Intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste

Abstract

This invention relates to the field of metal recycling technology, and more particularly to an intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste. The method includes: acquiring image data and weight data of the waste to be recycled; calculating an integrity index based on contour regularity and weight data to determine whether the recycling type is whole-piece recycling or dismantling recycling. If it is dismantling recycling, the dismantling area is located through the image, and the dismantling order of the area is determined by combining the degree of damage and color saturation parameters of each area. Further, candidate dismantling points are determined within each area, and the dismantling order of the points is determined by analyzing the damage risk parameters and structural complexity parameters of the points. Finally, precise dismantling is performed based on the area dismantling order and the point dismantling order. This invention, through multi-level visual recognition and intelligent sorting techniques, achieves the beneficial effects of reducing damage to non-ferrous metals during dismantling and improving recycling rate and efficiency.

Description

Technical Field

[0001] This invention relates to the field of metal recycling technology, and in particular to an intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste. Background Technology

[0002] With the rapid iteration and upgrading of electronic and electrical products, the amount of industrial electronic and electrical waste generated is increasing year by year. This type of waste contains a large amount of high-value non-ferrous metals such as copper, aluminum, gold, and silver, and has extremely high recycling value. However, electronic and electrical products have complex structures, diverse component connection methods, and the waste comes from a wide range of sources and varies in degree of damage, posing a huge challenge to automated recycling.

[0003] Traditional non-ferrous metal recycling methods mainly rely on manual dismantling or overall mechanical crushing. Manual dismantling is inefficient, labor-intensive, and poses safety hazards; while overall crushing, although fast, results in a mixture of non-ferrous metals and non-metallic materials, increasing the difficulty of subsequent sorting and potentially damaging high-value metal components, leading to resource waste. Furthermore, existing automated recycling systems struggle to intelligently adjust dismantling strategies based on the actual condition of the waste. When processing complex or partially damaged electronic and electrical waste, problems such as improper dismantling sequence and damage to non-ferrous metals often arise, affecting recycling purity and economic efficiency.

[0004] Chinese Patent Publication No. CN110523747A discloses an electronic waste recycling and processing device with complete component stripping. Its structure includes a main body, a separation device, a control panel, a crushing shaft, a feed hopper, and a discharge port. Beneficial effects: This invention utilizes a stripping structure. Under the action of the rotating cone-shaped pins along the stripping shaft, the pins act as a magnetic structure to attract metal parts. The rotation also causes friction, resulting in a scraping-like stripping of the metal parts from the plastic surface, thus achieving separation of plastic and metal. This effectively improves the efficiency of waste recycling. The invention also utilizes a cleaning structure. Under the interaction of a magnetic plate and a swing plate, falling metal parts are attracted, while falling plastic parts are pulled down by a rectangular frame on its surface, achieving complete separation of metal and plastic. The attracted metal parts are then swept to the inside of the circular hopper by the swing plate.

[0005] Therefore, the existing technology has the following problems: The problem is that it cannot intelligently adjust the dismantling strategy according to the individual differences of waste, and it is difficult to protect the integrity of high-value metals. Summary of the Invention

[0006] Therefore, this invention provides an intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste, in order to overcome the problems in the prior art that it is impossible to intelligently adjust the dismantling strategy according to the individual differences of the waste and that it is difficult to protect the integrity of high-value metals.

[0007] To achieve the above objectives, the present invention provides an intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste, comprising: Step S1: Obtain image data and weight data of the electronic and electrical waste to be recycled; Step S2: Based on the image data, obtain the contour regularity of the electronic and electrical waste to be recycled. Based on the contour regularity and combined with the weight data of the electronic and electrical waste to be recycled, determine the integrity index of the electronic and electrical waste to be recycled. Based on the integrity index, determine the recycling type of the electronic and electrical waste to be recycled, wherein the recycling type includes whole recycling type and dismantling recycling type. Step S3: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is dismantling recycling, several dismantling areas are located based on the image data, and the dismantling order of the several dismantling areas is determined based on the degree of damage and color saturation parameters of the several dismantling areas. Step S4: Determine candidate disassembly points for several disassembly areas, determine the damage risk parameters and structural complexity parameters of the candidate disassembly points based on the image data, and determine the disassembly order of the candidate disassembly points according to the damage risk parameters and structural complexity parameters. Step S5: Disassemble the electronic and electrical waste to be recycled based on the disassembly sequence of the area and the disassembly sequence of the location.

[0008] Further, in step S2, the step of obtaining the contour regularity of the electronic and electrical waste to be recycled based on the image data, and determining the integrity index of the electronic and electrical waste to be recycled based on the contour regularity and in combination with the weight data of the electronic and electrical waste to be recycled, includes: Step S21: Perform edge detection and contour extraction on the image data to determine the minimum bounding rectangle of the electronic and electrical waste to be recycled; Step S22: Calculate the ratio of the actual outline area of ​​the electronic and electrical waste to be recycled to the area of ​​the minimum bounding rectangle, and use the ratio as the outline regularity of the electronic and electrical waste to be recycled; Step S23: Obtain the weight data of the electronic and electrical waste to be recycled, and determine the weight deviation rate of the electronic and electrical waste to be recycled by combining it with the standard weight range of this type of electronic and electrical waste pre-stored in the database. Step S24: Determine the integrity index based on the contour regularity and the weight deviation rate.

[0009] Further, in step S2, determining the recycling type of the electronic and electrical waste to be recycled based on the integrity index includes: If the integrity index is greater than or equal to the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be the whole recycling type; If the integrity index is less than the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be dismantling and recycling.

[0010] Further, in step S3, locating several disassembly areas based on the image data, and determining the disassembly order of the several disassembly areas based on the degree of damage and color saturation parameters of the several disassembly areas, includes: Step S31: Perform semantic segmentation on the image data to locate several dismantling areas of the electronic and electrical waste to be recycled; Step S32: Determine the damage level index of each disassembly area based on the edge jaggedness, surface crack density, and missing area ratio of each disassembly area. Step S33: Convert the image data to the HSV color space, extract the saturation channel data of each of the disassembled regions, and determine the average saturation value of each of the disassembled regions as the color saturation parameter of each of the disassembled regions. Step S34: Determine the priority score for each dismantling area based on the damage level index and color saturation parameter; Step S35: Sort the several disassembly areas in descending order of priority score to generate the region disassembly order of the disassembly areas.

[0011] Further, in step S4, determining the candidate disassembly points of the disassembly area includes: Step S411: Local feature extraction is performed on the image data of the disassembly area, and a corner detection algorithm is used to identify several candidate disassembly points within the disassembly area; Step S412: For each candidate disassembly point, extract the geometric features of each candidate disassembly point. Step S413: Based on the geometric features, perform type identification for each candidate disassembly point; Step S414: Record the type identifier of each candidate dismantling point, its two-dimensional coordinates in the image, and its three-dimensional coordinates obtained through three-dimensional reconstruction, and generate an information set for each candidate dismantling point.

