Battery blue film defect detection method, device and equipment and storage medium

Abstract

The invention discloses a battery blue film defect detection method, device and equipment and a storage medium, and relates to the technical field of automation. The method comprises the following steps: scanning a to-be-detected battery under a preset light source through a preset camera to obtain battery blue film images of a plurality of cameras under different visual angles; the preset camera is perpendicular to the blue film surface of the outer surface of the to-be-detected battery; converting the battery blue film image to an HSV color space to obtain a blue film background area; and detecting the blue film background area based on the deep learning model to obtain blue film defect information. In this way, the battery blue film image on the surface of the battery blue film to be detected is converted from the RGB space to the HSV space in combination with color space conversion, the blue background area is extracted, and the contrast ratio of the defect and the background is improved in a targeted mode; and meanwhile, an image processing algorithm is optimized, the surface texture noise of the blue film is effectively inhibited while defect features are enhanced, and the misjudgment rate is reduced.

Description

Technical Field

[0001] This invention relates to the field of battery technology, and in particular to a method, apparatus, device, and storage medium for detecting defects in the blue film of batteries. Background Technology

[0002] As safety standards for power batteries rise, the industry's requirements for the detection accuracy of blue film defects have evolved from 0.2mm to 0.05mm. However, existing technologies face triple limitations in three-dimensional information acquisition, spectral matching, and algorithm adaptability, leading to a dual dilemma for production lines: "over-detection waste" and "missed detection risk." Missed, minute wrinkles in the blue film may cause it to rupture during battery cycling, potentially leading to a short circuit. Summary of the Invention

[0003] In view of this, the purpose of the present invention is to overcome the shortcomings of the prior art and provide a method, apparatus, device and storage medium for detecting blue film defects in batteries, for taking pictures of the battery to be tested, extracting the blue film background area on the surface of the blue film of the battery to be tested from the captured image using the characteristics of the blue film, and further detecting blue film defects based on a deep learning model.

[0004] This invention provides the following technical solution: In a first aspect, the present invention proposes a method for detecting defects in the blue film of a battery, comprising: The battery under test is scanned by a preset camera under a preset light source to obtain multiple images of the battery's blue film from different angles; the preset camera is perpendicular to the blue film surface of the outer surface of the battery under test. The battery blue film image is converted to the HSV color space to obtain the blue film background area; The blue film background region is detected using a deep learning model to obtain blue film defect information.

[0005] In one embodiment, the main light source of the preset light source is a blue LED light source with a preset peak wavelength; the light source layout of the preset light source is a ring light source tilted around the detection area of ​​the battery to be tested at a preset angle, and a coaxial light source is installed perpendicular to the surface of the blue film.

[0006] In one embodiment, before scanning the battery to be tested with a preset camera under a preset light source to obtain multiple images of the battery's blue film from different camera angles, the following steps are included: The camera parameters of the preset camera and the luminous intensity parameters of the preset light source are adjusted according to the blue film reflection characteristics, wherein the preset camera is a 2.5D area array camera.

[0007] In one embodiment, converting the battery blue film image to the HSV color space to obtain the blue film background area includes: Convert the battery blue film image to the HSV color space; The blue background area is extracted from the HSV color space based on a preset threshold range, which includes a hue threshold range, a saturation threshold range, and a brightness threshold range.

[0008] In one embodiment, extracting the blue background region from the HSV color space based on a preset threshold range includes: The blue component channel is obtained by separating it from the HSV color space based on the hue threshold range; The blue background region is extracted from the blue component channel based on the saturation threshold range and the brightness threshold range.

[0009] In one embodiment, the step of detecting the blue film background region based on a deep learning model to obtain blue film defect information includes: The blue film background region is denoised using a nonlocal mean filtering algorithm; Based on morphological operations, edge enhancement and morphological restoration are performed on the denoised blue film background region to obtain the image to be detected; The deep learning model is used to detect the image to be detected, and the blue film defect information is obtained.

[0010] In one embodiment, the method further includes: Construct an original training dataset, which includes battery blue film images of different batches and different battery models obtained under different lighting conditions, and their corresponding blue film defect annotation information. The deep learning model is obtained by training the original training dataset using a deep learning algorithm.

