings to the user; enable the user to know the degree of abnormality and effectively protect the user's asset safety.

[0043] Another aspect of the present application is a battery appearance abnormality detection system based on visual recognition. The system is implemented by applying the above-mentioned battery appearance abnormality detection method based on visual recognition. The system includes a battery data acquisition module, a database, an appearance comparison image generation module, an evaluation association model construction module, a parameter adjustment module, and an abnormality comparison and prompt module;

[0044] The battery data acquisition module uses a camera to capture the appearance image of the battery, and a sensor to collect the environmental information of the battery; the database is used to receive and store battery data;

[0045] The appearance comparison image generation module is used to extract the basic information of the battery in the database, construct a three-dimensional appearance model of the battery, and generate a battery appearance comparison image according to the environmental information and the three-dimensional appearance model;

[0046] The evaluation association model construction module is used to obtain the historical appearance images and historical environmental information of the battery, analyze the historical appearance images and historical environmental information of the battery, construct an image quality evaluation model based on the historical appearance images, and construct an image evaluation and environment association model according to the image quality evaluation results of the historical appearance images and the historical environmental information;

[0047] The parameter adjustment module uses the image quality evaluation model to evaluate the real-time appearance image of the battery, calculates the brightness value of the lighting device using the environment association model, and adjusts the parameters of the lighting device in real time;

[0048] The abnormality comparison and prompt module is used to compare the real-time appearance image of the battery with the battery appearance comparison image, determine the abnormal area in the real-time appearance image of the battery, generate an abnormality detection result based on the comparison of the abnormal area with the database, and prompt and alarm the user.

[0049] According to the above solution, the appearance comparison image generation module includes a three-dimensional model and mapping unit and a comparison image generation unit;

[0050] The three-dimensional model and mapping unit creates a three-dimensional appearance model of the battery according to the specification information of the battery; creates a three-dimensional coordinate system based on the three-dimensional appearance model, and maps the camera to the mapping point in the three-dimensional coordinate system based on the battery pose information and the relative position between the battery and the camera;

[0051] The comparison image generation unit takes the mapping point of the camera in the three-dimensional coordinate system as the starting point, and generates a shooting range ray according to the shooting range of the camera; generates a battery appearance comparison image based on the shooting range ray, the camera shooting parameters, and the three-dimensional appearance model.

[0052] According to the above solution, the evaluation association model construction module includes an image quality evaluation unit and an image evaluation and environment association unit;

[0053] The image quality evaluation unit is used to calculate the evaluation image quality parameters of historical appearance images, and construct an image quality evaluation model based on the evaluation image quality parameters of historical appearance images;

[0054] The image evaluation and environment association unit constructs an image evaluation and environment association set based on the image quality value, the brightness value of the lighting device, and the brightness value of the non-lighting device, and constructs an image evaluation and environment association model based on the image evaluation and environment association set.

[0055] According to the above solution, the anomaly comparison and prompt module includes an image anomaly comparison unit and an alarm prompt unit;

[0056] The image anomaly comparison unit is used to split the real-time appearance image of the battery and the appearance comparison image of the battery into pixel blocks of the same size, calculate the pixel value similarity of the two images, and generate an anomaly area;

[0057] The alarm prompt unit is used to compare the anomaly area with the defect database, generate the specific anomaly type of the anomaly area, and perform different levels of prompts and alarms set by the system for the user according to the specific anomaly type.

[0058] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: creating a three-dimensional appearance model of the battery based on 3D modeling software can effectively present the appearance of a standard battery without problems; based on the three-dimensional appearance model, the basic information of the battery appearance comparison image can be provided more accurately, improving the stability and accuracy of system anomaly detection; evaluating the overall image quality of historical appearance images based on multiple parameters can effectively improve the accuracy of evaluation; associating image evaluation with brightness information provides a numerical basis for subsequent adjustment of lighting device parameters, improving the automation level of the system; adjusting the lighting device parameters for the real-time appearance image of the battery with poor image quality to improve the accuracy and stability of battery appearance anomaly detection; splitting the real-time appearance image of the battery and the appearance comparison image of the battery and then calculating the similarity, and using a distributed system to calculate the pixel degree of multiple groups of pixel blocks together during the calculation to improve the calculation efficiency of the system; performing different levels of prompts and alarms set by the system for the user can enable the user to know the degree of anomaly and effectively protect the user's asset safety. Description of the Drawings

