A waste lithium battery positive and negative electrode sheet color selection method and device based on multi-modal fusion recognition

Abstract

The application discloses a kind of based on multimodal fusion identification's waste lithium battery positive and negative pole piece color selection method and device, comprising: in inert atmosphere, waste lithium battery is broken into fragment material;Fragment material is placed in low-temperature heat treatment environment, after keeping preset time, fragment material is screened, and the fragment material screened is single layer spread on transparent conveyor belt;Fragment material is lighted on transparent conveyor belt using adaptive light source, and fragment material is simultaneously subjected to RGB imaging and short-wave infrared detection, and obtain RGB image and short-wave infrared detection image;Material identification is carried out to RGB image and short-wave infrared detection image, and comprehensive material identification result is obtained;According to comprehensive material identification result, waste in fragment material is rejected and sorted, and positive pole piece and negative pole piece are obtained.The application effectively improves the sorting accuracy and robustness.

Description

Technical Field

[0001] This invention relates to the field of color sorting technology for positive and negative electrode sheets of lithium batteries, and in particular to a method and apparatus for color sorting positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition. Background Technology

[0002] Currently, the mainstream recycling technologies in the lithium battery recycling field include physical methods, hydrometallurgy, and pyrometallurgy. Physical methods separate battery components through a series of physical means such as crushing, magnetic separation, and eddy current separation; hydrometallurgy dissolves electrode materials through acid or alkali leaching, and then separates metal ions through extraction and precipitation steps; pyrometallurgy uses high-temperature pyrolysis to remove organic matter and then extract metal oxides.

[0003] In practical applications, these mainstream technologies often result in a mixture of positive and negative electrode powders. The presence of this mixture presents numerous challenges for subsequent processing. From a hydrometallurgical perspective, the characteristics of the negative electrode graphite powder differ significantly from those of the positive electrode material during the subsequent leaching process. The negative electrode graphite powder contains only lithium, and after reduction and roasting, it can be leached with water to obtain a lithium-ion solution. However, the positive electrode material, containing metals such as nickel, cobalt, and manganese, requires leaching with strong acids and reducing agents to convert it into a solution containing the corresponding metal ions. When the mixed positive and negative electrode powder is used as a raw material, the large amount of negative electrode graphite powder occupies the effective processing space of the leaching equipment, reducing its processing capacity. Furthermore, during the extraction and separation of valuable metals, the lithium element contained in the negative electrode is easily lost, resulting in a significant decrease in lithium recovery rate. Furthermore, graphite powder can only be discharged as an insoluble residue after leaching. During this process, the graphite powder carries valuable elements from the cathode as well as leaching acid, which not only consumes additional leaching material and increases recycling costs, but also further reduces the recovery rate of valuable components. More seriously, because graphite powder residue may contain metals and acids, it poses a risk of being classified as a hazardous chemical, which undoubtedly further increases the difficulty and cost of treatment.

[0004] Therefore, accurately separating the positive and negative electrode powders in the initial stages of lithium battery recycling is of paramount practical importance. This not only significantly improves the efficiency and economy of subsequent processing techniques, reduces resource waste and environmental pollution risks, but also lays a solid foundation for the sustainable development of the entire lithium battery recycling industry.

[0005] Chinese invention patent CN114147043A discloses a method for separating positive and negative electrode powders from recycled waste lithium batteries. This method achieves separation of positive and negative electrode powders from waste lithium batteries through one-time crushing, low-temperature heat treatment, hydrodynamic separation, color sorting, high-temperature pyrolysis, and wet or dry stripping. This invention utilizes the color difference between small, regularly shaped positive electrode sheets and negative electrode current collectors to separate the positive and negative electrode sheets through color sorting, thereby obtaining the separated positive and negative electrode powders. The electrode powder recovery rate is high, and all aluminum and copper foils are recycled. However, this method has the following drawbacks:

[0006] (1) When graphite or contamination remains on the surface of the negative electrode, the color characteristics of the copper foil are obscured, leading to misjudgment;

[0007] (2) The requirements for material dryness and shape regularity are strict, making it difficult to maintain a stable sorting effect in actual production;

[0008] (3) The influence of changes in ambient light on color sorting accuracy is not considered, and frequent calibration is required after long-term operation.