[0011] Secondly, the present invention provides a battery blue film defect detection device, comprising: The scanning module is used to scan the battery to be tested using a preset camera under a preset light source to obtain multiple images of the battery's blue film from different angles; the preset camera is perpendicular to the blue film surface of the outer surface of the battery to be tested. The conversion module is used to convert the battery blue film image to the HSV color space to obtain the blue film background area; The detection module is used to detect the blue film background area based on a deep learning model to obtain blue film defect information.

[0012] Thirdly, the present invention proposes a computer device including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the battery blue film defect detection method as described in the first aspect.

[0013] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery blue film defect detection method as described in the first aspect.

[0014] This invention discloses a method, apparatus, device, and storage medium for detecting blue film defects in batteries. The method involves scanning the battery under test using a preset camera under a preset light source, obtaining multiple images of the battery's blue film from different perspectives. The preset camera is perpendicular to the blue film surface of the battery's outer surface. The battery's blue film images are converted to the HSV color space to obtain the blue film background area. A deep learning model is used to detect the blue film background area to obtain blue film defect information. In this way, by combining the battery's blue film image with color space conversion (converting the image from RGB to HSV space) and extracting the blue background area, the contrast between defects and the background is specifically improved. Simultaneously, the image processing algorithm is optimized to effectively suppress surface texture noise while enhancing defect features, reducing the false positive rate. This allows for accurate differentiation of blue film defects such as bubbles and foreign objects without additional equipment, significantly improving detection efficiency while reducing equipment costs and minimizing the amount of manual re-inspection per hour on the production line. Attached Figure Description

[0015] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope of protection of the present invention. In the various drawings, similar components are numbered similarly.

[0016] Figure 1 A schematic flowchart of the battery blue film defect detection method proposed in this embodiment is shown. Figure 2 Another schematic diagram of the battery blue film defect detection method proposed in this embodiment is shown; Figure 3 This diagram illustrates another step in the process of the battery blue film defect detection method proposed in this embodiment. Figure 4 A schematic diagram of the battery blue film defect detection device proposed in this embodiment is shown.

[0017] Explanation of reference numerals in the attached diagram: 400 - Battery blue film defect detection device; 401 - Scanning module; 402 - Conversion module; 403 - Detection module. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] The components of the embodiments of the invention described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0020] In the following, the terms “comprising,” “having,” and their cognates, which may be used in various embodiments of the invention, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as excluding, firstly, the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more features, numbers, steps, operations, elements, components, or combinations thereof.

[0021] Furthermore, the terms "first," "second," and "third" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0022] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the invention pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be interpreted as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of the invention.

[0023] Example 1 This disclosure provides a method for detecting blue film defects in batteries. The method involves taking a picture of the battery to be tested, extracting the blue film background area from the captured image using the characteristics of the blue film, and further detecting blue film defects based on a deep learning model.

[0024] Please see Figure 1 The battery blue film defect detection method includes steps S101 to S103, and each step is described below.

[0025] Step S101: The battery to be tested is scanned by a preset camera under a preset light source to obtain multiple images of the battery blue film from different angles; the preset camera is perpendicular to the blue film surface of the outer surface of the battery to be tested.

[0026] In this embodiment, a preset camera scans the battery under test under a preset light source, obtaining multiple images of the battery's blue film from different camera angles. These blue film images are RGB images. The preset camera is perpendicular to the blue film surface of the battery under test, and in conjunction with a high-precision motion control platform, a complete one-time scan of the battery's surface is achieved. By optimizing the camera resolution and scanning accuracy parameters, it ensures that minute wrinkles at the 0.015mm level and three-dimensional morphological features such as weld slag protrusions less than 0.5mm in height can be captured, avoiding data stitching errors caused by line scanning.

[0027] It should be noted that by using two air pumps to misalign the batteries that are running side by side in the dual-station configuration, the preset camera can simultaneously take pictures and collect images of the two facades. In one specific embodiment, the main light source of the preset light source is a blue LED light source with a preset peak wavelength; the light source layout of the preset light source is a ring light source tilted around the detection area of ​​the battery to be tested at a preset angle, and a coaxial light source is installed perpendicular to the surface of the blue film.

[0028] In this embodiment, based on the high reflectivity of the blue film for light in the 450-500nm wavelength range, the main light source of the preset light source is a blue LED light source with a preset peak wavelength, which can be 470nm. The preset light source layout adopts a combination of multi-angle ring light sources and coaxial light sources. The ring light source is tilted around the detection area of ​​the battery under test at a preset angle to reduce specular reflection on the surface of the blue film. The preset angle can be 45°. The coaxial light source is installed perpendicular to the surface of the blue film to enhance the edge shadows of low-contrast defects such as electrolyte residue, thereby improving the visual difference between defects and the background.