[0059] The 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 to the present invention. In the drawings:

[0060] Figure 1 It is a schematic flowchart of a method for detecting battery appearance anomalies based on visual recognition according to the present invention;

[0061] Figure 2 It is a schematic structural diagram of an abnormal battery appearance detection system based on visual recognition of the present invention. Specific implementation mode

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0063] Embodiment 1: Please refer to Figure 1 , the present invention provides a technical solution: an abnormal battery appearance detection method based on visual recognition, the method includes the following steps:

[0064] S1: Use a camera to capture the appearance image of the battery, and a sensor to collect the environmental information of the battery. Upload the basic information, appearance image and environmental information of the battery to the database, extract the basic information of the battery in the database, construct a three-dimensional appearance model of the battery, and generate a battery appearance comparison image according to the environmental information and the three-dimensional appearance model;

[0065] According to the above solution, the basic information of the battery in step S1 includes the identification information, type information and specification information of the battery; the identification information corresponds to the battery one by one, and the identification information includes the label information of the battery. According to the label information, all information of the corresponding battery can be searched and extracted from the database; the specification information includes the model information and standard appearance information of the battery; there are no defects and quality problems in the standard appearance information;

[0066] The environmental information includes light information, battery pose information and the relative position between the battery and the camera; the light information includes the light value of the lighting equipment and the light value of the non-lighting equipment;

[0067] Create a three-dimensional appearance model of the battery using three-dimensional modeling software according to the specification information of the battery; create a three-dimensional coordinate system based on the three-dimensional appearance model, and map the camera to the mapping point in the three-dimensional coordinate system based on the battery pose information and the relative position between the battery and the camera;

[0068] Starting from the mapping point of the camera in the three-dimensional coordinate system, generate a shooting range ray according to the shooting range of the camera; generate a battery appearance comparison image based on the shooting range ray, camera shooting parameters and three-dimensional appearance model, and perform image grayscale processing on the original color battery appearance comparison image.

[0069] S2: Obtain the historical appearance images and historical environmental information of the battery, analyze the historical appearance images and historical environmental information of the battery, construct an image quality evaluation model based on the historical appearance images, and construct an image evaluation and environment association model according to the image quality evaluation results and historical environmental information of the historical appearance images;

[0070] According to the above solution, in step S2, perform image grayscale processing on the historical appearance images collected by a camera according to the same rules, and combine them to generate a set of historical appearance images, denoted as HAI = {HAI i |i ∈ [1, N]}, where HAI i represents the historical appearance image with serial number i in the set of historical appearance images, and N represents the total number of historical appearance images in the set of historical appearance images;

[0071] For the historical appearance image HAI i , calculate the evaluation image quality parameters of the historical appearance image respectively, and normalize each evaluation image quality parameter; the evaluation image quality parameters include the Laplace transform value SH i , the standard deviation CO of the histogram i , the average value BR of the pixels i , and the peak signal-to-noise ratio NO i ;

[0072] Construct an image quality evaluation model based on the evaluation image quality parameters of the historical appearance images:

[0073] IQV i = A 1 × SH i + A 2 × CO i + A 3 × BR i + A 4 × NO i ;

[0074] where IQV i represents the image quality value of the historical appearance image HAI i , A 1 , A 2 , A 3 , and A 4 are the weight values of the Laplace transform value, the standard deviation of the histogram, the average value of the pixels, and the peak signal-to-noise ratio of the historical appearance image respectively, where A 1 + A 2 + A 3 + A 4 = 1;

[0075] Based on the image quality evaluation model, for the historical appearance image HAI iEvaluate the image quality; based on the historical appearance image HAI i of the image quality value IQV i , the lighting device brightness value LFB i and the non - lighting device brightness value NLFB i Construct an image evaluation and environment association set, denoted as IEE = {(LFB i , NLFB i , IQV i )|i ∈ [1, N]}; Based on the image evaluation and environment association set, construct an image evaluation and environment association model: z = α 1 ×x + α 2 ×y + α 3 ; where α 1 , α 2 and α 3 represent fitting coefficients, x represents the independent variable of the lighting device brightness value, y represents the independent variable of the non - lighting device brightness value, z represents the dependent variable of the image quality value, and use the least - squares method to calculate and solve α 1 , α 2 and α 3 in the image evaluation and environment association model.