[0012] Further, in step S4, the damage risk parameters of the candidate dismantling points are determined based on the image data, including: Step S421: For each candidate disassembly point, identify non-ferrous metal components within a preset range adjacent to each candidate disassembly point based on the image data, and determine the type, size, and spatial orientation of the non-ferrous metal components. Step S422: Based on the information set of the candidate disassembly points, determine the force direction and stress propagation range when disassembling the candidate disassembly point; Step S423: Determine the spatial distance between the candidate disassembly point and each adjacent non-ferrous metal component; Step S424: Determine the force direction influence coefficient based on the angle between the force direction and the direction pointing towards the non-ferrous metal component; Step S425: Determine the stress overlap coefficient based on the degree of overlap between the stress propagation range and the spatial position of the non-ferrous metal component; Step S426: Based on the spatial distance, the force direction influence coefficient, and the stress overlap coefficient, determine the damage risk parameters of the candidate dismantling points.

[0013] Further, in step S422, determining the force direction and stress propagation range when dismantling a candidate dismantling point based on the information set of the candidate dismantling point includes: Step S4221: Based on the type identifier of the candidate disassembly point, query the preset disassembly process database to obtain the force direction corresponding to the type of the candidate disassembly point; Step S4222: Based on the geometric characteristics of the candidate disassembly point, the force direction is corrected to obtain the actual force direction of the candidate disassembly point; Step S4223: Based on the type identifier and geometric characteristics of the candidate disassembly point, query the preset stress propagation model database to obtain the stress propagation radius corresponding to the type of the candidate disassembly point; Step S4224: Determine the stress propagation range of the candidate disassembly point, centered on the candidate disassembly point and within the range of the stress propagation radius.

[0014] Further, in step S4, the structural complexity parameters of the candidate disassembly points are determined based on the image data, including: Step S431: For each candidate dismantling point, extract local image blocks within a preset range around the candidate dismantling point based on the image data, and perform edge detection on the local image blocks to obtain an edge distribution map; Step S432: Based on the edge distribution map, determine the edge density around the candidate disassembly point; Step S433: Perform texture analysis on the local image patch, extract the texture features of the local image patch, and calculate the texture entropy value; Step S434: Based on the image data, obtain the three-dimensional point cloud data of the area surrounding the candidate dismantling point through three-dimensional reconstruction, and determine the mean square error of the local surface curvature of the area surrounding the candidate dismantling point as a geometric complexity index. Step S435: Identify the number of components and the connection relationship between components within a preset range around the candidate disassembly point, and determine the component connection density; Step S436: Determine the structural complexity parameters of the candidate disassembly points based on the edge density, the texture entropy value, the geometric complexity index, and the component connection density.

[0015] Further, in step S4, the dismantling order of the candidate dismantling points is determined based on the damage risk parameter and the structural complexity parameter, including: Step S441: For each candidate disassembly point, the damage risk parameter and the structural complexity parameter are normalized respectively to obtain the normalized damage risk value and the normalized structural complexity value. Step S442: Based on the preset weight coefficients, the dismantling priority score of each candidate dismantling point is determined according to the normalized damage risk value and the normalized structural complexity value. Step S443: Sort all candidate disassembly points in the disassembly area according to the disassembly priority score from high to low, and generate the disassembly order of the candidate disassembly points.

[0016] Furthermore, the method also includes: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is the whole recycling type, the electronic and electrical waste to be recycled is transported to the crushing and sorting system for crushing and sorting to recover non-ferrous metals.

[0017] Compared with existing technologies, the advantages of this invention are as follows: This invention fuses visual image data with weight data for judgment. First, it obtains contour regularity based on edge detection and contour extraction, calculates the weight deviation rate by combining it with standard weight ranges in a pre-stored database, and then weights and fuses the two to obtain an integrity index. This index is then compared with a preset threshold to accurately distinguish between whole-item recycling and dismantled-item recycling types, thus achieving intelligent diversion of electronic and electrical waste. This method only requires conventional image acquisition and weighing equipment, without the need for complex and expensive sensors, and has the advantages of low cost and simple implementation. Contour regularity effectively identifies the shape integrity of objects, and weight deviation rate effectively identifies the internal structural integrity of objects. The two methods complement and verify each other, avoiding the shortcomings of relying solely on visual judgment, which is easily interfered with by surface patterns, or relying solely on weight, which is easily affected by impurities. This significantly improves the accuracy and robustness of the recycling type determination. For complete equipment determined to be of the whole recycling type, subsequent fine disassembly can be carried out to extract high-value non-ferrous metal components. For crushed materials determined to be of the disassembly recycling type, they can directly enter the efficient crushing and sorting process, realizing differentiated treatment of different types of waste. This avoids the value loss caused by directly crushing complete equipment and the efficiency waste caused by feeding crushed materials into fine disassembly. As a result, the overall recovery rate and efficiency of non-ferrous metals are improved, which has significant industrial application value.

[0018] Furthermore, this invention accurately locates each dismantling area on electronic and electrical waste using semantic segmentation technology, and prioritizes each area based on damage severity indicators and color saturation parameters, achieving intelligent planning of dismantling areas. The damage severity indicator comprehensively quantifies the physical damage state of the area by considering three dimensions: edge jaggedness, surface crack density, and missing area ratio. Areas with more severe damage are prioritized for dismantling, making full use of damaged areas as natural dismantling entry points and reducing dismantling difficulty. The color saturation parameter is extracted based on the HSV color space, effectively identifying the low saturation characteristics of non-ferrous metals, prioritizing the dismantling of areas with obvious metal characteristics, and ensuring priority recycling of high-value components. The priority score obtained by weighted fusion of these two parameters balances the feasibility of dismantling and the value of recycling, significantly improving dismantling efficiency and non-ferrous metal recovery rate, and has significant industrial application value.

[0019] Furthermore, this invention determines the dismantling order of candidate dismantling points by fusing damage risk parameters and structural complexity parameters, achieving an optimized balance between safety and efficiency in the dismantling process. The damage risk parameter comprehensively considers the spatial distance between the point and adjacent non-ferrous metal components, the influence coefficient of the force direction, and the stress overlap coefficient, and introduces the value weight of different metal types. This accurately quantifies the risk of damage to high-value non-ferrous metals such as gold, silver, and copper during dismantling operations, ensuring that high-risk points are prioritized to avoid accidental damage to valuable components. The structural complexity parameter integrates four dimensions: edge density, texture entropy value, local surface curvature mean square error, and component connection density. This comprehensively quantifies the complexity of the structure surrounding the point, prioritizing the dismantling of points with simple structures to improve dismantling efficiency and avoid complex points blocking the overall process. The two parameters are normalized and weighted to obtain a dismantling priority score, ensuring both the basic principle of safety priority and the actual need for maximizing efficiency, significantly reducing the damage rate of non-ferrous metals and increasing the overall value of the recycled materials. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment. Figure 2 This is a flowchart illustrating the process of determining the integrity index in the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment. Figure 3 This is a flowchart illustrating the process of determining the recycling type in the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment. Detailed Implementation

[0021] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0022] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0023] Please see Figures 1-3 As shown, Figure 1 This is a flowchart illustrating the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment. Figure 2 This is a flowchart illustrating the process of determining the integrity index in the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment. Figure 3 This is a flowchart illustrating the process of determining the recycling type in the intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste in this embodiment.

