Lithium battery step utilization intelligent sorting method based on visual detection

Abstract

The application discloses a lithium battery step utilization intelligent sorting method fusing visual detection, and relates to the technical field of lithium battery step utilization sorting, and comprises the following steps: acquiring basic appearance parameters of a square lithium battery, collecting three-dimensional point cloud of the surface of the square lithium battery, and obtaining battery surface point cloud data; carrying out denoising processing on the battery surface point cloud data to obtain standard surface point cloud data; acquiring actual thickness information based on the standard surface point cloud data, setting an adaptive threshold to carry out surface bulge analysis, and obtaining battery surface bulge information; and sorting the square lithium battery according to the battery surface bulge information of the square lithium battery; the application is used to solve the problem that the existing lithium battery step utilization sorting technology cannot adapt to slight deformation of the retired battery, is prone to detection errors, and causes errors in step sorting when step sorting is carried out based on the bulge state of the square lithium battery by using point cloud data and fixed threshold detection.

Description

Technical Field

[0001] This invention relates to the field of lithium battery cascade utilization sorting technology, specifically to an intelligent sorting method for lithium battery cascade utilization that integrates visual inspection. Background Technology

[0002] Lithium-ion battery cascade utilization sorting technology refers to a complete technical system that uses technologies such as appearance inspection, electrical performance testing, health status assessment and intelligent clustering to conduct safety screening, performance grading and value classification of power batteries retired from new energy vehicles, energy storage systems and other scenarios. This system accurately matches batteries in different states to different cascade application scenarios, while eliminating batteries with safety hazards or no cascade utilization value and putting them into the recycling process.

[0003] Existing lithium battery tiered utilization sorting technologies often rely on manual visual inspection to identify and sort bulges when sorting prismatic lithium batteries based on their bulging state. However, manual inspection is not very accurate, easily missing small bulges, and different inspectors have inconsistent judgment standards, making it difficult to guarantee sorting consistency. Furthermore, manually inspecting a single battery takes a long time, resulting in low overall sorting efficiency. Existing methods typically also utilize 3D vision to collect point cloud data of the prismatic lithium battery surface, thereby obtaining the surface thickness distribution, and then using a fixed thickness threshold to identify and sort bulges. However, a fixed thickness threshold cannot accommodate the slight deformation commonly found in retired batteries. It is easy to misjudge normal slight overall deformation as local bulges, and it is also easy to miss tiny bulges superimposed on the overall deformation. Furthermore, the fixed thickness threshold ignores the local differences in different areas of the battery surface, which can easily lead to a large number of false detections in some areas with large differences, such as scratched areas. Therefore, when the existing lithium battery cascade utilization sorting technology performs cascade sorting based on the bulging state of square lithium batteries, it uses point cloud data and fixed thresholds to detect bulges, which cannot accommodate the slight deformation of retired batteries, making it easy to detect errors and causing mistakes in cascade sorting. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in the prior art. It obtains basic appearance parameters of a square lithium battery and collects three-dimensional point clouds of the battery's surface to obtain battery surface point cloud data. Noise reduction processing is then performed to obtain standard surface point cloud data. Based on this standard surface point cloud data, actual thickness information is obtained, and an adaptive threshold is set to analyze surface bulging, yielding battery surface bulging information. The square lithium batteries are then sorted for secondary use. This addresses the problem that existing lithium battery secondary use sorting technologies, when sorting based on the bulging state of square lithium batteries, rely on point cloud data and fixed thresholds to detect bulging. This approach fails to accommodate slight deformations in retired batteries, leading to detection errors and sorting mistakes. To achieve the above objectives, this application provides a smart sorting method for the secondary utilization of lithium batteries that integrates visual inspection, comprising the following steps: The basic appearance parameters of the square lithium battery are obtained, and the three-dimensional point cloud of the surface of the square lithium battery is collected to obtain the point cloud data of the battery surface. The battery surface point cloud data is denoised, and outliers are filtered and removed to obtain standard surface point cloud data. The actual thickness information is obtained based on standard surface point cloud data, and surface bulging analysis is performed by setting an adaptive threshold to obtain battery surface bulging information. Based on the information about bulging on the surface of the square lithium batteries, the square lithium batteries are sorted for secondary use.

