Battery recycling method and system based on machine vision

Abstract

The invention provides a battery recycling method and system based on machine vision. The method comprises the following steps: acquiring sample appearance data of each sample storage battery; further determining a noise interference domain and a damage distribution domain of the corresponding sample storage battery, and obtaining a confidence defect feature of each sample storage battery according to each noise interference domain and the corresponding damage distribution domain; collecting to-be-detected battery data, and extracting appearance characteristics of a target waste storage battery from the to-be-detected battery data; according to the appearance characteristics and the confidence defect characteristics of each sample storage battery, determining the collaborative similarity of the shell damage between the target waste storage battery and each sample storage battery; and determining the appearance defect grade value of the target waste storage battery according to all the collaborative similarities, and classifying and recycling the target waste storage battery according to the appearance defect grade value. By the adoption of the scheme, the accuracy of classified recycling of the waste batteries such as the power batteries or the lithium batteries of the waste new energy automobile can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of battery recycling, and more specifically, to a battery recycling method and system based on machine vision. Background Art

[0002] Battery recycling refers to the process of collecting, processing and reusing used batteries to reduce environmental pollution and waste of resources. With the widespread application of batteries in various electronic devices, new energy vehicles and other fields, the number of used batteries has increased dramatically. In order to ensure the steady development of battery recycling, it is necessary to classify and recycle used batteries and ensure the safety of batteries in each recycling process.

[0003] The existing method of classifying and recycling used batteries still mainly relies on human judgment. However, there is subjectivity in human identification of battery defects. Different people will make different judgments on the degree of damage of used batteries due to different work experience. It is difficult to reach a unified standard and the efficiency is low. Although the existing intelligent classification and recycling of used batteries can automatically identify, quickly classify and process used batteries through unified standards, it is difficult to identify the small and numerous defects on the shells of used batteries (such as slight dents and oxidative discoloration, etc.), which makes the error large when detecting defects in used batteries, resulting in low accuracy of classification and recycling of used batteries. Therefore, how to enhance the ability to identify defects in used battery shells during intelligent classification and recycling of used batteries, thereby improving the accuracy of classification and recycling of used batteries has become a difficult problem faced by the industry. Summary of the invention

[0004] The present application provides a battery recycling method and system based on machine vision, which can enhance the ability to identify defects in the shells of used batteries during the intelligent classification and recycling of used batteries, and improve the accuracy of classification and recycling of used batteries.

[0005] In a first aspect, the present application provides a battery recycling method based on machine vision, comprising: Acquire sample appearance data of each sample battery in the sample battery group; The noise interference domain of the corresponding sample battery is determined by the noise characteristics in the appearance data of each sample, and the damage distribution domain of the corresponding sample battery is determined by the damage characteristics in the appearance data of each sample, and then the defect contour of each sample battery is binarized and covered according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect feature of the shell of each sample battery; Collecting battery parameters of the target used battery to obtain test battery data, and extracting appearance features of the target used battery shell from the test battery data; Determine the collaborative similarity of shell damage between the target waste battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell; The appearance defect grade value of the target waste storage battery is determined according to all the collaborative similarities, and then the target waste storage batteries are classified and recycled according to the appearance defect grade value.

[0006] In some embodiments, determining the noise interference domain of the corresponding sample battery by using the noise characteristics in each sample appearance data specifically includes: Determine a noisy binary image of a corresponding sample battery according to each sample appearance data; Perform noise feature extraction on all noise binary images to obtain the noise feature of each sample battery; Determine the noise area of ​​each noise feature; The noise interference domain of the corresponding sample battery is determined according to each noise area.

[0007] In some embodiments, determining the damage distribution domain of the corresponding sample battery by the damage features in each sample appearance data specifically includes: Determine a damaged binary image of a corresponding sample battery according to each sample appearance data; Extract damage features from all damaged binary images to obtain damage features of each sample battery; Determine the damaged area for each damage feature; The damage distribution domain of the corresponding sample battery is determined according to each damage area.

[0008] In some embodiments, extracting the appearance features of the target waste battery shell in the battery data to be tested specifically includes: Obtaining an appearance display image of a target used storage battery from the battery data to be tested; Determine the appearance binary image of the target waste battery according to the appearance display image; Feature extraction is performed on the appearance binary image to obtain appearance features of the target waste battery shell.

[0009] In some embodiments, the defect profile of each sample battery is binarized and covered according to the noise interference domain and the corresponding damage distribution domain of each sample battery, and the confidence defect feature of each sample battery shell is obtained, which specifically includes: For each sample battery, obtain the noise interference domain and damage distribution domain of the sample battery; Determine a defect profile of a sample battery according to the noise interference domain and the damage distribution domain; Feature extraction is performed on the area covered by the defect contour in the binary image of the sample battery to obtain the confident defect feature of the sample battery shell, and then the confident defect feature of each sample battery shell is determined.

[0010] In some embodiments, determining the collaborative similarity of shell damage between the target used battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell specifically includes: Determine a feature similarity domain through the appearance features and the confidence defect features of each sample battery casing; The collaborative similarity of shell damage between the target used battery and each sample battery is determined based on the feature similarity domain.

[0011] In some embodiments, the target waste storage batteries are waste new energy vehicle power batteries or waste lithium batteries.

[0012] In a second aspect, the present application provides a battery recycling system based on machine vision, and the battery recycling system based on machine vision includes: A collection module, for acquiring sample appearance data of each sample battery in the sample battery group; The processing module determines the noise interference domain of the corresponding sample battery through the noise characteristics in the appearance data of each sample, determines the damage distribution domain of the corresponding sample battery through the damage characteristics in the appearance data of each sample, and then performs binary coverage on the defect contour of each sample battery according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect feature of the shell of each sample battery; The processing module is also used to collect battery parameters of the target waste battery, thereby obtaining the battery data to be tested, and extracting the appearance features of the shell of the target waste battery from the battery data to be tested; The processing module is also used to determine the collaborative similarity of shell damage between the target waste battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell; The execution module is used to determine the appearance defect grade value of the target waste battery according to all the collaborative similarities, and then classify and recycle the target waste battery according to the appearance defect grade value.