[0024] This embodiment provides an intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste, including: Step S1: Obtain image data and weight data of the electronic and electrical waste to be recycled; Step S2: Based on the image data, obtain the contour regularity of the electronic and electrical waste to be recycled. Based on the contour regularity and combined with the weight data of the electronic and electrical waste to be recycled, determine the integrity index of the electronic and electrical waste to be recycled. Based on the integrity index, determine the recycling type of the electronic and electrical waste to be recycled, wherein the recycling type includes whole recycling type and dismantling recycling type. Step S3: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is dismantling recycling, several dismantling areas are located based on the image data, and the dismantling order of the several dismantling areas is determined based on the degree of damage and color saturation parameters of the several dismantling areas. Step S4: Determine candidate disassembly points for several disassembly areas, determine the damage risk parameters and structural complexity parameters of the candidate disassembly points based on the image data, and determine the disassembly order of the candidate disassembly points according to the damage risk parameters and structural complexity parameters. Step S5: Disassemble the electronic and electrical waste to be recycled based on the disassembly sequence of the area and the disassembly sequence of the location.

[0025] In this embodiment of the invention, the image data includes, but is not limited to: visible light images, multispectral images, infrared images, or three-dimensional point cloud data. Acquisition method: Multi-angle imaging is performed using a CCD industrial camera, hyperspectral camera, or three-dimensional laser scanner to collect the electronic and electrical waste to be recycled; the weight data includes, but is not limited to: the actual weight of a single piece of waste, the dynamic weight during continuous conveying, or the deviation from the standard weight. Simple acquisition method: Real-time measurement and acquisition are performed using a dynamic rail scale or a conveyor belt continuous weighing module.

[0026] Specifically, in step S2, obtaining the contour regularity of the electronic and electrical waste to be recycled based on the image data, and determining the integrity index of the electronic and electrical waste to be recycled based on the contour regularity and in combination with the weight data of the electronic and electrical waste to be recycled, includes: Step S21: Perform edge detection and contour extraction on the image data to determine the minimum bounding rectangle of the electronic and electrical waste to be recycled; Step S22: Calculate the ratio of the actual outline area of ​​the electronic and electrical waste to be recycled to the area of ​​the minimum bounding rectangle, and use the ratio as the outline regularity of the electronic and electrical waste to be recycled; Step S23: Obtain the weight data of the electronic and electrical waste to be recycled, and determine the weight deviation rate of the electronic and electrical waste to be recycled by combining it with the standard weight range of this type of electronic and electrical waste pre-stored in the database. Step S24: Determine the integrity index based on the contour regularity and the weight deviation rate. In this embodiment of the invention, the specific method for determining the contour regularity is as follows: The RGB image of the collected electronic and electrical waste to be recycled is converted to grayscale; the edge information in the image is extracted using the Canny edge detection algorithm; the outermost contour is obtained through a contour tracking algorithm; the total number of pixels in the area enclosed by the contour is calculated as the actual contour area; simultaneously, the minimum bounding rectangle of the contour is calculated, and the total number of pixels in the area enclosed by the rectangle is counted as the area of ​​the minimum bounding rectangle; the actual contour area is divided by the area of ​​the minimum bounding rectangle, and the resulting ratio is the contour regularity. This parameter ranges from 0 to 1. The closer the value is to 1, the more regular and closer to a complete rectangular shape the object's contour is; the closer the value is to 0, the more irregular and fragmented the object's contour is.

[0027] It should be noted that before obtaining the weight data in step S1, the electronic and electrical waste to be recycled is pre-treated: loose debris, dust, and temporary binding materials attached to the surface are removed by blowing, vibration, or simple cleaning to ensure the accuracy of the weight data. The weight data referred to in this technical solution refers to the net weight measured after the above pre-treatment, excluding externally attached non-fixed debris.

[0028] The specific method for determining the weight deviation rate is as follows: First, based on the contour information extracted in step S21 and combined with the pre-stored electronic and electrical waste type identification model, the possible types of electronic and electrical waste to be recycled are initially determined, such as desktop computers, laptops, servers, monitors, printers, etc.; then, the standard weight range of this type of electronic and electrical waste is queried from the database. This standard weight range is obtained by weighing and statistically analyzing a large number of intact electronic and electrical wastes of the same type, including the minimum standard weight and the maximum standard weight; the weight data of the electronic and electrical waste to be recycled collected by the actual weighing equipment is obtained, the absolute deviation between the actual weight and the median of the standard weight range is calculated, and then divided by half the width of the standard weight range to obtain the weight deviation rate. The value of this parameter is between 0 and 1. The closer the value is to 0, the closer the actual weight is to the standard weight. The closer the value is to 1, the farther the actual weight deviates from the standard weight.

[0029] The specific method for determining the integrity index is as follows: contour regularity and weight deviation rate are weighted and fused according to preset weight coefficients. Contour regularity is positively correlated with the integrity index, and weight deviation rate is negatively correlated with the integrity index. The specific calculation formula is that the integrity index equals contour regularity multiplied by the contour regularity weight coefficient plus one minus the weight deviation rate multiplied by the weight deviation rate weight coefficient. The sum of the contour regularity weight coefficient and the weight deviation rate weight coefficient is 1, and can be adjusted according to the actual application scenario. Typical values ​​are 0.6 for contour regularity weight coefficient and 0.4 for weight deviation rate weight coefficient.

[0030] It should be noted that the weight deviation rate is used to characterize the degree of deviation between the actual weight and the standard weight. The smaller the value, the closer the actual weight is to the standard weight, and the more likely the object is electronic waste with an intact internal structure; the larger the value, the farther the actual weight deviates from the standard weight, and the more likely the object is an empty shell, filler, or severely damaged. Therefore, in the integrity index calculation, the form (1 - weight deviation rate) is used, so that the weight deviation rate and the integrity index are negatively correlated.

[0031] Specifically, in step S2, determining the recycling type of the electronic and electrical waste to be recycled based on the integrity index includes: If the integrity index is greater than or equal to the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be the whole recycling type; If the integrity index is less than the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be dismantling and recycling.

[0032] In this embodiment of the invention, the preset integrity index is an empirical threshold pre-set based on historical recycling data and actual production needs, used to distinguish between complete equipment and equipment requiring dismantling. The specific method for determining this threshold is as follows: a certain number of historical electronic and electrical waste samples are collected, their integrity indices are calculated, and the actual recycling type of each sample is determined through manual labeling or actual processing results; through statistical analysis or machine learning methods, the threshold that optimally distinguishes the two types of samples is found as the preset integrity index. In practical applications, a typical value is between 0.6 and 0.8, for example, it can be set to 0.7.

[0033] It should be noted that the integrity index defined in this technical solution ranges from 0 to 1, with higher values ​​indicating that the object is closer to a complete electronic or electrical device. When the integrity index is greater than or equal to the preset threshold, it indicates that the electronic or electrical waste to be recycled has a regular outline and a weight close to the standard, and is likely a structurally complete device containing non-ferrous metal components that require fine dismantling for effective recycling. Therefore, it is classified as a whole-unit recycling type and enters the fine dismantling process. When the integrity index is less than the preset threshold, it indicates that the electronic or electrical waste to be recycled may have a broken outline, abnormal weight, etc., and is likely to be broken material, an empty shell, or a component made of a single material. Direct fine dismantling would be inefficient, so it is classified as a dismantling and recycling type and enters the crushing and sorting process.