[0005] Further, the basic appearance parameters of the square lithium battery are obtained, and the three-dimensional point cloud of the surface of the square lithium battery is collected to obtain the battery surface point cloud data, including the following sub-steps: Let any type of square lithium battery be referred to as the first type of battery; let any two parallel surfaces of the first type of battery be referred to as the corresponding surface group; let the corresponding surface groups to be tested be obtained and referred to as corresponding surface group 1 to corresponding surface group n in sequence; where n is the number of corresponding surface groups to be tested. For the corresponding surface group 1, the vertical straight-line distance between the two surfaces in the corresponding surface group 1 is recorded as the surface distance of the corresponding surface group 1, and marked as surface distance 1; the range in which the surface distance 1 of the qualified first type battery should be located is obtained, and recorded as the corresponding qualified range [A0, B0].

[0006] Furthermore, obtaining the basic appearance parameters of the square lithium battery and acquiring the three-dimensional point cloud of the square lithium battery surface to obtain the battery surface point cloud data also includes the following sub-steps: Repeatedly obtain the qualified intervals corresponding to the surface distances of all corresponding surface groups, and record them as the basic appearance parameters of the first type of battery; The two surfaces in the corresponding surface group 1 are respectively designated as surface 11 and surface 12; any one of the first type batteries to be sorted in the tier is designated as the first lithium battery; The three-dimensional point cloud of the surface 11 and surface 12 of the first lithium battery is collected simultaneously using a laser scanning device to obtain the point cloud data of the corresponding surface group 1 of the first lithium battery; the point cloud data of all corresponding surface groups of the first lithium battery are collected repeatedly to obtain the battery surface point cloud data of the first lithium battery.

[0007] Furthermore, the battery surface point cloud data is denoised by filtering and removing outliers to obtain standard surface point cloud data, including the following sub-steps: Establish a spatial rectangular coordinate system, and label the horizontal axis, vertical axis and vertical axis of the spatial rectangular coordinate system as the x-axis, y-axis and z-axis respectively; place the point clouds of surface 11 and surface 12 in the spatial rectangular coordinate system according to their relative relationship, and place them vertically and horizontally parallel to the xy plane of the spatial rectangular coordinate system, and label them as the upper point cloud and the lower point cloud respectively according to their vertical relationship; Obtain the z-axis coordinates of all points in the point cloud above, and calculate the median MA and median absolute deviation MD of all z-axis coordinates; denote any point in the point cloud above as the first point, and label the z-axis coordinates of the first point as AZ; If AZ is not located in [MA-3*MD, MA+3*MD], then mark the first point as a location error point; repeat the filtering of all location error points in the point cloud above and remove them to obtain the corresponding standard point cloud 1; Then, statistical filtering is performed on the standard point cloud 1 above to filter out the corresponding outliers and remove them, resulting in the corresponding standard point cloud 2; this is called the initial point cloud above. The same process is repeated on the point cloud below to obtain the corresponding standard point cloud 2, which is called the initial point cloud below.

[0008] Furthermore, the denoising process for the battery surface point cloud data, which involves filtering and removing outliers to obtain standard surface point cloud data, includes the following sub-steps: Obtain the projections of the upper and lower preliminary screening point clouds onto the xy plane, and divide the region where the projections are located on the xy plane into multiple grids of size a1*a1; the grid that simultaneously contains the projections of the upper and lower preliminary screening point clouds is denoted as a non-empty grid; where a1 is the set grid side length; Denote any non-empty grid as the first grid; considering only the x-axis and y-axis coordinates; obtain the upper and lower preliminary point clouds located in the first grid respectively, and denote them as the upper point set and lower point set of the first grid; Calculate the median z-axis coordinates of the upper and lower point sets respectively, denoted as MZD and MZU. Calculate MZD-MZU, denoted as the reference thickness TB of the first grid. Repeat the collection of reference thicknesses for all non-empty grids.

[0009] Furthermore, the denoising process for the battery surface point cloud data, which involves filtering and removing outliers to obtain standard surface point cloud data, includes the following sub-steps: Calculate the median TA and median absolute deviation TD for all reference thicknesses, and mark non-empty grids whose reference thicknesses are not located in [TA-3*TD, TA+3*TD] as anomalous grids; If the first grid is an abnormal grid, then based on the z-axis coordinate, the interquartile range method is used to filter and remove abnormal points in the upper and lower point sets respectively; repeat the process for all abnormal grids, filter and remove all abnormal points, and after completion, the upper standard point cloud and the lower standard point cloud are obtained, which are recorded as the standard point cloud data of the corresponding surface group 1. Repeatedly acquire standard point cloud data for all corresponding surface groups to obtain standard surface point cloud data.