[0013] In a third aspect, the present application provides a computer device, comprising a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned machine vision-based battery recycling method.

[0014] In a fourth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned battery recycling method based on machine vision is implemented.

[0015] The technical solution provided by the embodiments disclosed in this application has the following beneficial effects: In the machine vision-based battery recycling method and system provided by the present application, firstly, the sample appearance data of each sample battery in the sample battery group is obtained; the noise interference domain of the corresponding sample battery is determined by the noise characteristics in each sample appearance data, and the damage distribution domain of the corresponding sample battery is determined by the damage characteristics in each sample appearance data, and then the defect contour of each sample battery is binarized and covered according to the noise interference domain of each sample battery and the corresponding damage distribution domain to obtain the confidence defect characteristics of the shell of each sample battery; the battery parameters of the target waste battery are collected to obtain the battery data to be tested, and the appearance characteristics of the shell of the target waste battery are extracted from the battery data to be tested; the collaborative similarity of the shell damage between the target waste battery and each sample battery is determined by the appearance characteristics and the confidence defect characteristics of the shell of each sample battery; the appearance defect grade value of the target waste battery is determined according to all the collaborative similarities, and then the target waste battery is classified and recycled according to the appearance defect grade value.

[0016] It can be seen that in the present application, the target waste batteries are classified and recycled by the appearance defect grade value; wherein, first, the sample appearance data of each sample battery in the sample battery group is obtained; the confidence defect features of the shell of each sample battery are determined by all the sample appearance data, and a feature sequence used to describe the defects of the sample battery shell can be obtained, and the feature sequence may include the contour circumference, contour area, aspect ratio, convex hull area and circularity of the defect, and then a detailed confidence defect feature database can be established, which is helpful to accurately identify the appearance defects of the target waste batteries when detecting them, thereby improving the recognition accuracy; secondly, the battery parameters of the target waste batteries are collected, and then the battery data to be tested is obtained, and the appearance features of the target waste battery shell are extracted from the battery data to be tested; then, the appearance features and each The confidence defect feature of the sample battery shell determines the collaborative similarity of the shell damage between the target waste battery and each sample battery, and a numerical value describing the similarity between the target waste battery shell and the sample battery shell can be obtained. By comparing the similarity between the target waste battery and the sample battery shell, the degree of damage between the two can be accurately matched, avoiding the error that may be caused by a single feature, thereby helping to identify small and numerous defects on the waste battery shell (such as slight dents and oxidation discoloration, etc.); finally, the appearance defect grade value of the target waste battery is determined according to all the collaborative similarities, and then the target waste battery is classified and recycled according to the appearance defect grade value; in summary, the present application can enhance the ability to recognize defects in waste battery shells in the process of intelligent waste battery classification and recycling, thereby improving the accuracy of waste battery classification and recycling. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is an exemplary flow chart of a battery recycling method based on machine vision according to some embodiments of the present application; Figure 2 is an exemplary flow chart of determining a noise interference domain according to some embodiments of the present application; Figure 3 is an exemplary flow chart for determining appearance defect grade values ​​according to some embodiments of the present application; Figure 4 is a schematic diagram of exemplary hardware and / or software of a battery recycling system based on machine vision according to some embodiments of the present application; Figure 5 It is a schematic diagram of the structure of a computer device for implementing a battery recycling method based on machine vision according to some embodiments of the present application. DETAILED DESCRIPTION

[0018] The embodiment of the present application provides a battery recycling method and system based on machine vision, the core of which is to obtain sample appearance data of each sample battery in a sample battery group; then determine the noise interference domain and damage distribution domain of the corresponding sample battery, and obtain the confidence defect feature of each sample battery according to each noise interference domain and the corresponding damage distribution domain; collect test battery data, and extract the appearance features of the target waste battery from the test battery data; determine the collaborative similarity of shell damage between the target waste battery and each sample battery through the appearance features and the confidence defect features of each sample battery; determine the appearance defect grade value of the target waste battery according to all the collaborative similarities, and then classify and recycle the target waste battery according to the appearance defect grade value, which can enhance the recognition ability of waste battery shell defects in the intelligent waste battery classification and recycling process, thereby improving the accuracy of waste battery classification and recycling.

[0019] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. Figure 1 , which is an exemplary flow chart of a battery recycling method based on machine vision according to some embodiments of the present application. The battery recycling method 100 based on machine vision mainly includes the following steps: In step 101, sample appearance data of each sample storage battery in a sample storage battery group is obtained.

[0020] It should be noted that the sample appearance data in the present application is an appearance display image of the sample battery. As a preferred embodiment, the sample appearance data of each sample battery in the sample battery group in the present application can be obtained in the following manner, namely: first use the robot's built-in camera to photograph the outer shell of each sample battery in the sample battery group, obtain the appearance display image of the outer shell of each sample battery, and then use each appearance display image as the sample appearance data of the corresponding sample battery, wherein the sample battery group is a collection of sample batteries with different types of damage to the outer shell, which may include sample batteries with deformed outer shells, sample batteries with damaged outer shells, sample batteries with corroded outer shells, sample batteries with traces of fire on the outer shells, and sample batteries with traces of water immersion on the outer shells. In addition, the robot used can be selected according to the specific situation. For example, the JAKAZu5 robot is used in this application, and the type of robot is not specifically limited here.

[0021] In step 102, the noise interference domain of the corresponding sample battery is determined by the noise characteristics in the appearance data of each sample, and the damage distribution domain of the corresponding sample battery is determined by the damage characteristics in the appearance data of each sample. Then, the binary image of each sample battery is covered and checked according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect characteristics of the shell of each sample battery.