[0034] This invention fuses visual image data with weight data for judgment. First, it obtains contour regularity based on edge detection and contour extraction, then calculates the weight deviation rate using standard weight ranges in a pre-stored database. Finally, it weights and fuses the two data to obtain an integrity index, which is compared with a preset threshold to accurately distinguish between whole-item recycling and dismantled-item recycling types, thus achieving intelligent diversion of electronic and electrical waste. This method only requires conventional image acquisition and weighing equipment, eliminating the need for complex and expensive sensors, and has the advantages of low cost and simple implementation. Contour regularity effectively identifies the shape integrity of objects, while weight deviation rate effectively identifies the internal structural integrity of objects; the two complement each other for verification. This method avoids the shortcomings of relying solely on visual judgment, which is easily affected by surface patterns, or relying solely on weight, which is easily affected by impurities. It significantly improves the accuracy and robustness of recycling type determination. For complete equipment determined to be of the whole recycling type, subsequent fine disassembly can be carried out to extract high-value non-ferrous metal components. For crushed materials determined to be of the disassembly recycling type, they can directly enter the efficient crushing and sorting process. This achieves differentiated treatment of different types of waste, avoiding both the value loss caused by directly crushing complete equipment and the efficiency waste caused by feeding crushed materials into fine disassembly. As a result, it improves the overall recovery rate and efficiency of non-ferrous metals and has significant industrial application value.

[0035] Specifically, in step S3, locating several disassembly areas based on the image data, and determining the disassembly order of the several disassembly areas based on the degree of damage and color saturation parameters of the several disassembly areas, includes: Step S31: Perform semantic segmentation on the image data to locate several dismantling areas of the electronic and electrical waste to be recycled; Step S32: Determine the damage level index of each disassembly area based on the edge jaggedness, surface crack density, and missing area ratio of each disassembly area. Step S33: Convert the image data to the HSV color space, extract the saturation channel data of each of the disassembled regions, and determine the average saturation value of each of the disassembled regions as the color saturation parameter of each of the disassembled regions. Step S34: Determine the priority score for each dismantling area based on the damage level index and color saturation parameter; Step S35: Sort the several disassembly areas in descending order of priority score to generate the region disassembly order of the disassembly areas.

[0036] In this embodiment of the invention, the semantic segmentation specifically employs a deep learning-based semantic segmentation model, such as Mask R-CNN or DeepLab series models. This model is pre-trained using a large dataset of electronic and electrical waste images labeled with dismantling areas, enabling it to accurately identify and segment each dismantled area on the electronic and electrical waste to be recycled, including panel areas, interface areas, heat dissipation areas, connector areas, etc. Through semantic segmentation, each pixel is assigned a category label, and connected pixel regions of the same category constitute a dismantling region. Simultaneously, the boundary contour and pixel-level position information of each region can be obtained.

[0037] The specific method for determining the degree of damage index is as follows: For each segmented region, firstly, its edge jaggedness is calculated, that is, Fourier transform is performed on the boundary contour of the region to extract the proportion of high-frequency components. The higher the high-frequency components, the more uneven the edge and the more obvious the jaggedness. Secondly, the surface crack density is calculated, that is, crack detection is performed on the image inside the region, and the proportion of the number of detected crack pixels to the total number of pixels in the region is counted. Then, the missing area ratio is calculated, that is, void detection is performed on the region to identify the missing parts caused by damage, and the area of ​​the missing parts is calculated as an estimated proportion of the area of ​​the original complete region. The above three indicators are weighted and summed to obtain the degree of damage index of the segmented region, with a value range between 0 and 1. The larger the value, the more severe the damage.

[0038] The specific method for determining the color saturation parameter is as follows: the original RGB image of the electronic and electrical waste to be recycled is converted to the HSV color space, which includes three channels: hue (H), saturation (S), and brightness (V). For each segmented area, the values ​​of all pixels in the saturation (S) channel within that area are extracted, and the arithmetic mean of these values ​​is calculated as the color saturation parameter for that area. The saturation parameter ranges from 0 to 1; a lower value indicates that the color is closer to grayscale or metallic, while a higher value indicates that the color is more vibrant. Since non-ferrous metals such as gold, silver, copper, and aluminum typically exhibit low saturation characteristics, a lower color saturation parameter indicates a higher probability that the area contains non-ferrous metals.

[0039] The specific method for determining the priority score is as follows: The damage severity index and color saturation parameter are calculated comprehensively. The damage severity index is positively correlated with the priority score, meaning that areas with more severe damage are prioritized for disassembly because the damaged area provides a natural entry point for disassembly. The color saturation parameter is negatively correlated with the priority score, meaning that lower saturation indicates more obvious metallic characteristics, making it more worthy of priority disassembly. The specific calculation formula is: the priority score equals the damage severity index multiplied by its weighting coefficient, plus one minus the color saturation parameter multiplied by its weighting coefficient. The sum of the weighting coefficients of the damage severity index and the color saturation parameter is 1, and these values ​​are adjustable according to the actual application scenario. A typical value is 0.5 for both the damage severity index and the color saturation parameter.

[0040] This invention accurately locates various dismantling areas on electronic and electrical waste using semantic segmentation technology, and prioritizes each area based on damage severity indicators and color saturation parameters, achieving intelligent planning of dismantling areas. The damage severity indicator comprehensively quantifies the physical damage state of the area by considering three dimensions: edge jaggedness, surface crack density, and missing area ratio. Areas with more severe damage are prioritized for dismantling, making full use of damaged areas as natural dismantling entry points and reducing dismantling difficulty. The color saturation parameter is extracted based on the HSV color space, effectively identifying the low saturation characteristics of non-ferrous metals, prioritizing the dismantling of areas with obvious metal characteristics, and ensuring priority recycling of high-value components. The priority score obtained by weighted fusion of these two parameters balances the feasibility of dismantling and the value of recycling, significantly improving dismantling efficiency and non-ferrous metal recovery rate, and has significant industrial application value.

[0041] Specifically, in step S4, determining the candidate disassembly points of the disassembly area includes: Step S411: Local feature extraction is performed on the image data of the disassembly area, and a corner detection algorithm is used to identify several candidate disassembly points within the disassembly area; Step S412: For each candidate disassembly point, extract the geometric features of each candidate disassembly point. Step S413: Based on the geometric features, perform type identification for each candidate disassembly point; Step S414: Record the type identifier of each candidate dismantling point, its two-dimensional coordinates in the image, and its three-dimensional coordinates obtained through three-dimensional reconstruction, and generate an information set for each candidate dismantling point.

[0042] In this embodiment of the invention, the corner detection algorithm specifically employs either the Harris corner detection algorithm or the Shi-Tomasi corner detection algorithm. Taking the Harris corner detection algorithm as an example, its basic principle is to identify corners by calculating the intensity of grayscale changes in an image window in various directions. When the grayscale changes in all directions are significant, the center point of the window is considered a corner. In specific implementation, the image of the disassembly area is first converted to grayscale, the gradient magnitude and direction of each pixel are calculated, a structure tensor matrix is ​​constructed, and then the corner response value of each pixel is calculated. By setting a response threshold and non-maximum suppression, the final corners are selected as candidate disassembly points. These corners typically correspond to structural points with significant geometric features, such as screw holes, buckle edges, and interface corners.

[0043] The extraction of geometric features includes the following: for each detected candidate disassembly point, multi-dimensional geometric features within its local neighborhood are extracted, specifically including the curvature change rate of the point, obtained by calculating the gradient change between the point and surrounding pixels; local edge density, which is the number of edge pixels within a preset radius around the point; indentation depth, obtained through 3D reconstruction to calculate the depth information of the point and its height difference with the surrounding plane; and scale information, which is determined through multi-scale spatial analysis to identify the actual physical dimensions of the structure corresponding to the point. These geometric features together constitute the feature vector of the point, used for subsequent type identification.