[0076] S3: Based on the image quality evaluation model, evaluate the image quality of the real - time appearance image of the battery, and adjust the lighting device parameters in real time according to the image quality evaluation result and the image evaluation and environment association model;

[0077] According to the above - mentioned solution, in step S3, calculate the evaluation image quality parameters of the real - time appearance image of the battery respectively, and normalize each evaluation image quality parameter; substitute the normalized evaluation image quality parameters into the image quality evaluation model to evaluate the image quality of the real - time appearance image of the battery;

[0078] Set an image quality value threshold for the image quality value. If the image quality value of the real - time appearance image of the battery is greater than or equal to the image quality value threshold, directly jump to step S4;

[0079] If the image quality value of the real - time appearance image of the battery is less than the image quality value threshold, substitute the image quality value threshold and the non - lighting device brightness value into the fitted image evaluation and environment association model to calculate the lighting device brightness value, and adjust the lighting device parameters to reach the lighting device brightness value.

[0080] S4: After adjusting the lighting device parameters, obtain the real - time appearance image of the battery again, compare the real - time appearance image of the battery with the battery appearance comparison image, and determine the abnormal area in the real - time appearance image of the battery;

[0081] According to the above solution, in step S4, after adjusting the parameters of the lighting device, the real-time appearance image of the battery is acquired again. When the real-time appearance image of the battery and the comparison image of the battery appearance are collected, the pose information of the battery, the relative position between the battery and the camera, and the focal length of the camera are the same;

[0082] Compare the real-time appearance image of the battery that has been subjected to image grayscale processing with the comparison image of the battery appearance, including:

[0083] Split the real-time appearance image of the battery and the comparison image of the battery appearance into pixel blocks of size P×Q. Calculate the pixel value similarity PVS between the pixel blocks of the real-time appearance image of the battery and the corresponding pixel blocks of the comparison image of the battery appearance. The specific calculation formula is as follows:

[0084]

[0085] Where RPV (p，q) represents the pixel value when the pixel block coordinates in the real-time appearance image of the battery are (p, q), and CPV (p，q) represents the pixel value when the pixel block coordinates in the comparison image of the battery appearance are (p, q). Determine the abnormal area in the real-time appearance image of the battery, where p ∈ [1, P] and q ∈ [1, Q];

[0086] Set a similarity threshold for the pixel value similarity of the pixel blocks. Extract all the pixel blocks whose extracted pixel value similarity is greater than the similarity threshold, and arrange the pixel blocks in the original arrangement order of the real-time appearance image of the battery, which is recorded as the abnormal area of the real-time appearance image of the battery.

[0087] S5: Generate an abnormal detection result based on the abnormal area comparison database, and give a prompt and alarm to the user.

[0088] According to the above solution, in step S5, based on the abnormal area comparison defect database, the database includes images of common abnormal phenomena of the battery;

[0089] Generate the specific abnormal type of the abnormal area, and give different levels of prompts and alarms set by the system to the user according to the specific abnormal type.

[0090] Embodiment 2: Please refer to Figure 2 , The present invention provides a technical solution: A battery appearance abnormality detection system based on visual recognition, the system includes a battery data acquisition module, a database, an appearance comparison image generation module, an evaluation association model construction module, a parameter adjustment module, and an abnormality comparison and prompt module;

[0091] The battery data acquisition module uses a camera to capture the appearance image of the battery, and a sensor to collect the environmental information of the battery; the database is used to receive and store battery data;

[0092] The appearance comparison image generation module is used to extract the basic information of the battery in the database, construct the three-dimensional appearance model of the battery, and generate the battery appearance comparison image according to the environmental information and the three-dimensional appearance model;

[0093] The evaluation association model construction module is used to obtain the historical appearance images and historical environmental information of the battery, analyze the historical appearance images and historical environmental information of the battery, construct an image quality evaluation model based on the historical appearance images, and construct an image evaluation and environment association model according to the image quality evaluation results of the historical appearance images and the historical environmental information;

[0094] The parameter adjustment module uses the image quality evaluation model to evaluate the real-time appearance image of the battery, calculates the brightness value of the lighting device using the environment association model, and adjusts the parameters of the lighting device in real time;

[0095] The anomaly comparison and prompt module is used to compare the real-time appearance image of the battery with the battery appearance comparison image, determine the abnormal area in the real-time appearance image of the battery, generate an anomaly detection result based on the abnormal area comparison database, and prompt and alarm the user.