[0022] In some embodiments, reference Figure 2 , which is an exemplary flow chart of determining the noise interference domain according to some embodiments of the present application. In the present application, the noise interference domain of the corresponding sample battery is determined by the noise characteristics in each sample appearance data, which can be implemented by the following steps: In step 1021, a noise binary image of a corresponding sample battery is determined according to each sample appearance data; In step 1022, noise features are extracted from all noise binary images to obtain noise features of each sample battery; In step 1023, the noise area of ​​each noise feature is determined; In step 1024, the noise interference domain of the corresponding sample battery is determined according to each noise area.

[0023] It should be noted that, in the present application, the noise interference domain is a noise area where noise exists in a binary image, and the noise in the binary image can be removed by the noise interference domain; the noise binary image refers to a grayscale image in which each pixel has only two possible values ​​of black (pixel value 0) or white (pixel value 255) after processing the appearance display image of the sample battery; noise feature extraction is a means of extracting noise features in an image, and the noise feature is a set of all noise areas in a binary image; the noise area represents a parameter of the degree of noise proportion in a binary image, and the larger the noise area, the greater the proportion of noise in the corresponding binary image.

[0024] In specific implementation, determining the noise binary image of the corresponding sample battery according to each sample appearance data can be achieved in the following manner, namely: first select a sample battery, obtain the sample appearance data of the sample battery, obtain the RGB values ​​of all pixels in the appearance display image corresponding to the sample appearance data, then extract the components of the RGB value of each pixel on the red, green and blue channels, and add 0.299 times the component on the red channel, 0.587 times the component on the green channel and 0.144 times the component on the blue channel, and the obtained sum is used as the grayscale value of the corresponding pixel, and then the grayscale values ​​of all pixels are combined into an image, and the image is used as the grayscale image corresponding to the sample appearance data, and then each pixel in the grayscale image whose pixel value is greater than a preset fixed threshold (the default is 128) is set to white (pixel value is 255), and each pixel whose pixel value is less than the preset fixed threshold is set to black (pixel value is 0), thereby obtaining a new image, and then the new image is used as the noise binary image of the sample battery, and finally the noise binary images of the remaining sample batteries are determined.

[0025] It should be noted that in the present application, because the target shell of the used battery is photographed indoors, the lighting conditions are stable and consistent, and the fixed threshold can be directly set to 128, the middle value between the white pixel value (255) and the black pixel value (0). In other embodiments, for example, when the shell of the target used battery is photographed outdoors and the lighting conditions vary with the sunshine time, the histogram analysis method in the prior art can also be used to set the fixed threshold in real time.

[0026] In addition, in the specific implementation, noise features are extracted from all noise binary images, and the noise features of each sample battery can be obtained in the following way, namely: select a noise binary image of a sample battery, first traverse each pixel in the noise binary image from left to right and from top to bottom, if the pixel value of the pixel is 255, then check the adjacent pixels above and to the left of the pixel, if the adjacent pixels of the pixel have been marked, then give the pixel the same label as the adjacent pixels, if the adjacent pixels of the pixel have not been marked, then The pixel is assigned a new label (for example, the pixel is assigned a label of 1), and then each pixel in the noisy binary image is traversed from left to right and from top to bottom again, and the pixels with the same label are connected to obtain multiple connected areas, and the total number of pixels in each connected area is counted, and the total number of pixels counted is used as the area of ​​the connected area, and then an area threshold is preset, and the set of all connected areas with an area less than the area threshold is used as the noise feature of the noisy binary image. Finally, the above steps are repeated to continue to obtain the noise binary image of the remaining sample batteries. The noise feature of the image is extracted, that is, the noise feature extraction is performed on all the noise binary images; wherein, the value of the area threshold can be preset according to the specific situation. For example, in the present application, considering that a certain accuracy will be lost when the image is binarized, resulting in the area of ​​the noise area in the image being enlarged, the area threshold is preset to 5, and only the connected area with an area less than 5 will be set as the noise area. In other embodiments, the area threshold can also be preset to other values ​​according to different image processing methods, which is not specifically limited here; determining the noise area of ​​each noise feature can be achieved in the following manner, namely: taking the set of area areas of all noise areas in each noise feature as the noise area of ​​the noise feature; determining the noise interference domain of the corresponding sample battery according to each noise area can be achieved in the following manner, namely: first calculating the average value of all area areas (i.e., the area of ​​noise) in the noise area, and rounding down the obtained average value to obtain an integer A, then constructing a matrix of size A, in which the values ​​of all elements in the matrix are 1, and finally taking the obtained matrix as the noise interference domain of the sample battery, and finally repeating the above steps to obtain the noise interference domain of the remaining sample batteries.

[0027] In some embodiments, determining the damage distribution domain of the corresponding sample battery by the damage features in each sample appearance data can be achieved by using the following steps: Determine a damaged binary image of a corresponding sample battery according to each sample appearance data; Extract damage features from all damaged binary images to obtain damage features of each sample battery; Determine the damaged area for each damage feature; The damage distribution domain of the corresponding sample battery is determined according to each damage area.

[0028] It should be noted that in the present application, the damage distribution domain is the damaged area covered by small defects in the binary image, and the small defects in the binary image can be repaired through this damage distribution domain; the damaged binary image refers to a grayscale image in which each pixel value obtained after processing the appearance display image of the sample battery has only two possible values: black (pixel value is 0) or white (pixel value is 255). Damage feature extraction is a means of extracting damage features in an image. The damage feature is a set that contains all damaged areas in the binary image.