[0009] Therefore, there is an urgent need for a sorting technology that can overcome the limitations of color recognition, is more adaptable to material states, and is less affected by environmental interference, in order to improve the accuracy and stability of positive and negative electrode sorting in the lithium battery recycling process and meet the needs of actual industrial production. Summary of the Invention

[0010] To address the technical problems existing in the background art, this invention proposes a method and device for color sorting of positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition.

[0011] In a first aspect, the present invention proposes a color sorting method for positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition, comprising:

[0012] S1. In an inert atmosphere, waste lithium batteries are crushed into irregular fragments.

[0013] S2. Place the fragmented material in a low-temperature heat treatment environment of 150℃-200℃ and maintain it for a preset time;

[0014] S3. After maintaining the preset time, the fragmented material is screened, and the screened fragmented material that meets the preset size is spread in a single layer on the transparent conveyor belt.

[0015] S4. Illuminate the fragmented material on the transparent conveyor belt using an adaptive light source; simultaneously perform RGB imaging and short-wave infrared detection on the fragmented material on the transparent conveyor belt to obtain RGB images and short-wave infrared detection images; perform material recognition on the RGB images and short-wave infrared detection images to obtain a comprehensive material recognition result; based on the comprehensive material recognition result, remove waste materials from the fragmented material.

[0016] S5. The fragmented materials after waste removal are sorted to obtain positive electrode sheets and negative electrode sheets.

[0017] Preferably, material identification is performed on the RGB image and the short-wave infrared detection image to obtain a comprehensive material identification result, specifically including:

[0018] The RGB image is denoised and white balance corrected, then converted to the XYZ color space and then to the Lab color space to obtain the Lab image;

[0019] Background removal and instance segmentation are performed on the Lab image to obtain masks for each fragment material;

[0020] Based on the mask of each fragment material, perform a material identification on each fragment material to obtain a material identification result;

[0021] Based on the mask and short-wave infrared detection images of each fragment material, a first material identification is performed on each fragment material to obtain a second material identification result;

[0022] Based on the results of the first and second material identifications, a comprehensive material identification result is obtained.

[0023] Preferably, based on the mask of each fragment material, a material identification is performed on each fragment material to obtain a material identification result, specifically including:

[0024] The number of pixels within the mask of each fragment material is counted, and the masks of each fragment material are sequentially mapped onto the Lab image. Based on the number of pixels within the mask of each fragment material, the Lab value of each fragment material is calculated.

[0025] Each fragment material is classified according to its Lab value to obtain a primary material identification result. The primary material identification result includes: aluminum shell, separator, high-temperature tape, positive electrode sheet and negative electrode sheet. Aluminum shell, separator and high-temperature tape are waste materials that need to be removed. Positive electrode sheet and negative electrode sheet continue to undergo secondary waste removal.

[0026] Preferably, if and and The fragment was determined to be an aluminum shell; where, Indicates brightness, Indicates the red-green axis. Indicates the yellow and blue axis;

[0027] like and and The fragment was determined to be a diaphragm.

[0028] like and and The fragment was determined to be high-temperature tape.

[0029] like and and If there are local bright pixels with R≥220, then the fragment is determined to be a positive electrode; where R represents the grayscale value.

[0030] like and and If there are no bright pixels with R≥50, then the fragment is determined to be a negative electrode.

[0031] Preferably, the results of secondary material identification are divided into positive electrode sheet, negative electrode sheet, and high-temperature tape or separator;

[0032] The process involves a first material identification based on the mask and short-wave infrared detection images of each fragment, resulting in a second material identification result, which specifically includes:

[0033] The masks of each fragment material are mapped onto the short-wave infrared detection image, so that each fragment material corresponds to a pixel mask;

[0034] The reflectance of each fragment material is obtained based on all valid pixels within each pixel mask.

[0035] Based on the reflectivity of each fragment material and the preset reflectivity threshold, the results of secondary material identification for each fragment material are obtained.

[0036] Preferably, in the 1400–1600 nm wavelength band, if the reflectivity of a certain fragment material is 45–60%, then the fragment material is determined to be a positive electrode.

[0037] If the reflectivity of a certain fragment material is 15–25%, then the fragment material is determined to be a negative electrode.

[0038] If the reflectivity of a certain fragment is greater than 25% and less than 45%, then the fragment is determined to be a high-temperature tape or diaphragm.