[0029] In one specific embodiment, before step S101, the following steps are included: adjusting the camera parameters of the preset camera and the luminous intensity parameters of the preset light source according to the blue film reflection characteristics, wherein the preset camera is a 2.5D area array camera.

[0030] In this embodiment, the camera parameters of the preset camera and the luminous intensity parameters of the preset light source are adjusted according to the blue film's reflectivity. Specifically, the camera's exposure time, gain, and other imaging parameters are optimized based on the blue film's reflectivity. In areas with strong blue film reflectivity, the exposure time is reduced and the gain is increased to avoid overexposure and loss of defect information. In areas with low reflectivity, the exposure time is appropriately extended to ensure that weak defects such as electrolyte residue can be clearly imaged, thereby reducing specular reflection artifacts and improving the recognition rate of low-contrast defects such as electrolyte residue, fundamentally solving the problem of defect feature submersion caused by optical mismatch.

[0031] Furthermore, the default camera is a 2.5D area array camera. Based on the principle of structured light or photometric stereo, this camera projects a specifically coded structured light pattern onto the blue film surface and simultaneously acquires images from different perspectives using multiple cameras. By using 2.5D area array imaging to replace traditional 2D imaging and 2.5D line scanning, the limitations of two-dimensional planar detection are overcome, enabling rapid and complete acquisition of three-dimensional information of the blue film surface of large-size square batteries. The detection accuracy is improved from the traditional 0.2mm level to 0.05mm level, significantly reducing the missed detection rate of minute defects. Step S102: Convert the battery blue film image to the HSV color space to obtain the blue film background area.

[0032] In this embodiment, the battery blue film image is converted to the HSV color space to obtain the blue film background area. Through color space conversion, the accuracy, robustness and efficiency of color segmentation are significantly improved, while providing high-quality input for subsequent defect detection.

[0033] Please see Figure 2 In one specific embodiment, step S102 includes steps S1021 to S1022, and each step is described in detail below.

[0034] Step S1021: Convert the battery blue film image to the HSV color space.

[0035] In this embodiment, converting the battery blue film image to the HSV color space can solve the problems of the RGB model being sensitive to light and having incomplete color separation.

[0036] Step S1022: Extract the blue film background area from the HSV color space based on a preset threshold range, wherein the preset threshold range includes a hue threshold range, a saturation threshold range, and a brightness threshold range.

[0037] In this embodiment, the blue film background region is located in the HSV space using hue threshold range, saturation threshold range, and brightness threshold range. The hue threshold range is used to filter all cyan-blue pixels in the image, the saturation threshold range is used to retain high-purity blue, and the brightness threshold range is used to identify overly dark or overly bright areas. The hue threshold range (H) can be 180°-240°, the saturation threshold range (S) can be 30-100, and the brightness threshold range (V) can be 20-80, accurately extracting the blue film region and highlighting the contours of defects such as wrinkles and weld spatter bumps.

[0038] In one specific embodiment, step S1022 includes: separating the blue component channel from the HSV color space based on the hue threshold range; and extracting the blue film background region from the blue component channel based on the saturation threshold range and the brightness threshold range.

[0039] In this embodiment, the blue component channel is separated from the HSV color space using a hue threshold range. Further, saturation and brightness threshold ranges are used to exclude pixels from the blue component channel that have a blue hue but insufficient saturation (too gray) or unsuitable brightness (too dark / too bright). These excluded pixels may be the blue portion of the object itself, wrinkles / reflective / stain areas of the blue film, shadow areas, or overexposed areas, ultimately resulting in an ideal, uniform blue film background area.

[0040] Step S103: Detect the blue film background region based on a deep learning model to obtain blue film defect information.

[0041] In this embodiment, a deep learning model is used to detect the blue film background area to obtain blue film defect information, thereby effectively identifying battery blue film defects, achieving automated detection, effectively reducing the risk of missed detections due to human inspection negligence, ensuring the efficiency and reliability of the appearance inspection of large square battery blue film, and providing strong technical support for the large-scale production of new energy power batteries. The blue film defect information includes defect location, size, type, etc., and the deep learning model can be YOLOv11.