[0029] In specific implementation, determining the damaged binary image of the corresponding sample battery according to each sample appearance data can be implemented in the following manner, namely: firstly, selecting a sample appearance data, obtaining the RGB values ​​of all pixels in the appearance display image corresponding to the sample appearance data, then extracting the components of the RGB value of each pixel on the red, green and blue channels, and adding 0.299 times the component on the red channel, 0.587 times the component on the green channel and 0.144 times the component on the blue channel, and the obtained sum is used as the gray value of the corresponding pixel, and then the gray values ​​of all pixels are combined into an image, and the image is used as the gray image corresponding to the sample appearance data, and then each pixel in the gray image whose pixel value is greater than a preset fixed threshold (the default is 128) is set to white (pixel value is 255), and each pixel whose pixel value is less than the preset fixed threshold is set to black (pixel value is 0), thereby obtaining a new image, and then the new image is used as the damaged binary image of the sample battery corresponding to the sample appearance data, and finally the damaged binary images of the sample batteries corresponding to the remaining sample appearance data are determined.

[0030] It should be noted that in the present application, because the target shell of the used battery is photographed indoors, the lighting conditions are stable and consistent, and the fixed threshold can be directly set to 128, the middle value between the white pixel value (255) and the black pixel value (0). In other embodiments, for example, when the shell of the target used battery is photographed outdoors and the lighting conditions vary with the sunshine time, the histogram analysis method in the prior art can also be used to set the fixed threshold in real time.

[0031] In addition, in the specific implementation, damage feature extraction is performed on all damaged binary images, and the damage feature of each sample battery can be obtained in the following manner, namely: select a damaged binary image of a sample battery, first traverse each pixel in the damaged binary image from left to right and from top to bottom, if the pixel value of the pixel is 255, then check the adjacent pixels above and to the left of the pixel, if the adjacent pixels of the pixel have been marked, then assign the pixel the same label as the adjacent pixels, if the adjacent pixels of the pixel have not been marked, then assign a new label to the pixel (for example, assign label 1 to the pixel), then traverse each pixel in the damaged binary image from left to right and from top to bottom again, connect the pixels with the same label to obtain multiple connected regions, and count the holes existing in each connected region. The number of holes is counted as the number of holes in the connected area, and then a hole threshold is preset. The connected areas with holes greater than the hole threshold are taken as damaged areas and formed into a set. The set is then taken as the damage feature of the damaged binary image. Finally, the above steps are repeated to continue to obtain the damage features of the damaged binary images of the remaining sample batteries, that is, to achieve damage feature extraction for all damaged binary images; wherein, the value of the hole threshold can be preset according to specific circumstances. For example, in the present application, considering that a certain accuracy will be lost when binarizing the image, holes that do not actually exist may appear in the image, the hole threshold is preset to 3, and only areas with holes greater than 3 will be marked as damaged areas. In other embodiments, the hole threshold can also be preset to other values ​​according to different image processing methods, which are not specifically limited here.

[0032] It should be noted that the damaged area in the present application is the area of ​​all damaged regions in the damage feature, and the area of ​​the damaged region may be the total value of the pixels in the damaged region.

[0033] In specific implementation, the damage distribution domain of the corresponding sample battery can be determined according to each damage area in the following way, namely: first, the average area of ​​all regions in the damage area is calculated, and the average is rounded down to an integer B, and then a matrix of size B is constructed, in which all elements of the matrix have a value of 1, and finally the obtained matrix is ​​used as the damage distribution domain of the sample battery, and finally the damage distribution domains of the remaining sample batteries are determined.

[0034] In some embodiments, the defect profile of each sample battery is binarized and covered according to the noise interference domain and the corresponding damage distribution domain of each sample battery, and the confidence defect feature of each sample battery shell is obtained by the following steps: For each sample battery, obtain the noise interference domain and damage distribution domain of the sample battery; Determine a defect profile of a sample battery according to the noise interference domain and the damage distribution domain; Feature extraction is performed on the area covered by the defect contour in the binary image of the sample battery to obtain the confident defect feature of the sample battery shell, and then the confident defect feature of each sample battery shell is determined.

[0035] It should be noted that in the present application, the confidence defect feature is a feature sequence used to describe the defects of the sample battery shell, and the feature sequence may include the contour circumference, contour area, aspect ratio, convex hull area and circularity of the defect; the primary inspection image and the secondary inspection image are intermediate process images obtained when the binary image of the sample battery is covered for inspection, and have no practical significance; binary coverage is a means of extracting features from the area covered by the defect contour in the binary image of the sample battery, and the defect contour refers to the contour of all defective areas on the sample battery shell.

[0036] In the specific implementation, first, for each sample battery, the noise interference domain and damage distribution domain of the sample battery are obtained; secondly, the defect contour of the sample battery is determined according to the noise interference domain and the damage distribution domain; finally, the area covered by the defect contour in the binary image of the sample battery is subjected to feature extraction, and then the confidence defect feature of the outer shell of each sample battery is obtained. The following method can be used for the implementation, namely: firstly, a sample battery is selected to determine the defect contour of the sample battery, and then the contour perimeter of the defect contour (that is, the sum of the distances between each pair of adjacent pixel points on the defect contour), the contour area (that is, the total number of pixel points inside and on the defect area corresponding to the defect contour), and the aspect ratio (that is, the quotient obtained by dividing the width of the bounding box of the defect contour by the height of the bounding box, where the bounding box is A rectangle with the minimum area that completely surrounds the defect contour, the width of the bounding box is the total number of pixels between the right and left boundaries of the bounding box, and the height of the bounding box is the total number of pixels between the upper and lower boundaries of the bounding box), convex hull area (i.e., the area of ​​the minimum convex polygon that can completely surround the defect contour) and circularity (i.e., the quotient obtained by multiplying the contour area of ​​the defect contour by the square of the perimeter), and then taking the obtained contour perimeter, contour area, aspect ratio, convex hull area and circularity as the features of the defect contour, respectively, and forming a sequence in the order of obtaining, and taking the sequence as the confidence defect feature of the sample battery casing, and finally continuing to obtain the confidence defect features of the remaining sample battery casings; thereby achieving binary coverage of the defect contour of each sample battery.