[0096] According to the above solution, the appearance comparison image generation module includes a three-dimensional model and mapping unit and a comparison image generation unit;

[0097] The three-dimensional model and mapping unit creates a three-dimensional appearance model of the battery using three-dimensional modeling software according to the specification information of the battery; creates a three-dimensional coordinate system based on the three-dimensional appearance model, and maps the camera to the mapping point in the three-dimensional coordinate system based on the battery pose information and the relative position between the battery and the camera;

[0098] The comparison image generation unit takes the mapping point of the camera in the three-dimensional coordinate system as the starting point, and generates a shooting range ray according to the shooting range of the camera; generates a battery appearance comparison image based on the shooting range ray, the camera shooting parameters, and the three-dimensional appearance model.

[0099] According to the above solution, the evaluation association model construction module includes an image quality evaluation unit and an image evaluation and environment association unit;

[0100] The image quality evaluation unit is used to calculate the evaluation image quality parameters of the historical appearance images, and construct an image quality evaluation model based on the evaluation image quality parameters of the historical appearance images;

[0101] The image evaluation and environment association unit constructs an image evaluation and environment association set based on the image quality value, the brightness value of the lighting device, and the brightness value of the non-lighting device, and constructs an image evaluation and environment association model based on the image evaluation and environment association set.

[0102] According to the above solution, the anomaly comparison and prompt module includes an image anomaly comparison unit and an alarm prompt unit;

[0103] The image anomaly comparison unit is used to split the real-time appearance image of the battery and the appearance comparison image of the battery into pixel blocks of the same size, perform pixel value similarity on the two images, and generate an anomaly area;

[0104] The alarm prompt unit is used to compare the anomaly area with the defect database, generate the specific anomaly type of the anomaly area, and give different levels of prompts and alarms set by the system to the user according to the specific anomaly type.