[0042] Please see Figure 3 In one specific embodiment, step S103 includes steps S1031 to S1033, and each step is described in detail below.

[0043] Step S1031: Denoise the blue film background region based on the nonlocal mean filtering algorithm.

[0044] In this embodiment, the background region of the blue film is denoised based on a nonlocal mean filtering algorithm, which preserves the edge information of defects while suppressing the texture noise on the surface of the blue film.

[0045] Step S1032: Based on morphological operations, edge enhancement and morphological restoration are performed on the denoised blue film background region to obtain the image to be detected.

[0046] In this embodiment, edge enhancement and shape restoration are performed on the denoised blue film background area based on morphological operations to obtain the image to be detected, making the continuous contour of the fine wrinkles clearer and the shape features of the welding slag protrusions more obvious.

[0047] Step S1033: Detect the image to be detected based on the deep learning model to obtain the blue film defect information.

[0048] In this embodiment, the image to be detected is enhanced based on the characteristics of the blue film to improve the defect contrast, thereby obtaining the blue film defect information. The false judgment rate is reduced, and no additional equipment such as infrared cameras is required. This not only reduces equipment costs but also significantly improves detection efficiency and reduces the amount of manual re-inspection per hour on the production line.

[0049] In one specific embodiment, the method further includes: constructing an original training dataset, the original training dataset including battery blue film images corresponding to different battery models in different batches obtained under different lighting conditions and their corresponding blue film defect annotation information; training the original training dataset based on a deep learning algorithm to obtain the deep learning model.

[0050] In this embodiment, an original training dataset is constructed, which includes blue film images of different battery models from different batches acquired under different lighting conditions, along with their corresponding blue film defect annotation information. The blue film defect annotation information includes information such as defect location, size, and type. The defect types need to include various defects such as wrinkles, weld slag bumps, and electrolyte residue.

[0051] The original training dataset was further trained using a deep learning algorithm to obtain a deep learning model for identifying blue film defects in batteries.

[0052] It should be noted that the blue membrane defect information detected by the deep learning model can be continuously optimized through cross-validation and multiple experiments to improve overall performance by adjusting model parameters and system settings. Furthermore, when the detection results are uncertain or unrecognizable, the system can promptly issue an alert or call for manual verification.

[0053] The first exemplary device is a modified 2D inspection device that uses a white ring light source, which is then modified to use the optical system of this invention. The new system uses a combination of a blue LED ring light source (50W power) with a peak wavelength of 470nm and a coaxial light source (30W power). The tilt angle of the ring light source is adjusted to 45°, and the light intensity of the light source and the camera exposure time are controlled (the exposure time is set to 1 / 500s and the gain is 20dB).

[0054] The test subjects were 1000 large rectangular battery blue films with minute electrolyte residues (approximately 0.1-0.3 mm² in area) and weld slag bumps (approximately 0.3-0.5 mm in height). Testing was conducted using equipment before and after the modification. The results showed that before the modification, due to specular reflection artifacts, the missed detection rate for electrolyte residues reached 32%, and the false detection rate for weld slag bumps was 25%. After adopting the optical system of this embodiment, the electrolyte residue recognition rate increased to 98%, the false detection rate for weld slag bumps decreased to 8%, and the surface illumination uniformity error of the blue film was optimized from ±12% to ±4%.

[0055] Secondly, in a laboratory setting, 500 sets of images of the blue film on a large rectangular battery, including wrinkles (0.015-0.03 mm in width), electrolyte residue, and weld slag bumps, were collected. These images were then processed using both a traditional grayscale thresholding algorithm system and the image processing system based on HSV color space conversion and deep learning, as described in this embodiment.

[0056] Test results show that the traditional algorithm has a correct recognition rate of only 55% for electrolyte residue, a 40% false negative rate for wrinkles, and an error of approximately ±0.1mm in identifying weld slag bumps. This embodiment, through HSV color space threshold extraction (H: 180°-240°, S: 30-100, V: 20-80), combined with nonlocal mean filtering and morphological operations, improves the accuracy of electrolyte residue recognition to 95%, increases wrinkle contour clarity by 62%, and reduces the error in identifying weld slag bump shape features to ±0.015mm. The false positive rate of the deep learning model for overall defects decreased from 18% to 4.5%.