[0037] In the above embodiment, the defect profile of the sample battery can be determined according to the noise interference domain and the damage distribution domain by using the following steps: Inspecting each pixel in the binary image of the sample battery according to the noise interference domain of the sample battery, and transforming the grayscale value of the corresponding pixel according to the inspection result, thereby obtaining a primary inspection image of the sample battery; Inspecting each pixel in the primary inspection image according to the damage distribution domain of the sample battery, and transforming the grayscale value of the corresponding pixel according to the inspection result, thereby obtaining a secondary inspection image of the sample battery; A defect profile of the sample battery is determined based on the secondary inspection image.

[0038] In a specific implementation, each pixel in the binary image of the sample battery is checked according to the noise interference domain of the sample battery, and the grayscale value of the corresponding pixel is transformed according to the inspection result, so as to obtain an inspection image of the sample battery. This can be implemented in the following way, namely: first determine the noise interference domain and the binary image of the sample battery, then align the center of the noise interference domain with each pixel in the binary image, if all the pixels covered by the noise interference domain are white, then the pixel remains white, if there is black among all the pixels covered by the noise interference domain, then the pixel is changed to black to obtain a new image, and finally the new image is used as the inspection image of the sample battery. Inspection image; inspecting each pixel in the primary inspection image according to the damage distribution domain of the sample battery, and transforming the grayscale value of the corresponding pixel according to the inspection result, so as to obtain a secondary inspection image of the sample battery, which can be achieved in the following manner, namely: first determining the damage distribution domain and the binary image of the sample battery, and then aligning the center of the damage distribution domain with each pixel in the binary image, if all the pixels covered by the damage distribution domain are white, then the pixel remains white, if there is black among all the pixels covered by the damage distribution domain, then the pixel is changed to black to obtain a new image, and finally the new image is used as the primary inspection image of the sample battery.

[0039] In addition, in the specific implementation, the defect outline of the sample battery can be determined based on the secondary inspection image in the following manner, namely: first traverse all pixel points in the secondary inspection image from top to bottom and from left to right, determine the connected area in the secondary inspection image, and use the connected area as the defect area of ​​the sample battery shell, and then use the outline surrounded by the outermost pixel points of the defect area as the defect outline of the sample battery.

[0040] In step 103, the battery parameters of the target used battery are collected to obtain the battery data to be tested, and the appearance features of the shell of the target used battery are extracted from the battery data to be tested.

[0041] It should be noted that in the present application, the target used battery may be a used new energy vehicle power battery or a used lithium battery, etc., which is not specifically limited here. The battery parameters may include the insulation resistance value, temperature value, cell voltage value, cell current value and appearance display diagram of the battery. In other embodiments, the battery parameters may also include other battery parameters related to the aging condition of the target used battery, which is not limited here. In specific implementation, existing tools (for example: insulation tester and JAKAZu5 robot) can be used to detect all battery parameters of the target used battery, and the collection of all battery parameters is used as the battery data to be tested. In some embodiments, extracting the appearance features of the target used battery shell from the battery data to be tested can be achieved by using the following steps: Obtaining an appearance display image of a target used storage battery from the battery data to be tested; Determine the appearance binary image of the target waste battery according to the appearance display image; Feature extraction is performed on the appearance binary image to obtain appearance features of the target waste battery shell.

[0042] It should be noted that, in the present application, the appearance feature is a characteristic quantity used to describe the defects of the target waste battery shell; the appearance binary image is a grayscale image used to describe the appearance color distribution in the target waste battery.

[0043] In specific implementation, determining the appearance binary image of the target used battery according to the appearance display image can be implemented in the following manner, namely: obtaining the RGB values ​​of all pixels in the appearance display image, then extracting the components of the RGB value of each pixel in the red, green and blue channels, and adding 0.299 times the component on the red channel, 0.587 times the component on the green channel and 0.144 times the component on the blue channel, and using the sum as the grayscale value of the corresponding pixel, and then forming an image with the grayscale values ​​of all pixels, and using the image as the grayscale image of the appearance display image, and then setting each pixel in the grayscale image whose pixel value is greater than a preset fixed threshold value to white (pixel value is 255), and setting each pixel whose pixel value is less than the preset fixed threshold value to black (pixel value is 0), thereby obtaining a new image, and then using the new image as the appearance binary image of the target used battery.

[0044] It should be noted that in the present application, because the target shell of the used battery is photographed indoors, the lighting conditions are stable and consistent, and the fixed threshold can be directly set to 128, the middle value between the white pixel value (255) and the black pixel value (0). In other embodiments, for example, when the shell of the target used battery is photographed outdoors and the lighting conditions vary with the sunshine time, the histogram analysis method in the prior art can also be used to set the fixed threshold in real time.

[0045] In addition, in a specific implementation, feature extraction is performed on the appearance binary image to obtain the appearance features of the target waste battery shell, which can be achieved in the following manner, namely: first, each pixel point in the appearance binary image is traversed from left to right and from top to bottom. If the pixel value of the pixel point is 255, the adjacent pixels above and to the left of the pixel point are checked. If the adjacent pixels of the pixel point have been marked, the pixel point is assigned the same label as the adjacent pixels. If the adjacent pixels of the pixel point have not been marked, a new label is assigned to the pixel point (for example, the pixel point is assigned a label of 1). Then, each pixel point in the appearance binary image is traversed from left to right and from top to bottom again, and the pixels with the same label are connected to obtain multiple connected areas, and the multiple connected areas obtained are respectively used as the defective areas of the target waste battery shell, and then the contour perimeter of the contour surrounded by the outermost pixels of each defective area is calculated (that is, each pair of adjacent pixels on the contour is ), contour area (i.e., the total number of pixels inside and on the contour of the defective area), aspect ratio (i.e., the quotient of the width of the bounding box of the contour divided by the height of the bounding box, where the bounding box is a rectangle with the minimum area that completely surrounds the contour, the width of the bounding box is the total number of pixels between the right and left boundaries of the bounding box, and the height of the bounding box is the total number of pixels between the upper and lower boundaries of the bounding box), convex hull area (i.e., the area of ​​the minimum convex polygon that can completely surround the contour) and circularity (i.e., the quotient obtained by multiplying the contour area of ​​the contour by the square of the perimeter), and then the contour perimeter, contour area, aspect ratio, convex hull area and circularity of each contour are combined into a sequence in the order of obtaining, and the obtained sequence is used as the feature of each defective area, and finally all features are sorted from small to large according to the contour area to obtain a feature sequence, and the feature sequence is used as the appearance feature of the target waste battery shell.