[0039] Preferably, based on the comprehensive material identification results, waste materials are removed from the fragmented materials, specifically including:

[0040] When a fragment of material is identified as an aluminum shell, diaphragm, or high-temperature tape in the first or second material identification results, the fragment of material is removed from the transparent conveyor belt.

[0041] If a fragment of material is identified as either a positive electrode or a negative electrode in both the primary and secondary material identification results, then the fragment of material is retained on the transparent conveyor belt.

[0042] Preferably, while using an adaptive light source to illuminate the fragmented material located on the transparent conveyor belt, the method also includes:

[0043] The reflectivity signal of the surface of the fragment material is monitored in real time using an optical fiber sensor in an adaptive light source.

[0044] Based on the reflectivity signal of the fragment material surface, the light source spectrum of the adaptive light source and the recognition threshold in the material recognition process are automatically adjusted.

[0045] Preferably, the fragmented material after waste removal is sorted to obtain positive and negative electrode sheets, specifically including:

[0046] The fragmented material after waste removal is sorted using an airflow sorting component to obtain positive and negative electrode sheets.

[0047] Preferably, after using an optical fiber sensor in an adaptive light source to monitor the reflectivity signal of the surface of the fragment material in real time, the method further includes:

[0048] The PWM duty cycle of the airflow nozzles in the airflow sorting component is dynamically adjusted based on the reflectivity signal of the fragment material surface.

[0049] Secondly, the present invention also proposes a color sorting device for positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition, including a multimodal fusion recognition mechanism;

[0050] The multimodal fusion recognition mechanism includes an adaptive light source, an RGB imaging unit, a short-wave infrared detection unit, and a recognition unit.

[0051] Adaptive light sources are used to illuminate fragmented materials located on a transparent conveyor belt;

[0052] The RGB imaging unit and the short-wave infrared detection unit are used to simultaneously perform RGB imaging and short-wave infrared detection on the fragmented material located on the transparent conveyor belt to obtain RGB images and short-wave infrared detection images.

[0053] The recognition unit is used to identify materials from RGB images and short-wave infrared detection images to obtain a comprehensive material recognition result.

[0054] The proposed method and apparatus for color sorting of positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition in this invention first crushes the charged waste lithium batteries in an inert atmosphere, eliminating the need for pre-discharge, thus shortening the process and reducing energy consumption. Then, after low-temperature drying and screening, the fragmented materials that meet the preset size are spread in a single layer on a transparent conveyor belt. By simultaneously performing RGB imaging and short-wave infrared detection, RGB-SWIR dual-spectrum collaborative material recognition is achieved, resulting in accurate material recognition results. This significantly reduces graphite dust and ambient light interference, and significantly reduces the dependence on a single color feature, achieving a positive and negative electrode sheet recognition accuracy of ≥99%. Finally, based on the accurate material recognition results, the fragmented materials are discarded and sorted, effectively improving the sorting accuracy and robustness. Attached Figure Description

[0055] Figure 1 This is a flowchart illustrating a color sorting method for positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition, as proposed in one embodiment of the present invention.

[0056] Figure 2 This is a schematic diagram of the reflectance of fragmented material in one embodiment of the present invention. Detailed Implementation

[0057] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0058] Firstly, referring to Figure 1 This invention proposes a color sorting method for positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition, comprising:

[0059] S1. In an inert atmosphere, waste lithium batteries are crushed into irregular fragments.

[0060] S2. Place the fragmented material in a low-temperature heat treatment environment of 150℃-200℃ and maintain it for a preset time;

[0061] S3. After maintaining the preset time, the fragmented material is screened to remove fragmented material smaller than the preset size, and the screened fragmented material that meets the preset size is spread in a single layer on the transparent conveyor belt.

[0062] S4. Remove fragmented materials located on the transparent conveyor belt as waste material;

[0063] S5. The fragmented materials after waste removal are sorted to obtain positive electrode sheets and negative electrode sheets.

[0064] In this embodiment, in step S4, waste material removal is performed on the fragmented material located on the transparent conveyor belt, specifically including:

[0065] An adaptive light source is used to illuminate the fragmented material located on a transparent conveyor belt;

[0066] Simultaneously perform RGB imaging and shortwave infrared detection on the fragmented material located on the transparent conveyor belt to obtain RGB image and shortwave infrared detection (SWIR) image;

[0067] Material identification is performed on RGB images and short-wave infrared detection images to obtain a comprehensive material identification result; based on the comprehensive material identification result, waste materials in the fragmented materials are removed.