[0044] The type identification specifically employs a machine learning-based classifier, such as a support vector machine, random forest, or lightweight neural network. A large number of labeled disassembly point samples are pre-collected. Each sample contains the extracted geometric feature vector and a corresponding type label, including screws, clips, adhesive points, weld points, interfaces, etc. The trained classifier can output a probability distribution of the type to which the candidate disassembly point belongs based on the input feature vector, and the type with the highest probability is taken as the identification result for that point. Points that are difficult to classify clearly can be marked as unknown types, and subsequently verified manually or processed specially.

[0045] The 3D reconstruction specifically employs stereo vision or multi-view 3D reconstruction technology. For the disassembly area, the 2D coordinates of each candidate disassembly point in different images are extracted using images acquired from multiple perspectives. The 3D spatial coordinates of each point are calculated using feature matching and triangulation principles. Alternatively, a structured light or time-of-flight depth camera can be used to directly acquire the scene's depth map, which is then combined with the 2D image coordinates to obtain the 3D coordinates. The recorded point information set includes: a unique point identifier, a type identifier, 2D image coordinates (u, v), 3D spatial coordinates (x, y, z), and an optional point location confidence score.

[0046] Specifically, in step S4, the damage risk parameters of the candidate dismantling points are determined based on the image data, including: Step S421: For each candidate disassembly point, identify non-ferrous metal components within a preset range adjacent to each candidate disassembly point based on the image data, and determine the type, size, and spatial orientation of the non-ferrous metal components. Step S422: Based on the information set of the candidate disassembly points, determine the force direction and stress propagation range when disassembling the candidate disassembly point; Step S423: Determine the spatial distance between the candidate disassembly point and each adjacent non-ferrous metal component; Step S424: Determine the force direction influence coefficient based on the angle between the force direction and the direction pointing towards the non-ferrous metal component; Step S425: Determine the stress overlap coefficient based on the degree of overlap between the stress propagation range and the spatial position of the non-ferrous metal component; Step S426: Based on the spatial distance, the force direction influence coefficient, and the stress overlap coefficient, determine the damage risk parameters of the candidate dismantling points.

[0047] In this embodiment of the invention, the identification of non-ferrous metal components specifically employs a deep learning-based target detection and segmentation model. First, the image region within a preset range surrounding the candidate dismantling point is magnified and analyzed. The preset range can be set as a circular region with a radius of 3 to 5 centimeters or a square region with a side length of 5 to 8 centimeters, depending on the actual situation. The local image is input into a trained instance segmentation model, such as Mask R-CNN or YOLACT. This model is trained using a large dataset of labeled images of non-ferrous metal components and can accurately identify and segment various non-ferrous metal components within the region, including copper pins, gold-plated contacts, aluminum heat sinks, and solder joints. The model outputs the mask contour, category label, and confidence score for each non-ferrous metal component. Based on the mask contour, the dimensions of each component, such as area, equivalent diameter, and aspect ratio, can be calculated. Based on the mask's position in the image combined with depth information or 3D reconstruction results, the spatial orientation of each component, including its center point coordinates and orientation, can be determined.

[0048] The specific calculation method for the spatial distance is as follows: For each candidate disassembly point, calculate the Euclidean distance between its three-dimensional spatial coordinates and the center point or the nearest point on the surface of each adjacent non-ferrous metal component. If the non-ferrous metal component has a precise mask profile and three-dimensional point cloud, the minimum distance from the point to all points on the component surface can be calculated as the spatial distance. The smaller the spatial distance, the closer the point is to the non-ferrous metal component, and the higher the risk of damaging the component during disassembly.

[0049] The specific method for determining the force direction influence coefficient is as follows: For each adjacent non-ferrous metal component, calculate the direction vector pointing from the candidate dismantling point to the center of the component; then calculate the cosine of the angle between this direction vector and the force direction vector determined in step S422. If the angle is less than 90 degrees, the force direction influence coefficient is equal to the cosine of the angle, ranging from 0 to 1, indicating that a certain component of the force direction points towards the component; if the angle is greater than or equal to 90 degrees, the force direction influence coefficient is 0, indicating that the force direction is away from or perpendicular to the component, and will basically not cause direct impact to the component. The smaller the angle, the larger the force direction influence coefficient, and the higher the risk of damage.

[0050] The specific method for determining the stress overlap coefficient is as follows: Determine whether the stress propagation range determined in step S422 covers the spatial region where the non-ferrous metal component is located. Model the stress propagation range as a spatial region, such as a sphere, cylinder, or a specific shaped envelope; also model the non-ferrous metal component as a spatial region or a set of spatial points; calculate the intersection volume or degree of overlap between the two spatial regions. If there is overlap, the stress overlap coefficient is 1 or a continuous value between 0 and 1 depending on the degree of overlap; if there is no overlap, the value is 0. The larger the stress overlap coefficient, the more likely the stress during disassembly is to be transmitted to the non-ferrous metal component, and the higher the risk of damage.

[0051] The specific method for determining the damage risk parameter is as follows: For each candidate dismantling point, considering the influence of all adjacent non-ferrous metal components, the damage risk parameter is calculated according to the following formula: The damage risk parameter is equal to the sum of the values ​​of each adjacent non-ferrous metal component. The contribution of each component is equal to the component's value weight divided by the spatial distance, multiplied by the force direction influence coefficient, and multiplied by the stress overlap coefficient. The value weight is set according to the type of non-ferrous metal component, for example, gold has a weight of 1.0, silver 0.8, copper 0.6, and aluminum 0.4. The smaller the spatial distance, the larger the force direction influence coefficient, the larger the stress overlap coefficient, and the higher the component value, the larger the damage risk parameter, indicating a higher risk of damaging high-value non-ferrous metal components when dismantling that point. The calculated damage risk parameter needs to be normalized so that its value ranges between 0 and 1, facilitating subsequent integration with the structural complexity parameter.

[0052] Specifically, in step S422, determining the force direction and stress propagation range when disassembling a candidate disassembly point based on the information set of the candidate disassembly points includes: Step S4221: Based on the type identifier of the candidate disassembly point, query the preset disassembly process database to obtain the force direction corresponding to the type of the candidate disassembly point; Step S4222: Based on the geometric characteristics of the candidate disassembly point, the force direction is corrected to obtain the actual force direction of the candidate disassembly point; Step S4223: Based on the type identifier and geometric characteristics of the candidate disassembly point, query the preset stress propagation model database to obtain the stress propagation radius corresponding to the type of the candidate disassembly point; Step S4224: Determine the stress propagation range of the candidate disassembly point, centered on the candidate disassembly point and within the range of the stress propagation radius.