[0105] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0106] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A battery appearance abnormality detection method based on visual recognition, characterized in that: The method comprises the following steps: S1: Use a camera to capture the appearance image of the battery, use a sensor to collect the environmental information of the battery, upload the basic information, appearance image and environmental information of the battery to a database, extract the basic information of the battery in the database, build a three-dimensional appearance model of the battery, and generate a battery appearance comparison image based on the environmental information and the three-dimensional appearance model; S2: Obtain historical appearance images and historical environmental information of the battery, parse the historical appearance images and historical environmental information of the battery, build an image quality assessment model based on the historical appearance images, and build an image assessment and environment association model based on the image quality assessment results of the historical appearance images and the historical environmental information; S3: Perform image quality assessment on the real-time appearance image of the battery based on the image quality assessment model, and adjust the lighting equipment parameters in real time according to the image quality assessment result and the image assessment and environment association model; S4: After adjusting the lighting equipment parameters, the real-time appearance image of the battery is acquired again, and the real-time appearance image of the battery acquired again is compared with the battery appearance comparison image to determine the abnormal area in the real-time appearance image of the battery; S5: Generate anomaly detection results based on the abnormal area comparison database, and give prompts and alarms to users; In step S2, the historical appearance images collected by a camera are subjected to the same rule of image grayscale processing, and then combined to generate a historical appearance image set, denoted as HAI={HAI i |i∈[1,N]}, where HAI i represents the historical appearance image with sequence number i in the historical appearance image set, and N represents the total number of historical appearance images in the historical appearance image set; For historical appearance images HAI i , respectively calculate the evaluation image quality parameters of the historical appearance images, and normalize each evaluation image quality parameter; the evaluation image quality parameter includes the Laplace transform value SH i , standard deviation of the histogram CO i , the average value of pixels BR i and peak signal-to-noise ratio NO i ; Construct an image quality assessment model based on the assessed image quality parameters of historical appearance images: IQV i =A1×SH i +A2×CO i +A3×BR i +A4×NO i ; IQV i Represents the historical appearance image HAI i A1, A2, A3 and A4 are the Laplace transform value, the standard deviation of the histogram, the average value of the pixel and the weight value of the peak signal-to-noise ratio of the historical appearance image, where A1+A2+A3+A4=1; HAI of historical appearance images based on image quality assessment model i Assess image quality; based on historical appearance image HAI i Image Quality Value IQV i , Lighting equipment brightness value LFB i and non-lighting equipment brightness value NLFB i Construct an image evaluation and environment association set, denoted as IEE={(LFB i ,NLFB i , IQV i )|i∈[1,N]}; Based on the image evaluation and environment association set, an image evaluation and environment association model is constructed: z=α1×x+α2×y+α3; where α1, α2 and α3 represent fitting coefficients, x represents the independent variable of the brightness value of the lighting equipment, y represents the independent variable of the brightness value of the non-lighting equipment, and z represents the dependent variable of the image quality value. The least squares method is used to calculate and solve α1, α2 and α3 in the image evaluation and environment association model; In step S3, the evaluation image quality parameters of the real-time appearance image of the battery are calculated respectively, and each evaluation image quality parameter is normalized; the evaluation image quality parameters after normalization are substituted into the image quality evaluation model to perform image quality evaluation on the real-time appearance image of the battery; An image quality value threshold is set for the image quality value. If the image quality value of the battery real-time appearance image is greater than or equal to the image quality value threshold, the battery real-time appearance image is directly compared with the battery appearance comparison image to determine the abnormal area in the battery real-time appearance image; If the image quality value of the real-time appearance image of the battery is less than the image quality value threshold, the image quality value threshold and the brightness value of the non-lighting device are substituted into the fitted image evaluation and environment association model, the brightness value of the lighting device is calculated, and the lighting device parameters are adjusted to achieve the lighting device brightness value.

2. The method for detecting battery appearance abnormality based on visual recognition according to claim 1, characterized in that: The basic information of the battery in step S1 includes identification information, type information and specification information of the battery; the identification information corresponds to the battery one by one, the identification information includes the label information of the battery, and all information of the corresponding battery can be found and extracted from the database according to the label information; the specification information includes the model information and standard appearance information of the battery; the standard appearance information does not have defects and quality problems; The environmental information includes light information, battery posture information, and the relative position of the battery and the camera; the light information includes the light value of the lighting device and the light value of the non-lighting device; A 3D appearance model of the battery is created using 3D modeling software according to the specification information of the battery; a 3D coordinate system is created based on the 3D appearance model, and the camera is mapped to a mapping point in the 3D coordinate system based on the battery posture information and the relative position between the battery and the camera; Taking the mapping point of the camera in the three-dimensional coordinate system as the starting point, a shooting range ray is generated according to the shooting range of the camera; a battery appearance comparison image is generated based on the shooting range ray, the camera shooting parameters and the three-dimensional appearance model, and the original color battery appearance comparison image is grayed out.

3. The method for detecting battery appearance abnormality based on visual recognition according to claim 2, characterized in that: In step S4, after adjusting the lighting device parameters, the real-time appearance image of the battery is acquired again, wherein the real-time appearance image of the battery and the battery appearance comparison image have the same battery posture information, the relative position between the battery and the camera, and the focal length of the camera when the real-time appearance image of the battery and the battery appearance comparison image are acquired; Compare the real-time appearance image of the battery that has been subjected to image grayscale processing with the battery appearance comparison image, including: The real-time battery appearance image and the battery appearance comparison image are split into pixel blocks of size P×Q. The pixel blocks of the real-time battery appearance image and the corresponding pixel blocks of the battery appearance comparison image are used to calculate the pixel value similarity PVS. The specific calculation formula is as follows: PVS=[∑Q q=1∑P p=1(RPV （p，q） -CPV （p，q） )]÷(P×Q); RPV （p，q） Indicates the pixel value when the pixel block coordinates are (p, q) in the real-time appearance image of the battery, CPV （p，q） Represents the pixel value when the pixel block coordinates in the battery appearance comparison image are (p, q), and determines the abnormal area in the real-time appearance image of the battery, p∈[1, P], q∈[1, Q]; A similarity threshold is set for the pixel value similarity of the pixel blocks, all pixel blocks whose pixel value similarity is less than the similarity threshold are extracted, and all the extracted pixel blocks are arranged in the order of the original battery real-time appearance image and recorded as abnormal areas of the battery real-time appearance image.