[0068] This invention first crushes charged waste lithium batteries in an inert atmosphere, eliminating the need for pre-discharge, thus shortening the process and reducing energy consumption. Then, after low-temperature drying and screening, fragments conforming to a preset size are spread in a single layer on a transparent conveyor belt. Simultaneous RGB imaging and short-wave infrared detection enable RGB-SWIR dual-spectrum collaborative material identification, yielding accurate material identification results. This significantly reduces graphite dust and ambient light interference, and significantly reduces reliance on a single color feature, achieving a positive and negative electrode identification accuracy of ≥99%. Based on the accurate material identification results, the fragments are then discarded and sorted, effectively improving sorting accuracy and robustness.

[0069] The process involves material identification from RGB images and short-wave infrared detection images to obtain a comprehensive material identification result, specifically including:

[0070] The RGB image is denoised and white balance corrected, then converted to the XYZ color space and then to the Lab color space to obtain the Lab image;

[0071] Background removal and instance segmentation are performed on the Lab image to obtain masks for each fragment material;

[0072] Based on the mask of each fragment material, perform a material identification on each fragment material to obtain a material identification result;

[0073] Each fragment material is identified once using the mask and short-wave infrared detection image, resulting in a second material identification result. Based on the first and second material identification results, a comprehensive material identification result is obtained.

[0074] In one specific embodiment, a CNN-based real-time classification model is used to simultaneously identify materials from RGB images and shortwave infrared detection images.

[0075] In this embodiment, the adaptive light source uses a combination of LEDs and filters, and can be preset with a program for adjusting the spectral output.

[0076] In one specific embodiment, the adaptive light source also includes a feedback fiber for real-time feedback of reflectivity.

[0077] In a further embodiment, based on the mask of each fragment material, a material identification is performed on each fragment material to obtain a material identification result, specifically including:

[0078] The number of pixels within the mask of each fragment material is counted, and the masks of each fragment material are sequentially mapped onto the Lab image. Based on the number of pixels within the mask of each fragment material, the Lab value of each fragment material is calculated.

[0079] Each fragment material is classified according to its Lab value to obtain a primary material identification result. The primary material identification result includes: aluminum shell, separator, high-temperature tape, positive electrode sheet and negative electrode sheet. Aluminum shell, separator and high-temperature tape are waste materials that need to be removed. Positive electrode sheet and negative electrode sheet continue to undergo secondary waste removal.

[0080] During the background removal process, this embodiment can use threshold, edge, or depth map to remove the conveyor belt background so that only the material area is retained.

[0081] During instance segmentation, this embodiment can use a pre-trained instance segmentation network, such as Mask R-CNN, YOLOv8-seg, or U-Net++, to cut out each aluminum shell, positive electrode, negative electrode, separator, and high-temperature tape separately and output a mask.

[0082] The Lab values ​​in this embodiment include: and Indicates brightness, Indicates the red-green axis. Indicates the yellow and blue axis; where, This indicates taking the average value.

[0083] During the classification process, if and and The fragment was determined to be an aluminum shell; if and and The fragment was determined to be a diaphragm.

[0084] like and and The fragment was determined to be high-temperature tape.

[0085] like and and And there are localized bright pixels. If the aluminum substrate shows reflective spots, then the fragment is determined to be a positive electrode sheet.

[0086] like and and And no bright pixels If so, the fragment is determined to be a negative electrode.

[0087] The results of secondary material identification are divided into positive electrode sheet, negative electrode sheet and high-temperature tape or diaphragm.

[0088] If the identification result of a certain fragment material is high-temperature tape or diaphragm, the fragment material is rejected.

[0089] The process involves a first material identification based on the mask and short-wave infrared detection images of each fragment, resulting in a second material identification result, which specifically includes:

[0090] The masks of each fragment material are mapped onto the short-wave infrared detection image, so that each fragment material corresponds to a pixel mask;

[0091] The reflectance of each fragment material is obtained based on all valid pixels within each pixel mask.

[0092] Based on the reflectivity of each fragment material and the preset reflectivity threshold, the results of secondary material identification for each fragment material are obtained.