[0046] In step 104, the collaborative similarity of shell damage between the target used battery and each sample battery is determined through the appearance feature and the confidence defect feature of each sample battery shell.

[0047] In some embodiments, determining the collaborative similarity of shell damage between the target used battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell can be achieved by the following steps: Determine a feature similarity domain through the appearance features and the confidence defect features of each sample battery casing; The collaborative similarity of shell damage between the target used battery and each sample battery is determined based on the feature similarity domain.

[0048] It should be noted that, in the present application, collaborative similarity is a numerical value that describes the similarity of the shell damage between the target used battery shell and the sample battery shell. The larger the collaborative similarity, the more similar the shell damage between the target used battery shell and the sample battery shell. The feature similarity domain is a matrix that represents the similarity between the appearance features of the battery data to be tested and the confidence defect features of the sample battery shell.

[0049] In specific implementation, determining the feature similarity domain through the appearance feature and the confidence defect feature of each sample battery shell can be achieved in the following manner, namely: first, sorting the confidence defect feature of each sample battery shell from small to large according to the contour area, and then calculating the similarity between the feature of each defect area in the appearance feature and the confidence defect feature of each sample battery shell, and filling all the obtained similarities into a matrix to obtain a similarity matrix, wherein the element in the i-th row and j-th column of the similarity matrix represents the similarity between the feature of the i-th defect area in the appearance feature and the confidence defect feature of the j-th sample battery shell, and the similarity matrix is ​​used as the feature similarity domain, wherein the confidence defect feature of each defect area in the appearance feature is calculated. The similarity between the features of the domain and the confidence defect features of each sample battery shell can be achieved by calculating the dot product between the two features; determining the collaborative similarity of shell damage between the target used battery and each sample battery based on the feature similarity domain can be achieved in the following manner, namely: first select a sample battery, find the similarity between the confidence defect features of the sample battery shell and the features of all defective areas in the appearance features of the target used battery in the feature similarity domain, then average all the similarities, and use the obtained mean as the collaborative similarity of shell damage between the target used battery and the sample battery, and finally continue to determine the collaborative similarity of shell damage between the target used battery and the remaining sample batteries.

[0050] In step 105, the appearance defect grade value of the target waste battery is determined according to all the collaborative similarities, and then the target waste battery is classified and recycled according to the appearance defect grade value.

[0051] In some embodiments, reference Figure 3 , which is an exemplary flow chart of determining a label hierarchy path according to some embodiments of the present application. In the present application, determining the appearance defect level value of the target waste battery according to all collaborative similarities can be implemented by the following steps: In step 1051, the confidence cost of each sample battery in the sample battery group is determined by all the collaborative similarities; In step 1052, the appearance defect level value of the target used battery is determined according to all the confidence costs.

[0052] It should be noted that, in the present application, the appearance defect grade value is a parameter describing the degree of damage to the target used battery shell. The larger the appearance defect grade value, the more serious the damage to the target used battery shell. The confidence cost is a probability value describing whether the target used battery shell has the same defects as the sample battery shell. The larger the confidence cost, the greater the probability that the target used battery shell has the same defects as the sample battery shell.

[0053] In specific implementation, the confidence cost of each sample battery in the sample battery group is determined by all the collaborative similarities, which can be implemented in the following way, namely: the sum of the powers of the natural constants of all the collaborative similarities is taken as the similarity sum, and then a sample battery is selected in the sample battery group, the collaborative similarity corresponding to the sample battery is determined, and the collaborative similarity is taken as the power of the natural constant, and the obtained value is divided by the value of the similarity sum to obtain a quotient, and then the obtained quotient is used as the confidence cost of the sample battery, and finally the confidence cost of the remaining sample batteries in the sample battery group is determined; the appearance defect level value of the target waste battery is determined according to all the confidence costs, which can be implemented in the following way, namely: first a confidence threshold is preset, and then a value greater than the confidence threshold is selected from all the confidence costs. The confidence cost is determined, and the total number of the selected confidence costs is determined, and finally the total number is used as the appearance defect grade value of the appearance damage of the target used battery, wherein the value of the confidence threshold can be preset according to the actual situation. For example, in the present application, because the confidence cost is a probability value, the value range is between 0 and 1. The larger the confidence cost is, the greater the probability that the shell of the target used battery has the same defects as the sample battery shell. At the same time, considering the area of ​​the target used battery shell and the complexity of the defects, the confidence threshold can be preset to 0.7. In other embodiments, the confidence threshold can also be preset to other values ​​between 0 and 1. For example, when the area of ​​the used battery is small and the complexity of the defects is not high, the value of the confidence threshold can also be preset to 0.9. No specific limitation is made here.

[0054] In some embodiments, the classification and recycling of target waste batteries according to the appearance defect grade value can be achieved by the following steps: Determine the cell defect amount of the internal cell of the target waste battery based on the cell parameter data in the battery data to be tested; Determining a battery recycling grade of the target waste battery according to the battery cell defect amount and the appearance defect grade value; The target waste batteries are classified and recycled according to the battery recycling level.