[0057] Thirdly, in a large-scale power battery production workshop, two parallel production lines were selected. Line A retained the traditional manual-assisted 2D inspection equipment (manual re-inspection capacity of 25 pieces / hour, inspection efficiency of 150 pieces / hour), while Line B deployed the inspection equipment of this embodiment that integrates 2.5D area array imaging, optimized optical system and automated control algorithm.

[0058] After 8 hours of continuous operation, the statistics showed that 12 minor wrinkle defects were missed and 35 were misjudged. Line B achieved fully automated inspection, increasing the inspection capacity to 350 pieces per hour, reducing manual re-inspection to 8 pieces per hour, and improving inspection efficiency by 133%. Simultaneously, Line B maintained a stable inspection accuracy of over 99% for wrinkles of 0.015mm and high weld slag protrusions of 0.3mm, effectively verifying the high efficiency and reliability of the method in this embodiment in large-scale production.

[0059] The battery blue film defect detection method proposed in this embodiment scans the battery under test using a preset camera under a preset light source, obtaining multiple images of the battery blue film from different perspectives. The preset camera is perpendicular to the blue film surface of the battery under test. The battery blue film images are converted to the HSV color space to obtain the blue film background area. The blue film background area is then detected based on a deep learning model to obtain blue film defect information. In this way, the battery blue film image on the surface of the battery under test is combined with color space conversion, converting the battery blue film image from RGB space to HSV space, extracting the blue background area, and specifically improving the contrast between defects and background. At the same time, the image processing algorithm is optimized to effectively suppress blue film surface texture noise while enhancing defect features, reducing the false judgment rate. It can accurately distinguish blue film defects such as bubbles and foreign objects without additional equipment, significantly improving detection efficiency while reducing equipment costs and reducing the amount of manual re-inspection per hour on the production line.

[0060] Example 2 Furthermore, this disclosure provides a battery blue film defect detection device 400, please refer to [link to relevant documentation]. Figure 4 ,include: The scanning module 401 is used to scan the battery to be tested under a preset light source using a preset camera to obtain images of the battery's blue film from multiple cameras at different angles; the preset camera is perpendicular to the blue film surface of the outer surface of the battery to be tested. The conversion module 402 is used to convert the battery blue film image to the HSV color space to obtain the blue film background area; The detection module 403 is used to detect the blue film background area based on a deep learning model to obtain blue film defect information.

[0061] Optionally, the main light source of the preset light source is a blue LED light source with a preset peak wavelength; the light source layout of the preset light source is a ring light source tilted around the detection area of ​​the battery to be tested at a preset angle, and a coaxial light source is installed perpendicular to the surface of the blue film.

[0062] Optionally, the device further includes an adjustment module for adjusting the camera parameters of the preset camera and the luminous intensity parameters of the preset light source according to the blue film reflection characteristics, wherein the preset camera is a 2.5D area array camera.

[0063] Optionally, the conversion module 402 is further configured to convert the battery blue film image to the HSV color space; and extract the blue film background area from the HSV color space based on a preset threshold range, wherein the preset threshold range includes a hue threshold range, a saturation threshold range, and a brightness threshold range.

[0064] Optionally, the conversion module 402 is further configured to separate the blue component channel from the HSV color space based on the hue threshold range; and to extract the blue film background region from the blue component channel based on the saturation threshold range and the brightness threshold range.

[0065] Optionally, the detection module 403 is further configured to denoise the blue film background region based on a nonlocal mean filtering algorithm; perform edge enhancement and morphological repair on the denoised blue film background region based on morphological operations to obtain an image to be detected; and detect the image to be detected based on the deep learning model to obtain the blue film defect information.

[0066] Optionally, the detection module 403 is further configured to construct an original training dataset, which includes battery blue film images corresponding to different battery models in different batches obtained under different lighting conditions and their corresponding blue film defect annotation information; and to train the original training dataset based on a deep learning algorithm to obtain the deep learning model.

[0067] The apparatus provided in this embodiment can perform the steps of the battery blue film defect detection method provided in Embodiment 1. To avoid repetition, it will not be described again.