4. The method for detecting battery appearance abnormality based on visual recognition according to claim 3, characterized in that: In step S5, a defect database is compared based on the abnormal area, wherein the database includes images of common abnormal phenomena of batteries; Generate specific abnormal types for abnormal areas, and provide users with different levels of prompts and alarms set by the system according to the specific abnormal types.

5. A battery appearance abnormality detection system based on visual recognition, the system is applied to a battery appearance abnormality detection method based on visual recognition according to any one of claims 1 to 4, characterized in that: The system includes a battery data acquisition module, a database, an appearance comparison image generation module, an evaluation association model construction module, a parameter adjustment module and an abnormality comparison and prompt module; The battery data acquisition module uses a camera to capture the appearance image of the battery, and the sensor collects the environmental information of the battery; the database is used to receive and store battery data; The appearance comparison image generation module is used to extract basic information of the battery in the database, construct a three-dimensional appearance model of the battery, and generate a battery appearance comparison image according to the environmental information and the three-dimensional appearance model; The evaluation association model construction module is used to obtain the historical appearance images and historical environmental information of the battery, parse the historical appearance images and historical environmental information of the battery, construct an image quality evaluation model based on the historical appearance images, and construct an image evaluation and environmental association model based on the image quality evaluation results of the historical appearance images and the historical environmental information; the evaluation association model construction module includes an image quality evaluation unit and an image evaluation and environmental association unit; the image quality evaluation unit is used to calculate the evaluation image quality parameters of the historical appearance images, and construct an image quality evaluation model based on the evaluation image quality parameters of the historical appearance images; the image evaluation and environmental association unit constructs an image evaluation and environmental association set based on the image quality value, the brightness value of the lighting device and the brightness value of the non-lighting device, and constructs an image evaluation and environmental association model based on the image evaluation and environmental association set; The parameter adjustment module uses an image quality assessment model to assess the real-time appearance image of the battery, uses the image assessment and environment association model to calculate the brightness value of the lighting device, and adjusts the lighting device parameters in real time; The abnormal comparison and prompt module is used to compare the real-time appearance image of the battery with the battery appearance comparison image, determine the abnormal area in the real-time appearance image of the battery, generate an abnormal detection result based on the abnormal area comparison database, and prompt and alarm the user.

6. The battery appearance abnormality detection system based on visual recognition according to claim 5 is characterized in that: The appearance contrast image generation module includes a three-dimensional model and mapping unit and a contrast image generation unit; The three-dimensional model and mapping unit creates a three-dimensional appearance model of the battery using three-dimensional modeling software according to the specification information of the battery; creates a three-dimensional coordinate system based on the three-dimensional appearance model, and maps the camera to a mapping point in the three-dimensional coordinate system based on the battery posture information and the relative position between the battery and the camera; The contrast image generation unit takes the mapping point of the camera in the three-dimensional coordinate system as the starting point, generates a shooting range ray according to the shooting range of the camera, and generates a battery appearance contrast image based on the shooting range ray, the camera shooting parameters and the three-dimensional appearance model.

7. The battery appearance abnormality detection system based on visual recognition according to claim 5 is characterized in that: The abnormality comparison and prompting module includes an image abnormality comparison unit and an alarm prompting unit; The image abnormality comparison unit is used to split the battery real-time appearance image and the battery appearance comparison image into pixel blocks of the same size, perform pixel value similarity analysis on the two images, and generate an abnormal area; The alarm prompt unit is used to compare the abnormal area with the abnormal area comparison database, generate a specific abnormal type of the abnormal area, and provide the user with different levels of prompts and alarms set by the system according to the specific abnormal type.

Images