[0053] In this embodiment of the invention, the disassembly process database is a pre-built knowledge base that stores standard force direction information corresponding to various disassembly point types. This database is established based on experimental research and mechanical analysis of a large number of electronic and electrical waste disassembly processes, covering common point types such as screws, clips, adhesive joints, welded joints, and interfaces. For screws, the standard force direction is the rotational direction around the screw axis, specifically including clockwise rotation (loosening) and counterclockwise rotation (tightening). Disassembly is usually performed by loosening, therefore the standard force direction is the clockwise tangential direction. For clips, based on their structure, they can be divided into press-type clips and pry-type clips. The standard force direction for press-type clips is the downward pressure direction perpendicular to the clip surface, while the standard force direction for pry-type clips is the prying force direction along the clip arm. For adhesive joints, the standard force direction is the tearing force direction normal to the adhesive surface. For welded joints, the standard force direction may be the shear force direction or the peeling force direction depending on the welding method. The database stores the standard force direction for each type in vector form, along with a description of that direction relative to the local coordinate system of the point.

[0054] The correction of the force direction is specifically based on the geometric features of the candidate disassembly point, including the normal direction, local surface orientation, indentation depth, and surrounding structural constraints. First, the precise normal direction of the point is obtained through 3D reconstruction; this direction is perpendicular to the local surface where the point is located. Then, the obtained standard force direction is transformed from the local coordinate system to the global coordinate system, considering the actual spatial orientation of the point. For example, for a screw point, the standard force direction is the tangential direction around the screw axis, and the screw axis direction is the normal direction of the point. Therefore, the actual force direction is a specific direction in the tangential plane perpendicular to the normal direction, and the specific direction depends on the screw's rotation direction. For a latch point, the actual force direction needs to be adjusted according to the actual orientation of the latch and surrounding spatial constraints. If there are obstacles around the latch, the force direction may need to avoid the direction of the obstacles. The correction process is achieved through a geometric transformation matrix, multiplying the standard force direction vector by the rotation matrix of the point's orientation to obtain the force direction vector in actual space.

[0055] The stress propagation model database is a pre-built knowledge base that stores stress propagation radius information for various disassembly point types under different geometric features. This database is built based on finite element analysis and actual disassembly experimental data, considering factors such as point type, point size, and surrounding material properties. For screw points, the stress propagation radius mainly depends on the screw diameter and tightening torque, with a typical range of 2 to 5 times the screw diameter. For example, a screw with a diameter of 4 mm has a stress propagation radius of approximately 8 to 20 mm. For snap-fit ​​points, the stress propagation radius depends on the length of the snap-fit ​​arm and the material's elastic modulus, with a typical range of 0.5 to 1.5 times the snap-fit ​​arm length. For bonded points, the stress propagation radius depends on the bond area and bond strength, with a typical range of 1 to 2 times the equivalent diameter of the bonded area. The database stores reference stress propagation radii corresponding to different point types and different geometric size ranges, and provides methods for interpolation or adjustment based on actual geometric features.

[0056] The determination of the stress propagation range is based on the candidate disassembly point as the center and the modified stress propagation radius as the radius, constructing a spatial region to represent the range affected by stress when disassembling that point. The shape of this spatial region can be adjusted according to the type of point. For screw points, a spherical region is usually used because the stress generated by the rotational torque propagates uniformly in all directions; for snap-fit ​​points, an ellipsoidal or cylindrical region can be used, elongated along the direction of force to reflect the characteristic of stress propagation along the snap-fit ​​arm; for adhesive points, a flattened ellipsoidal region can be used, compressed along the normal direction of the adhesive surface to reflect the characteristic of stress concentration near the adhesive interface.

[0057] Specifically, in step S4, determining the structural complexity parameters of the candidate disassembly points based on the image data includes: Step S431: For each candidate dismantling point, extract local image blocks within a preset range around the candidate dismantling point based on the image data, and perform edge detection on the local image blocks to obtain an edge distribution map; Step S432: Based on the edge distribution map, determine the edge density around the candidate disassembly point; Step S433: Perform texture analysis on the local image patch, extract the texture features of the local image patch, and calculate the texture entropy value; Step S434: Based on the image data, obtain the three-dimensional point cloud data of the area surrounding the candidate dismantling point through three-dimensional reconstruction, and determine the mean square error of the local surface curvature of the area surrounding the candidate dismantling point as a geometric complexity index. Step S435: Identify the number of components and the connection relationship between components within a preset range around the candidate disassembly point, and determine the component connection density; Step S436: Determine the structural complexity parameters of the candidate disassembly points based on the edge density, the texture entropy value, the geometric complexity index, and the component connection density.

[0058] In this embodiment of the invention, the extraction of the local image patch is centered on the candidate disassembly point, taking an image region with a preset side length or radius. The preset size can be set according to the actual application scenario, typically a square region with a side length of 64 to 128 pixels or a circular region with a radius of 32 to 64 pixels. This local image patch contains environmental information around the point, which is used for subsequent edge detection and texture analysis.

[0059] The edge detection specifically employs either the Canny edge detection algorithm or the Sobel edge detection algorithm. Taking the Canny algorithm as an example, firstly, Gaussian filtering is applied to local image patches to smooth them and reduce image noise; then, the gradient magnitude and direction of each pixel are calculated; next, non-maximum suppression is performed, retaining pixels with the largest local gradient magnitude as candidate edge points; finally, through double threshold detection and edge connection, the final binary edge image, i.e., the edge distribution map, is obtained. In the edge distribution map, white pixels represent edge positions, and black pixels represent non-edge regions.

[0060] The specific method for determining the edge density is as follows: The number of edge pixels in the edge distribution map is counted, and then divided by the total number of pixels in the local image block to obtain the edge density. The edge density value ranges from 0 to 1. A larger value indicates richer edges and a more complex structure in the region; a smaller value indicates sparser edges and a simpler structure in the region. For example, an area full of screws and interfaces has a higher edge density, while a smooth plastic panel area has a lower edge density.

[0061] The texture analysis specifically employs the gray-level co-occurrence matrix (GLCM) method. First, local image patches are converted to grayscale images. Then, the GLCM is calculated, which statistically represents the probability of a specific gray-level combination occurring at a specific direction and distance. Based on the GLCM, texture entropy is calculated. Texture entropy is a measure of the randomness and complexity of image texture, calculated as the sum of the negative values ​​of each element in the GLCM multiplied by its logarithm. A larger texture entropy value indicates a more complex and random image texture; a smaller value indicates a more regular and uniform image texture. For example, a plastic casing area with regular texture has a smaller texture entropy value, while an area filled with various components and wiring has a larger texture entropy value.

[0062] The 3D point cloud data is acquired using stereo vision matching or direct acquisition via a depth camera. For the area surrounding the candidate disassembly point, 3D point cloud data is obtained through multi-view image matching or depth map mapping. Each point in the point cloud contains 3D spatial coordinate information. Based on the point cloud data, local surface curvature is calculated: for each point, a local plane or surface is fitted using its neighboring points to calculate the principal curvature or average curvature of that point. Then, the curvature values ​​of all points within the region are statistically analyzed, and the root mean square error (RMSE) of these curvature values ​​is calculated as an indicator of geometric complexity. A larger MRMSE indicates more dramatic surface undulations and a more complex geometric structure; a smaller MRMSE indicates a flatter surface and a simpler geometric structure.

[0063] The identification of the number of components specifically employs object detection or instance segmentation models to analyze local image blocks and identify the different components contained within them, such as screws, clips, interfaces, chips, capacitors, connectors, etc. The number of identified components is then counted. The connection relationships between components are determined by analyzing their spatial adjacency and relative positions, such as whether adjacent components are in contact, have interlocking relationships, or show welding marks. Component connection density is defined as the sum of the number of components and the number of connection points between components divided by the area of ​​the local image block, or as the total number of edges in the adjacency graph of the components divided by the area. Higher component connection density indicates a denser concentration of components and more complex interconnections in the area, making disassembly more difficult.