[0068] The battery blue film defect detection device proposed in this embodiment scans the battery under test using a preset camera under a preset light source, obtaining multiple images of the battery blue film from different perspectives. The preset camera is perpendicular to the blue film surface of the battery under test. The battery blue film images are converted to the HSV color space to obtain the blue film background area. The blue film background area is then detected based on a deep learning model to obtain blue film defect information. In this way, the battery blue film image on the surface of the battery under test is combined with color space conversion, converting the battery blue film image from RGB space to HSV space, extracting the blue background area, and specifically improving the contrast between defects and background. At the same time, the image processing algorithm is optimized to effectively suppress blue film surface texture noise while enhancing defect features, reducing the false judgment rate. It can accurately distinguish blue film defects such as bubbles and foreign objects without additional equipment, significantly improving detection efficiency while reducing equipment costs and reducing the amount of manual re-inspection per hour on the production line.

[0069] Example 3 Furthermore, this disclosure provides a computer device including a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the battery blue film defect detection method described in Embodiment 1.

[0070] The device provided in this embodiment can perform the steps of the battery blue film defect detection method provided in Embodiment 1. To avoid repetition, the steps will not be repeated.

[0071] Example 4 This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the battery blue film defect detection method described in Embodiment 1.

[0072] In this embodiment, the computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0073] The computer-readable storage medium provided in this embodiment can implement the battery blue film defect detection method provided in Embodiment 1. To avoid repetition, it will not be described again here.

[0074] In all examples shown and described herein, any specific values ​​should be interpreted as merely exemplary and not as limitations; therefore, other examples of exemplary embodiments may have different values.

[0075] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0076] The above-described embodiments are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention.

Claims

1. A method for detecting defects in the blue film of a battery, characterized in that, include: By scanning the battery under test with a preset camera under a preset light source, multiple images of the battery's blue film from different camera angles are obtained. The preset camera is perpendicular to the blue film surface of the outer surface of the battery to be tested; The battery blue film image is converted to the HSV color space to obtain the blue film background area; The blue film background region is detected using a deep learning model to obtain blue film defect information.

2. The battery blue film defect detection method according to claim 1, characterized in that, The main light source of the preset light source is a blue LED light source with a preset peak wavelength; the light source layout of the preset light source is a ring light source tilted around the detection area of ​​the battery to be tested at a preset angle, and a coaxial light source is installed perpendicular to the surface of the blue film.

3. The battery blue film defect detection method according to claim 2, characterized in that, Before scanning the battery under a preset light source using a preset camera to obtain multiple images of the battery's blue film from different camera angles, the process includes: The camera parameters of the preset camera and the luminous intensity parameters of the preset light source are adjusted according to the blue film reflection characteristics, wherein the preset camera is a 2.5D area array camera.

4. The battery blue film defect detection method according to claim 1, characterized in that, The step of converting the battery blue film image to the HSV color space to obtain the blue film background area includes: Convert the battery blue film image to the HSV color space; The blue background area is extracted from the HSV color space based on a preset threshold range, which includes a hue threshold range, a saturation threshold range, and a brightness threshold range.

5. The battery blue film defect detection method according to claim 4, characterized in that, Extracting the blue background region from the HSV color space based on a preset threshold range includes: The blue component channel is obtained by separating it from the HSV color space based on the hue threshold range; The blue background region is extracted from the blue component channel based on the saturation threshold range and the brightness threshold range.

6. The battery blue film defect detection method according to claim 1, characterized in that, The detection of the blue film background region based on a deep learning model to obtain blue film defect information includes: The blue film background region is denoised using a nonlocal mean filtering algorithm; Based on morphological operations, edge enhancement and morphological restoration are performed on the denoised blue film background region to obtain the image to be detected; The deep learning model is used to detect the image to be detected, and the blue film defect information is obtained.

7. The battery blue film defect detection method according to claim 1, characterized in that, The method further includes: Construct an original training dataset, which includes battery blue film images of different batches and different battery models obtained under different lighting conditions, and their corresponding blue film defect annotation information. The deep learning model is obtained by training the original training dataset using a deep learning algorithm.

8. A battery blue film defect detection device, characterized in that, include: The scanning module is used to scan the battery under test using a preset camera under a preset light source, and obtain images of the battery blue film from multiple cameras at different angles. The preset camera is perpendicular to the blue film surface of the outer surface of the battery to be tested; The conversion module is used to convert the battery blue film image to the HSV color space to obtain the blue film background area; The detection module is used to detect the blue film background area based on a deep learning model to obtain blue film defect information.

9. A computer device, characterized in that, The device includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the battery blue film defect detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the battery blue film defect detection method as described in any one of claims 1 to 7.

Images