[0055] It should be noted that in the present application, the battery recycling grade is the recycling status of the target waste battery obtained after determining the safety status of the target waste battery based on the appearance and battery operating condition of the target waste battery. When recycling the battery, the target waste batteries belonging to different battery recycling grades need to be treated differently in terms of operating requirements in packaging, transportation, warehousing and other links to ensure the safety of the recycling link; the cell defect amount is a parameter that describes the degree of damage to the internal cell of the target waste battery. The larger the cell defect amount, the more serious the degree of damage to the internal cell of the target waste battery; the cell parameter data refers to the specific values ​​of the key parameters that can affect the performance of the target waste battery cell. In some preferred embodiments, the cell parameter data may be the insulation resistance, temperature, cell voltage, and cell current of the battery. In other embodiments, the cell parameter data may also be other key parameters in the battery that can affect the performance of the target waste battery cell, which is not limited here.

[0056] Preferably, in some embodiments, determining the cell defect quantity of the internal cells of the target used battery based on the cell parameter data in the battery data to be tested can be achieved by using the following steps: Determining the deviation of each key parameter in the internal battery cell of the target waste battery according to the battery cell parameter data; The cell defect quantity of the internal cells of the target waste battery is determined through all the deviations.

[0057] It should be noted that, in the present application, the deviation is a parameter used to measure the degree of difference between the internal cells of the target used battery and the internal cells of the normal target used battery. The larger the deviation, the greater the difference in the value of the key parameter between the target used battery and the normal used battery, and the higher the defect level of the target used battery.

[0058] In specific implementation, the deviation of each key parameter in the internal battery cell of the target used battery can be determined according to the battery cell parameter data in the following manner, namely: firstly, a key parameter in the internal battery cell of the target used battery can be selected, and the standard value of the battery of the same model as the target used battery on the key parameter can be queried in the instruction manual of the target used battery, and then the parameter value corresponding to the key parameter is obtained from the battery cell parameter data, and the difference between the parameter value and the standard value is used as the deviation of the key parameter. If the standard value of the key parameter is an interval, the parameter value is compared with the standard value. The difference between the upper bounds of the intervals corresponding to the quasi-values ​​is taken as the deviation of the key parameter, and the above steps are repeated to continue to obtain the deviations of the remaining key parameters in the battery cell parameter data; determining the cell defect amount of the target battery cell inside the used battery through all the deviations can be achieved in the following way, that is, the average value of all the deviations can be taken as the cell defect amount of the target battery cell inside the used battery; determining the cell defect amount of the target battery cell inside the used battery through all the deviations can be achieved in the following way, that is, obtaining the average value of all the deviations, and taking the average value as the cell defect amount of the target battery cell inside the used battery.

[0059] In specific implementation, the battery recycling grade of the target waste battery is determined according to the battery cell defect amount and the appearance defect grade value, which can be implemented in the following manner, namely: if the battery cell defect amount and the appearance defect grade value of the target waste battery are both zero, the battery recycling grade of the target waste battery is set to grade A; if the battery cell defect amount of the target waste battery is zero, but the appearance defect grade value is not zero, the battery recycling grade of the target waste battery is set to grade B; if the battery cell defect amount and the appearance defect grade value of the target waste battery are both not zero, the battery recycling grade of the target waste battery is set to grade C; the classified recycling of the target waste batteries according to the battery recycling grade can be implemented in the following manner, namely: using a robot to package the target waste batteries according to the battery recycling grade, and The target used batteries are placed in a transport vehicle of the corresponding battery recycling level, waiting for transportation and later processing. The requirements for packaging and transportation of used target batteries of different battery recycling levels are different. For example, Class A used batteries should use anti-static and moisture-proof packaging materials to avoid short circuits and overheating. Extrusion and vibration should be avoided during transportation. Class B used batteries should be packaged in leak-proof and corrosion-resistant containers to avoid electrolyte leakage. During transportation, the batteries should be prevented from tipping over and damage to avoid electrolyte leakage. Each battery of Class C used batteries should be individually packaged with a plastic bag or plastic film, and placed in a leak-proof plastic box or metal box for separate storage. During transportation, the batteries should also be prevented from tipping over and damage to avoid electrolyte leakage. In addition, the robot can be selected according to the specific situation. For example, the JAKAZu5 robot is used in this application.

[0060] In addition, in another aspect of the present application, in some embodiments, the present application provides a battery recycling system based on machine vision, referring to Figure 4 , which is a schematic diagram of exemplary hardware and / or software of a battery recycling system based on machine vision according to some embodiments of the present application. The battery recycling system based on machine vision 400 includes: an acquisition module 401, a processing module 402 and an execution module 403, which are described as follows: Acquisition module 401, in this application, acquisition module 401 is mainly used to obtain sample appearance data of each sample battery in the sample battery group; Processing module 402, in the present application, the processing module 402 is mainly used to determine the noise interference domain of the corresponding sample battery through the noise characteristics in each sample appearance data, determine the damage distribution domain of the corresponding sample battery through the damage characteristics in each sample appearance data, and then perform binary coverage on the defect contour of each sample battery according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect feature of each sample battery shell; It should be noted that the processing module 402 in the present application is also used to collect battery parameters of the target waste battery, thereby obtaining the battery data to be tested, and extracting the appearance characteristics of the target waste battery shell from the battery data to be tested; In addition, it should be noted that the processing module 402 in the present application is also used to determine the collaborative similarity of the shell damage between the target waste battery and each sample battery through the appearance features and the confidence defect features of each sample battery shell; Execution module 403, in the present application, execution module 403 is mainly used to determine the appearance defect grade value of the target waste battery according to all collaborative similarities, and then classify and recycle the target waste battery according to the appearance defect grade value.

[0061] In addition, the present application also provides a computer device, which includes a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the above-mentioned machine vision-based battery recycling method.