[0093] The reflectivity of each fragment is... ;

[0094] ;

[0095] In the formula, DNs(i,j) represents the shortwave infrared reflectance (%) of pixel (i,j) after dark current correction; DNs(i,j) represents the original digital number of pixel (i,j) in the sample frame; DNd(i,j) represents the original digital number of pixel (i,j) in the dark frame, used to subtract dark current; DNw(i,j) represents the white reference plate, 99% represents the calibrated reflectance of the white reference plate in this band, if a 50% gray plate is used, multiply by 50%.

[0096] It is important to understand that the sample frame represents the fragment material captured by short-wave infrared detection; the white reference frame is obtained by flipping the white reference plate into the field of view at the same time or by automatically flipping it in before the start of each shift; the dark frame is obtained by collecting data with the lights off / blocked, and is used to deduct dark current.

[0097] like Figure 2As shown, during the secondary material identification process, in the 1400–1600 nm band, if the reflectance is 45–60%, the fragment material is determined to be a positive electrode (aluminum substrate); if the reflectance is 15–25%, the fragment material is determined to be a negative electrode (copper-graphite composite substrate), which can be accurately distinguished even if covered with graphite dust; if the reflectance is greater than 25% and less than 45%, the fragment is determined to be a high-temperature tape or diaphragm.

[0098] Among them, the reflectivity of each fragment material in the short-wave infrared detection image is as follows: Figure 2 As shown.

[0099] In a further embodiment, based on the comprehensive material identification results, waste materials are removed from the fragmented materials, specifically including:

[0100] When a fragment of material is identified as an aluminum shell, diaphragm, or high-temperature tape in the first or second material identification results, the fragment of material is removed from the transparent conveyor belt.

[0101] When a fragment of material is identified as a positive electrode or a negative electrode in the first and second material identification results, the fragment of material is retained on the transparent conveyor belt.

[0102] In this embodiment, the fragmented material after waste removal is sorted to obtain positive and negative electrode sheets, specifically including:

[0103] The fragmented material after waste removal is sorted using an airflow sorting component to obtain positive and negative electrode sheets.

[0104] In a further embodiment, while illuminating the fragmented material located on the transparent conveyor belt using an adaptive light source, the method also includes:

[0105] The reflectivity signal of the surface of the fragment material is monitored in real time using an optical fiber sensor in an adaptive light source.

[0106] Based on the reflectivity signal of the fragment material surface, the light source spectrum of the adaptive light source and the recognition threshold in the material recognition process are automatically adjusted.

[0107] This embodiment, by configuring the system in this way, eliminates drift caused by dust, humidity, and LED aging, significantly reducing reliance on a single color feature and improving sorting accuracy and robustness. This dynamic correction mechanism ensures that the system can operate continuously for more than 8 hours with a recognition threshold drift of less than ±1%.

[0108] In a further embodiment, after using an optical fiber sensor in an adaptive light source to monitor the reflectivity signal of the surface of the fragment material in real time, the method further includes:

[0109] Based on the reflectivity signal of the fragment material surface, the PWM duty cycle of the airflow nozzle in the airflow sorting component is dynamically adjusted to reduce the blowing force to prevent breakage of thin fragment materials and increase the blowing force to ensure separation of thick fragment materials, thereby reducing the secondary breakage rate of the electrode sheets and improving the recovery rate of valuable metals.

[0110] In one specific embodiment, the PWM duty cycle of the airflow nozzle is 30%–70%.

[0111] In this embodiment, the oxygen content in the inert atmosphere is 3% or less.

[0112] Secondly, the present invention also proposes a color sorting device for positive and negative electrode sheets of waste lithium batteries based on multimodal fusion recognition, including a multimodal fusion recognition mechanism;

[0113] The multimodal fusion recognition mechanism includes an adaptive light source, an RGB imaging unit, a short-wave infrared detection unit, and a recognition unit.

[0114] Adaptive light sources are used to illuminate fragmented materials located on a transparent conveyor belt;

[0115] The RGB imaging unit and the short-wave infrared detection unit are used to simultaneously perform RGB imaging and short-wave infrared detection on the fragmented material located on the transparent conveyor belt to obtain RGB images and short-wave infrared detection images.

[0116] The recognition unit is used to identify materials from RGB images and short-wave infrared detection images to obtain a comprehensive material recognition result.