[0064] The specific method for determining the structural complexity parameter is as follows: Edge density, texture entropy, geometric complexity index, and component connection density are normalized to a range of 0 to 1. Then, they are weighted and fused according to preset weight coefficients to obtain the structural complexity parameter of the candidate disassembly point. The weight coefficients can be adjusted according to the actual application scenario; typical values ​​are: edge density weight 0.2, texture entropy weight 0.3, geometric complexity index weight 0.3, and component connection density weight 0.2. The structural complexity parameter ranges from 0 to 1; a larger value indicates a more complex surrounding structure and greater disassembly difficulty; a smaller value indicates a simpler structure and easier disassembly.

[0065] Specifically, in step S4, the dismantling order of the candidate dismantling points is determined based on the damage risk parameters and structural complexity parameters, including: Step S441: For each candidate disassembly point, the damage risk parameter and the structural complexity parameter are normalized respectively to obtain the normalized damage risk value and the normalized structural complexity value. Step S442: Based on the preset weight coefficients, the dismantling priority score of each candidate dismantling point is determined according to the normalized damage risk value and the normalized structural complexity value. Step S443: Sort all candidate disassembly points in the disassembly area according to the disassembly priority score from high to low, and generate the disassembly order of the candidate disassembly points.

[0066] In this embodiment of the invention, the specific method of normalization is as follows: For all candidate disassembly points within the same disassembly area, the maximum and minimum values ​​of their damage risk parameters and structural complexity parameters are calculated respectively. For each candidate disassembly point, its normalized damage risk value is equal to the damage risk parameter of that point minus the minimum damage risk parameter, and then divided by the difference between the maximum and minimum damage risk parameters, so that the normalized damage risk value ranges between 0 and 1. The normalized structural complexity value is calculated similarly. Through normalization, the differences in dimensions and numerical ranges between different parameters are eliminated, enabling them to be weighted and fused on the same scale.

[0067] The relationship between the damage risk parameter, structural complexity parameter, and dismantling priority is as follows: A higher damage risk parameter indicates a greater risk of damaging adjacent non-ferrous metal components during dismantling. From a safety perspective, high-risk locations need to be prioritized to avoid accidental triggering of risks in subsequent operations. Therefore, the damage risk parameter is positively correlated with dismantling priority. Conversely, a higher structural complexity parameter indicates a more complex surrounding structure and greater dismantling difficulty. From an efficiency perspective, complex locations may require more time and delicate operations, and are usually handled later to avoid blocking the dismantling of simpler locations. Therefore, the structural complexity parameter is negatively correlated with dismantling priority. Both parameters together determine the dismantling order of the locations.

[0068] The specific method for setting the weighting coefficients is as follows: Based on the different emphases on disassembly safety and efficiency in actual application scenarios, adjust the weights of the damage risk parameter and the structural complexity parameter, with their sum being 1. A typical value is a damage risk parameter weight of 0.6 and a structural complexity parameter weight of 0.4, reflecting the principle of prioritizing safety. In some scenarios with extremely high efficiency requirements, the weighting can be adjusted to a damage risk parameter weight of 0.4 and a structural complexity parameter weight of 0.6, prioritizing the disassembly of simpler points to accelerate the disassembly process. The weighting coefficients can be optimized through statistical analysis of historical disassembly data or expert experience.

[0069] The specific formula for calculating the dismantling priority score is as follows: the dismantling priority score equals the normalized damage risk value multiplied by the damage risk parameter weight, plus a subtraction of the normalized structural complexity value multiplied by the structural complexity parameter weight. By subtracting the structural complexity value, the contribution of this factor decreases as the structural complexity value increases, thus achieving a design where structural complexity and priority are negatively correlated. The calculated dismantling priority score ranges from 0 to 1, with a higher score indicating that the location should be dismantled more preferentially.

[0070] This invention determines the dismantling order of candidate dismantling points by fusing damage risk parameters and structural complexity parameters, achieving an optimized balance between safety and efficiency in the dismantling process. The damage risk parameter comprehensively considers the spatial distance between the point and adjacent non-ferrous metal components, the influence coefficient of the force direction, and the stress overlap coefficient, and introduces the value weight of different metal types. This accurately quantifies the risk of damage to high-value non-ferrous metals such as gold, silver, and copper during dismantling operations, ensuring that high-risk points are prioritized to avoid accidental damage to valuable components. The structural complexity parameter integrates four dimensions: edge density, texture entropy value, local surface curvature root mean square error, and component connection density. This comprehensively quantifies the complexity of the structure surrounding the point, prioritizing the dismantling of points with simple structures to improve dismantling efficiency and avoiding complex points from blocking the overall process. The two parameters are normalized and weighted to obtain a dismantling priority score, ensuring both the basic principle of safety priority and the practical need for maximizing efficiency, significantly reducing the damage rate of non-ferrous metals and increasing the overall value of the recycled materials.

[0071] Specifically, the method further includes: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is the whole recycling type, the electronic and electrical waste to be recycled is transported to the crushing and sorting system for crushing and sorting to recover non-ferrous metals.

[0072] In this embodiment of the invention, the crushing and sorting system includes a coarse crushing unit, a fine crushing unit, a screening unit, and a multi-sensor fusion sorting unit. When the electronic and electrical waste to be recycled is determined to be of the whole-unit recycling type, although it has a regular outline and a weight close to the standard, it may be an empty shell, filler, or low-value intact equipment, which is not suitable for fine dismantling resources, and therefore directly enters the efficient crushing and sorting process.

[0073] In practice, the electronic and electrical waste to be recycled is first transported to the coarse crushing unit, where it is initially crushed using a twin-shaft shear crusher or hammer crusher, breaking large pieces of waste into blocks of 50 to 100 millimeters. The coarsely crushed material is then conveyed by a belt conveyor to the fine crushing unit, where it undergoes secondary crushing using a high-speed rotating hammer mill or impact crusher, further breaking the material into particles of 10 to 30 millimeters. This process effectively separates components of different materials, creating conditions for subsequent sorting.

[0074] The crushed material enters the screening unit, where it is classified by particle size using a vibrating screen or a multi-layer grading screen. It is typically divided into three grades: fine particles (less than 10 mm), medium particles (10 mm to 30 mm), and coarse particles (greater than 30 mm). Materials of different particle sizes are then transported to their respective sorting equipment for further processing to improve sorting efficiency and accuracy.

[0075] The multi-sensor fusion sorting unit integrates various sorting technologies: For medium to coarse particles, a magnetic separator first separates ferromagnetic substances; then, an eddy current separator uses an alternating magnetic field to generate eddy currents in non-ferromagnetic metals, causing non-ferromagnetic metal particles to be separated from the material flow by repulsion, primarily recovering aluminum and copper; the remaining material enters an X-ray fluorescence separator, which accurately separates different types of non-ferromagnetic metal particles, such as those containing copper, zinc, and tin, by identifying the characteristic X-rays of different elements. For fine particles, an airflow separation combined with electrostatic separation is used, utilizing differences in density and conductivity to separate metals from non-metals. The entire sorting process is monitored and parameters are adjusted in real time by a central control system to ensure the recovery rate and purity of non-ferrous metals.