[0062] In some embodiments, reference Figure 5 , which is a schematic diagram of the structure of a computer device for implementing a battery recycling method based on machine vision according to some embodiments of the present application. The battery recycling method based on machine vision in the above embodiments can be Figure 5 The computer device 500 shown in the figure is implemented, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503 and at least one communication interface 504.

[0063] The processor 501 may be a general-purpose central processing unit (CPU) or an application specific integrated circuit (ASIC).

[0064] The communication bus 502 may be used to transmit information between the above-mentioned components.

[0065] The memory 503 may be a read only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read only memory (EEPROM), a compact disc read only memory (CD ROM) or other optical disc storage, an optical disc storage (including a compressed optical disc, a laser disc, an optical disc, a digital versatile disc, a Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory 503 may exist independently and be connected to the processor 501 via the communication bus 502. The memory 503 may also be integrated with the processor 501.

[0066] The memory 503 is used to store the program code for executing the solution of the present application, and the execution is controlled by the processor 501. The processor 501 is used to execute the program code stored in the memory 503. The program code may include one or more software modules. The battery recycling method based on machine vision in the above embodiment can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0067] The communication interface 504 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0068] In a specific implementation, as an embodiment, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. The processor here may refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0069] The above-mentioned computer device may be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device may be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device or an embedded device. The embodiment of the present application does not limit the type of computer device.

[0070] In addition, the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned battery recycling method based on machine vision is implemented.

[0071] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0072] Obviously, those skilled in the art may make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A battery recycling method based on machine vision, It is characterized in that The steps include: Acquire sample appearance data of each sample battery in the sample battery group; The noise interference domain of the corresponding sample battery is determined by the noise characteristics in the appearance data of each sample, and the damage distribution domain of the corresponding sample battery is determined by the damage characteristics in the appearance data of each sample, and then the defect contour of each sample battery is binarized and covered according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect feature of the shell of each sample battery; Collecting battery parameters of the target used battery to obtain test battery data, and extracting appearance features of the target used battery shell from the test battery data; Determine the collaborative similarity of shell damage between the target waste battery and each sample battery through the appearance features and the confidence defect features of each sample battery shell; The appearance defect grade value of the target waste storage battery is determined according to all the collaborative similarities, and then the target waste storage batteries are classified and recycled according to the appearance defect grade value.

2. The method according to claim 1, It is characterized in that The noise interference domain of the corresponding sample battery is determined by the noise characteristics in each sample appearance data, specifically including: Determine a noise binary image of a corresponding sample battery according to each sample appearance data; Perform noise feature extraction on all noise binary images to obtain the noise feature of each sample battery; Determine the noise area of ​​each noise feature; The noise interference domain of the corresponding sample battery is determined according to each noise area.

3. The method according to claim 1, It is characterized in that The damage distribution domain of the corresponding sample battery is determined by the damage characteristics in the appearance data of each sample, specifically including: Determine a damaged binary image of a corresponding sample battery according to each sample appearance data; Extract damage features from all damaged binary images to obtain damage features of each sample battery; Determine the damaged area for each damage feature; The damage distribution domain of the corresponding sample battery is determined according to each damage area.

4. The method according to claim 1, It is characterized in that The defect profile of each sample battery is binarized and covered according to the noise interference domain and the corresponding damage distribution domain of each sample battery, and the confidence defect features of each sample battery shell are obtained, which specifically include: For each sample battery, obtain the noise interference domain and damage distribution domain of the sample battery; Determine a defect profile of a sample battery according to the noise interference domain and the damage distribution domain; Feature extraction is performed on the area covered by the defect contour in the binary image of the sample battery to obtain the confident defect feature of the sample battery shell, and then the confident defect feature of each sample battery shell is determined.

5. The method according to claim 1, It is characterized in that The appearance features of the target waste battery shell extracted from the battery data to be tested specifically include: Obtaining an appearance display image of a target used storage battery from the battery data to be tested; Determine the appearance binary image of the target waste battery according to the appearance display image; Feature extraction is performed on the appearance binary image to obtain appearance features of the target waste battery shell.

6. The method according to claim 1, It is characterized in that Determining the coordinated similarity of shell damage between the target waste battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell specifically includes: Determine a feature similarity domain through the appearance features and the confidence defect features of each sample battery casing; The collaborative similarity of shell damage between the target used battery and each sample battery is determined based on the feature similarity domain.

7. The method according to claim 1, It is characterized in that The target waste batteries are waste new energy vehicle power batteries or waste lithium batteries.

8. A battery recycling system based on machine vision, It is characterized in that include: A collection module, for acquiring sample appearance data of each sample battery in the sample battery group; The processing module determines the noise interference domain of the corresponding sample battery through the noise characteristics in the appearance data of each sample, determines the damage distribution domain of the corresponding sample battery through the damage characteristics in the appearance data of each sample, and then performs binary coverage on the defect contour of each sample battery according to the noise interference domain and the corresponding damage distribution domain of each sample battery to obtain the confidence defect characteristics of the shell of each sample battery; The processing module is also used to collect battery parameters of the target waste battery, thereby obtaining the battery data to be tested, and extracting the appearance features of the shell of the target waste battery from the battery data to be tested; The processing module is also used to determine the collaborative similarity of shell damage between the target waste battery and each sample battery through the appearance feature and the confidence defect feature of each sample battery shell; The execution module is used to determine the appearance defect grade value of the target waste battery according to all the collaborative similarities, and then classify and recycle the target waste battery according to the appearance defect grade value.

9. A computer device, It is characterized in that The computer device includes a memory and a processor, wherein the memory stores codes, and the processor is configured to obtain the codes and execute the battery recycling method based on machine vision according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program. It is characterized in that When the computer program is executed by a processor, the battery recycling method based on machine vision as described in any one of claims 1 to 7 is implemented.

Images