[0117] The process involves material identification from RGB images and short-wave infrared detection images to obtain a comprehensive material identification result, specifically including:

[0118] The RGB image is denoised and white balance corrected, then converted to the XYZ color space and then to the Lab color space to obtain the Lab image;

[0119] Background removal and instance segmentation are performed on the Lab image to obtain masks for each fragment material;

[0120] Based on the mask of each fragment material, perform a material identification on each fragment material to obtain a material identification result;

[0121] Based on the mask and short-wave infrared detection images of each fragment material, a first material identification is performed on each fragment material to obtain a second material identification result;

[0122] Based on the results of the first and second material identifications, a comprehensive material identification result is obtained.

[0123] Specifically, based on the mask of each fragment material, a material identification is performed on each fragment material to obtain a material identification result, which includes:

[0124] The number of pixels within the mask of each fragment material is counted, and the masks of each fragment material are sequentially mapped onto the Lab image. Based on the number of pixels within the mask of each fragment material, the Lab value of each fragment material is calculated.

[0125] Each fragment material is classified according to its Lab value to obtain a primary material identification result. The primary material identification result includes: aluminum shell, separator, high-temperature tape, positive electrode sheet and negative electrode sheet. Aluminum shell, separator and high-temperature tape are waste materials that need to be removed. Positive electrode sheet and negative electrode sheet continue to undergo secondary waste removal.

[0126] During the classification process, if and and The fragment was determined to be an aluminum shell; where, Indicates brightness, Indicates the red-green axis. Indicates the yellow and blue axis;

[0127] like and and The fragment was determined to be a diaphragm.

[0128] like and and The fragment was determined to be high-temperature tape.

[0129] like and and If there are local bright pixels with R≥220, then the fragment is determined to be a positive electrode; where R represents the grayscale value.

[0130] like and and If there are no bright pixels with R≥50, then the fragment is determined to be a negative electrode.

[0131] The results of secondary material identification are divided into positive electrode sheets, negative electrode sheets, and high-temperature tapes or diaphragms.

[0132] The process involves a first material identification based on the mask and short-wave infrared detection images of each fragment, resulting in a second material identification result, which specifically includes:

[0133] The masks of each fragment material are mapped onto the short-wave infrared detection image, so that each fragment material corresponds to a pixel mask; the reflectivity of each fragment material is obtained based on all effective pixels within each pixel mask.

[0134] Based on the reflectivity of each fragment material and the preset reflectivity threshold, the results of secondary material identification for each fragment material are obtained.

[0135] It should be understood that in the 1400–1600nm wavelength band, if the reflectivity of a certain fragment material is 45–60%, then the fragment material is determined to be a positive electrode.

[0136] If the reflectivity of a certain fragment material is 15–25%, then the fragment material is determined to be a negative electrode.

[0137] If the reflectivity of a certain fragment is greater than 25% and less than 45%, then the fragment is determined to be a high-temperature tape or diaphragm.

[0138] In this embodiment, the waste identification result is obtained based on the primary material identification result and the secondary material identification result, specifically including:

[0139] If a fragment of material is identified as an aluminum shell, diaphragm, or high-temperature tape in a first or second material identification result, the fragment of material is determined to be waste.

[0140] If a fragment of material is identified as either a positive electrode or a negative electrode in both the primary and secondary material identification results, then the fragment of material is retained on the transparent conveyor belt.

[0141] In this embodiment, the adaptive light source also utilizes its own fiber optic sensor to monitor the reflectivity signal of the surface of the fragment material in real time; based on the reflectivity signal of the surface of the fragment material, it automatically adjusts the light source spectrum of the adaptive light source and the recognition threshold in the material recognition process.

[0142] In one embodiment, the recognition unit is pre-configured with a CNN-based real-time classification model. RGB images and short-wave infrared detection images are input into the CNN-based real-time classification model, and a comprehensive material recognition result based on the primary material recognition result and the secondary material recognition result is output.

[0143] The RGB imaging unit uses an RGB camera (400–700 nm).

[0144] The short-wave infrared detection unit uses an InGaAs infrared camera (1400–1600 nm).

[0145] Of course, the waste lithium battery positive and negative electrode color sorting device based on multimodal fusion recognition in this embodiment also includes: a battery crushing mechanism for crushing waste lithium batteries into irregular fragments in an inert atmosphere, a drying mechanism for performing low-temperature heat treatment on the fragments, a screening mechanism for screening the fragments, a vibrating feeding mechanism for spreading the screened fragments that meet the preset size onto a transparent conveyor belt in a single layer, and a waste removal mechanism and a sorting mechanism for removing waste located on the transparent conveyor belt.