[0010] Furthermore, based on standard surface point cloud data, actual thickness information is obtained, and an adaptive threshold is set to perform surface bulging analysis. The resulting battery surface bulging information includes the following sub-steps: Ignore the z-axis coordinate, obtain the position of each point in the standard point cloud above, and record it as the thickness point. Record any thickness point as the second point. The distribution is based on the lower standard point cloud and the upper standard point cloud to perform surface fitting, and the lower fitting surface and the upper fitting surface are obtained respectively; the vertical distance between the second point in the lower fitting surface and the upper fitting surface is obtained, and denoted as the actual thickness AF of the second point; Obtain all thickness points in the fitted surface and record them as upper thickness points; if the actual thickness AF of the second point is greater than 1.2*B0 or ​​less than A0, mark the upper thickness point at the corresponding position as an abnormal deformation point. Repeatedly filter all abnormal deformation points in the upper thickness points and remove them. Then, use the remaining upper thickness points to perform surface fitting to obtain the trend surface corresponding to the upper fitted surface, which is denoted as the upper trend surface. Based on the lower fitted surface, repeatedly obtain the trend surface corresponding to the lower fitted surface, which is denoted as the lower trend surface.

[0011] Furthermore, based on standard surface point cloud data, actual thickness information is obtained, and an adaptive threshold is set to perform surface bulging analysis. The resulting battery surface bulging information includes the following sub-steps: Obtain the corresponding point of the second point on the upper trend surface, and take the corresponding point as the endpoint to obtain the vertical displacement from the corresponding point to the upper fitting surface, which is denoted as the upper residual DC of the second point. Obtain the corresponding point of the second point on the lower trend surface, and take the corresponding point as the endpoint to obtain the vertical displacement from the corresponding point to the lower fitting surface, which is denoted as the lower residual UC of the second point. Repeatedly obtain the upper and lower residuals of all thickness points to obtain the upper residual set and the lower residual set respectively; Set three circular windows with radii e1, e2, and e3, respectively, and label them small window, medium window, and large window. Here, e1, e2, and e3 represent the set radii, e1... <e2<e3。

[0012] Furthermore, based on standard surface point cloud data, actual thickness information is obtained, and an adaptive threshold is set to perform surface bulging analysis. The resulting battery surface bulging information includes the following sub-steps: Based on the upper residual set, with the second point as the center of the small window, calculate the median RA and median absolute deviation RD of the upper residuals of all thickness points within the small window; if DC is greater than RA+k1*RD, then mark the second point as a suspected point, where k1 is a set coefficient. Repeat the judgment with the second point as the center of the middle window and the large window. If the second point is marked as a suspected point under any circular window, then mark the second point as the upper bulge point. Based on the lower residual set, the judgment is repeatedly performed with the second point as the center of each circular window. If the second point is marked as a suspected point under any circular window, then the second point is marked as the lower bulge point. Repeatedly obtain all upper and lower drum point locations to obtain upper drum set and lower drum set respectively; Connectivity analysis is performed based on the upper bulge set. Adjacent and continuous upper bulge points are merged to obtain upper bulge regions. The area of ​​each upper bulge region is calculated and denoted as the bulge area. The average value of the upper residual of each upper bulge point in the upper bulge region is denoted as the bulge height. Based on the repeated processing of the lower bulge set, the lower bulge region, as well as the corresponding bulge area and bulge height, are obtained; Record all upper and lower bulge areas, along with their corresponding bulge areas and heights, as the bulge detection information for the corresponding surface group 1; repeat this process for all corresponding surface groups to obtain the battery surface bulge information.

[0013] Furthermore, based on the information regarding the bulges on the surface of the square lithium batteries, the sorting of the square lithium batteries for secondary use includes the following sub-steps: Set sorting criteria corresponding to the number of bulges, bulge area, and bulge height; obtain the number of bulges and the corresponding total bulge area based on the bulge information on the surface of the first lithium battery; Based on the set sorting criteria, the sorting criteria met by the first lithium battery are judged, and the first lithium battery is sorted for secondary use; the battery surface bulge information of the first type of battery is repeatedly obtained, and the corresponding secondary use sorting is performed.