[0076] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A smart recycling method for recovering non-ferrous metals from industrial electronic and electrical waste, characterized in that, include: Step S1: Obtain image data and weight data of the electronic and electrical waste to be recycled; Step S2: Based on the image data, obtain the contour regularity of the electronic and electrical waste to be recycled. Based on the contour regularity and combined with the weight data of the electronic and electrical waste to be recycled, determine the integrity index of the electronic and electrical waste to be recycled. Based on the integrity index, determine the recycling type of the electronic and electrical waste to be recycled, wherein the recycling type includes whole recycling type and dismantling recycling type. Step S3: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is dismantling recycling, several dismantling areas are located based on the image data, and the dismantling order of the several dismantling areas is determined based on the degree of damage and color saturation parameters of the several dismantling areas. Step S4: Determine candidate disassembly points for several disassembly areas, determine the damage risk parameters and structural complexity parameters of the candidate disassembly points based on the image data, and determine the disassembly order of the candidate disassembly points according to the damage risk parameters and structural complexity parameters. Step S5: Disassemble the electronic and electrical waste to be recycled based on the disassembly sequence of the area and the disassembly sequence of the location.

2. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S2, the process of obtaining the contour regularity of the electronic and electrical waste to be recycled based on the image data, and determining the integrity index of the electronic and electrical waste to be recycled based on the contour regularity and in combination with the weight data of the electronic and electrical waste to be recycled, includes: Step S21: Perform edge detection and contour extraction on the image data to determine the minimum bounding rectangle of the electronic and electrical waste to be recycled; Step S22: Calculate the ratio of the actual outline area of ​​the electronic and electrical waste to be recycled to the area of ​​the minimum bounding rectangle, and use the ratio as the outline regularity of the electronic and electrical waste to be recycled; Step S23: Obtain the weight data of the electronic and electrical waste to be recycled, and determine the weight deviation rate of the electronic and electrical waste to be recycled by combining it with the standard weight range of this type of electronic and electrical waste pre-stored in the database. Step S24: Determine the integrity index based on the contour regularity and the weight deviation rate.

3. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S2, determining the recycling type of the electronic and electrical waste to be recycled based on the integrity index includes: If the integrity index is greater than or equal to the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be the whole recycling type; If the integrity index is less than the preset integrity index, then the recycling type of the electronic and electrical waste to be recycled is determined to be dismantling and recycling.

4. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S3, locating several disassembly areas based on the image data and determining the disassembly order of the several disassembly areas based on the degree of damage and color saturation parameters of the several disassembly areas includes: Step S31: Perform semantic segmentation on the image data to locate several dismantling areas of the electronic and electrical waste to be recycled; Step S32: Determine the damage level index of each disassembly area based on the edge jaggedness, surface crack density, and missing area ratio of each disassembly area. Step S33: Convert the image data to the HSV color space, extract the saturation channel data of each of the disassembled regions, and determine the average saturation value of each of the disassembled regions as the color saturation parameter of each of the disassembled regions. Step S34: Determine the priority score for each dismantling area based on the damage level index and color saturation parameter; Step S35: Sort the several disassembly areas in descending order of priority score to generate the region disassembly order of the disassembly areas.

5. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 4, characterized in that, In step S4, determining the candidate disassembly points of the disassembly area includes: Step S411: Local feature extraction is performed on the image data of the disassembly area, and a corner detection algorithm is used to identify several candidate disassembly points within the disassembly area; Step S412: For each candidate disassembly point, extract the geometric features of each candidate disassembly point. Step S413: Based on the geometric features, perform type identification for each candidate disassembly point; Step S414: Record the type identifier of each candidate dismantling point, its two-dimensional coordinates in the image, and its three-dimensional coordinates obtained through three-dimensional reconstruction, and generate an information set for each candidate dismantling point.

6. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S4, damage risk parameters for the candidate dismantling points are determined based on the image data, including: Step S421: For each candidate disassembly point, identify non-ferrous metal components within a preset range adjacent to each candidate disassembly point based on the image data, and determine the type, size, and spatial orientation of the non-ferrous metal components. Step S422: Based on the information set of the candidate disassembly points, determine the force direction and stress propagation range when disassembling the candidate disassembly point; Step S423: Determine the spatial distance between the candidate disassembly point and each adjacent non-ferrous metal component; Step S424: Determine the force direction influence coefficient based on the angle between the force direction and the direction pointing towards the non-ferrous metal component; Step S425: Determine the stress overlap coefficient based on the degree of overlap between the stress propagation range and the spatial position of the non-ferrous metal component; Step S426: Based on the spatial distance, the force direction influence coefficient, and the stress overlap coefficient, determine the damage risk parameters of the candidate dismantling points.

7. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 6, characterized in that, In step S422, determining the force direction and stress propagation range when dismantling a candidate dismantling point based on the information set of the candidate dismantling points includes: Step S4221: Based on the type identifier of the candidate disassembly point, query the preset disassembly process database to obtain the force direction corresponding to the type of the candidate disassembly point; Step S4222: Based on the geometric characteristics of the candidate disassembly point, the force direction is corrected to obtain the actual force direction of the candidate disassembly point; Step S4223: Based on the type identifier and geometric characteristics of the candidate disassembly point, query the preset stress propagation model database to obtain the stress propagation radius corresponding to the type of the candidate disassembly point; Step S4224: Determine the stress propagation range of the candidate disassembly point, centered on the candidate disassembly point and within the range of the stress propagation radius.

8. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S4, the structural complexity parameters of the candidate disassembly points are determined based on the image data, including: Step S431: For each candidate dismantling point, extract local image blocks within a preset range around the candidate dismantling point based on the image data, and perform edge detection on the local image blocks to obtain an edge distribution map; Step S432: Based on the edge distribution map, determine the edge density around the candidate disassembly point; Step S433: Perform texture analysis on the local image patch, extract the texture features of the local image patch, and calculate the texture entropy value; Step S434: Based on the image data, obtain the three-dimensional point cloud data of the area surrounding the candidate dismantling point through three-dimensional reconstruction, and determine the mean square error of the local surface curvature of the area surrounding the candidate dismantling point as a geometric complexity index. Step S435: Identify the number of components and the connection relationship between components within a preset range around the candidate disassembly point, and determine the component connection density; Step S436: Determine the structural complexity parameters of the candidate disassembly points based on the edge density, the texture entropy value, the geometric complexity index, and the component connection density.

9. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, In step S4, the dismantling order of the candidate dismantling points is determined based on the damage risk parameters and structural complexity parameters, including: Step S441: For each candidate disassembly point, the damage risk parameter and the structural complexity parameter are normalized respectively to obtain the normalized damage risk value and the normalized structural complexity value. Step S442: Based on the preset weight coefficients, the dismantling priority score of each candidate dismantling point is determined according to the normalized damage risk value and the normalized structural complexity value. Step S443: Sort all candidate disassembly points in the disassembly area according to the disassembly priority score from high to low, and generate the disassembly order of the candidate disassembly points.

10. The intelligent recycling method for recovering non-ferrous metals from industrial electronic and electrical waste according to claim 1, characterized in that, The method further includes: In response to the fact that the recycling type of the electronic and electrical waste to be recycled is the whole recycling type, the electronic and electrical waste to be recycled is transported to the crushing and sorting system for crushing and sorting to recover non-ferrous metals.

Images