[0146] The sorting mechanism in this embodiment is an airflow sorting mechanism. The airflow sorting mechanism is also used to dynamically adjust the PWM duty cycle of the airflow nozzles in the airflow sorting component based on the reflectivity signal of the surface of the fragment material.

[0147] In one specific embodiment, each orifice in the airflow nozzle has an independent PWM with a duty cycle of 30–70% and a response time of <2ms.

[0148] Of course, this embodiment also includes a positive electrode material box for collecting positive electrode sheets, a negative electrode material box for collecting negative electrode sheets, and a waste box for collecting waste materials.

[0149] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0150] Example 1

[0151] 200 kg of waste ternary lithium batteries were crushed to an average particle size of 18 mm under a nitrogen atmosphere (oxygen content 2.8%). The crushed material was volatilized at 180 ℃ for 30 min and then passed through a 1 mm sieve, with a fine powder removal rate of 7.5%. Subsequently, it was spread in a single layer at a linear velocity of 0.8 m / s, and RGB cameras were used to capture data at 2048×1536@60 fps, and SWIR cameras were used to capture data at 640×512@100 fps.

[0152] In this embodiment, the RGB camera and SWIR camera are coaxially arranged above the conveyor belt. The SWIR camera parameters are: typical 1400–1600 nm bandpass, InGaAs sensor, 14-bit output; a white reference plate (PTFE or BaSO4, known reflectivity ≈99%) is used; a sealed chamber with an active SWIR light source (halogen tungsten lamp or LED array) is used to ensure uniform illumination and immunity to ambient light interference. Hard-triggered synchronization of the camera, light source, and encoder pulses ensures consistent lighting conditions for each frame of image acquisition.

[0153] The results showed that the positive electrode recognition rate was 99.3%, the negative electrode recognition rate was 99.1%, the secondary breakage rate was reduced by 34%, and the threshold drift was 0.7% after 12 hours of continuous operation.

[0154] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition, characterized in that, The method comprises the following steps: S1, crushing the waste lithium battery into irregular fragment materials in an inert atmosphere; S2, placing the fragment materials in a low-temperature heat treatment environment of 150-200°C and maintaining for a preset time; S3, after maintaining for a preset time, screening the fragment materials, and spreading the screened fragment materials in a single layer on a transparent conveying belt; S4, using an adaptive light source to light the fragment materials on the transparent conveying belt; at the same time, RGB imaging and short-wave infrared detection are performed on the fragment materials on the transparent conveying belt to obtain an RGB image and a short-wave infrared detection image; the RGB image is denoised and white balance corrected, and the RGB image is converted to XYZ color space and then to Lab color space to obtain a Lab image; the Lab image is subjected to background removal and instance segmentation to obtain a mask of each fragment material; the pixel amount in each fragment material mask is counted, and each fragment material mask is sequentially mapped onto the Lab image, and the Lab value of each fragment material is calculated according to the pixel amount in each fragment material mask; The various fragment materials are classified according to Lab values of the various fragment materials, to obtain a primary material identification result; wherein the primary material identification result includes: aluminum shell, diaphragm, high-temperature adhesive tape, positive electrode sheet and negative electrode sheet, and the aluminum shell, diaphragm and high-temperature adhesive tape are waste materials that need to be removed; wherein if and and , it is determined as an aluminum shell; in the formula, represents lightness, represents a red-green axis, represents a yellow-blue axis; if and and , it is determined as a diaphragm; if and and , it is determined as a high-temperature adhesive tape; if and and and there is a local highlight pixel of R≥220, it is determined as a positive electrode sheet; in the formula, R represents a gray value; if and and , and there is no highlight pixel of R≥50, it is determined as a negative electrode sheet; based on the mask of each fragment material and the short-wave infrared detection image, a first material identification of each fragment material is performed to obtain a second material identification result; based on the first material identification result and the second material identification result, a comprehensive material identification result is obtained; based on the comprehensive material identification result, waste materials in the fragment materials are removed; S5, sorting the fragment materials after the waste materials are removed to obtain positive electrode sheets and negative electrode sheets.

2. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 1, characterized in that, The second material identification result is divided into positive electrode sheets, negative electrode sheets, and high-temperature adhesive tapes or separators; wherein, based on the mask of each fragment material and the short-wave infrared detection image, a first material identification of each fragment material is performed to obtain a second material identification result, which specifically comprises: mapping each fragment material mask into the short-wave infrared detection image, so that each fragment material corresponds to a pixel mask; based on all effective pixels in each pixel mask, the reflectivity of each fragment material is obtained; based on the reflectivity of each fragment material and a preset reflectivity threshold, the second material identification result of each fragment material is obtained.

3. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 2, characterized in that, Under the 1400-1600nm waveband, if the reflectivity of a certain fragment material is 45-60%, it is determined that the fragment material is a positive electrode sheet; if the reflectivity of a certain fragment material is 15-25%, it is determined that the fragment material is a negative electrode sheet; if the reflectivity of a certain fragment material is greater than 25% and less than 45%, it is determined that the fragment is a high-temperature adhesive tape or a separator.

4. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 1, characterized in that, Based on the comprehensive material identification result, the waste materials in the fragment materials are removed, which specifically comprises: when a certain fragment material is identified as an aluminum shell, a separator, or a high-temperature adhesive tape in the first material identification result or the second material identification result, the fragment material is removed from the transparent conveying belt; when a certain fragment material is identified as a positive electrode sheet or a negative electrode sheet in the first material identification result and the second material identification result, the fragment material is retained on the transparent conveying belt.

5. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 1, characterized in that, The use of an adaptive light source to light the fragment materials on the transparent conveying belt also includes: using a fiber sensor in the adaptive light source to monitor the reflectivity signal of the surface of the fragment materials in real time; According to the reflectivity signal of the surface of the fragment material, the light spectrum of the adaptive light source and the identification threshold in the material identification process are automatically adjusted.

6. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 5, characterized in that, The fragment material after waste removal is sorted to obtain positive and negative electrode sheets, specifically including: The fragment material after waste removal is sorted by using the airflow sorting assembly to obtain positive and negative electrode sheets.

7. The waste lithium battery positive and negative electrode sheet color selection method based on multi-modal fusion recognition according to claim 6, characterized in that, After the reflectivity signal of the surface of the fragment material is monitored in real time by using the optical fiber sensor in the adaptive light source, it further includes: According to the reflectivity signal of the surface of the fragment material, the PWM duty cycle of the airflow nozzle in the airflow sorting assembly is dynamically adjusted.

8. A waste lithium battery positive and negative electrode sheet color selection device based on multi-modal fusion recognition, characterized in that, It includes a multi-modal fusion identification mechanism; The multi-modal fusion identification mechanism includes an adaptive light source, an RGB imaging unit, a short-wave infrared detection unit, and an identification unit. The adaptive light source is used to light the fragment material on the transparent conveying belt; The RGB imaging unit and the short-wave infrared detection unit are used to simultaneously perform RGB imaging and short-wave infrared detection on the fragment material on the transparent conveying belt to obtain RGB images and short-wave infrared detection images; The identification unit is configured to perform noise reduction and white balance correction on the RGB image, convert the RGB image to XYZ color space and then to Lab color space to obtain a Lab image, perform background removal and instance segmentation on the Lab image to obtain a mask of each piece of material, count the number of pixels in each mask, and sequentially map each mask to the Lab image, calculate the Lab value of each piece of material according to the number of pixels in each mask, and classify each piece of material according to the Lab value of each piece of material to obtain a first material identification result; wherein the first material identification result includes: aluminum shell, diaphragm, high-temperature adhesive tape, positive electrode sheet and negative electrode sheet, and the aluminum shell, diaphragm and high-temperature adhesive tape are waste materials that need to be removed; wherein if and and , the aluminum shell is determined; in the formula, represents lightness, represents the red-green axis, represents the yellow-blue axis; if and and , the diaphragm is determined; if and and , the high-temperature adhesive tape is determined; if and and and there is a local highlight pixel with R≥220, the positive electrode sheet is determined; in the formula, R represents a gray value; if and and , and there is no highlight pixel with R≥50, the negative electrode sheet is determined; According to the mask and the short-wave infrared detection image of each fragment material, a first material identification is performed on each fragment material to obtain a second material identification result; and according to the first material identification result and the second material identification result, a comprehensive material identification result is obtained.

Images