[0014] The beneficial effects of this invention are as follows: This invention obtains the basic appearance parameters of a square lithium battery and collects the three-dimensional point cloud of the battery surface to obtain battery surface point cloud data; it performs noise reduction processing on the battery surface point cloud data, filters and removes abnormal points in the point cloud to obtain standard surface point cloud data; it obtains actual thickness information based on the standard surface point cloud data and sets an adaptive threshold to perform surface bulging analysis to obtain battery surface bulging information; it sorts the square lithium batteries for secondary use based on the battery surface bulging information; when sorting the square lithium batteries based on the bulging state, the threshold can be adaptively set in different areas based on the surface point cloud data, thereby reliably detecting surface bulging and improving the accuracy and reliability of secondary sorting; This invention uses the median absolute deviation to initially screen for location errors, and further uses statistical filtering and outlier grid identification to remove local outliers, thus obtaining standard surface point cloud data. This effectively removes erroneous points from the point cloud, reducing interference with subsequent thickness calculations. Simultaneously, the grid-based local anomaly detection retains the true characteristics of slight deformations on the battery surface, avoiding misclassification of minor deformations as anomalies. The actual thickness is calculated through surface fitting, and anomalous deformation points are removed. Then, trend surface and residual analysis, combined with circular windows of different sizes, are used to establish an adaptive threshold for identifying bulges on the upper and lower surfaces. Compared to direct determination using a fixed thickness threshold, the threshold can be dynamically adjusted according to changes in the local state of the battery surface through local residual statistical features and multi-scale window constraints. This reduces false alarms caused by overall minor deformations, improves the detection rate of small bulges, and provides a more accurate and reliable basis for subsequent sorting. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the steps of the method of the present invention; Figure 2 This is a flowchart illustrating the standard point cloud data acquisition process of the present invention. Figure 3 This is a flowchart of the bulge detection information acquisition process of the present invention; Figure 4 This is a schematic diagram of the electronic device of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, please refer to Figure 1As shown, this application provides a smart sorting method for the secondary utilization of lithium batteries that integrates visual inspection, including the following steps: Step S1 involves obtaining the basic appearance parameters of the square lithium battery and collecting the three-dimensional point cloud of the battery surface to obtain the battery surface point cloud data. Step S1 includes the following sub-steps: Step S101: Any type of square lithium battery is designated as the first type of battery; any two parallel surfaces of the first type of battery are designated as corresponding surface groups; the corresponding surface groups to be tested are obtained and designated as corresponding surface group 1 to corresponding surface group n in sequence; where n is the number of corresponding surface groups to be tested; for example, the upper surface and the lower surface, the left surface and the right surface, and the front surface and the back surface are corresponding surface groups respectively. Step S102: For the corresponding surface group 1, the vertical straight-line distance between two surfaces in the corresponding surface group 1 is recorded as the surface distance of the corresponding surface group 1, and marked as surface distance 1; that is, the thickness between the two surfaces; obtain the range in which the surface distance 1 of the qualified first type battery should be located, and record it as the corresponding qualified range [A0, B0].

[0018] Step S103: Repeatedly obtain the qualified intervals corresponding to the surface distances of all corresponding surface groups and record them as the basic appearance parameters of the first type of battery. Step S104: The two surfaces in the corresponding surface group 1 are successively designated as surface 11 and surface 12; any one of the first type batteries to be sorted in stages is designated as the first lithium battery; Step S105: Simultaneously acquire the three-dimensional point cloud of the surface 11 and surface 12 of the first lithium battery using a laser scanning device to obtain the point cloud data of the corresponding surface group 1 of the first lithium battery; repeatedly acquire the point cloud data of all corresponding surface groups of the first lithium battery to obtain the battery surface point cloud data of the first lithium battery. Simultaneous acquisition can reduce the problems of pose deviation, data misalignment and local missing data caused by acquisition time difference, and improve the matching accuracy between the upper and lower surface point clouds. In practice, lithium battery bulging is one of the most common failure modes of lithium batteries and the most intuitive external manifestation of internal safety hazards. It is an important screening indicator in the tiered utilization sorting process. The batteries to be sorted are usually retired batteries that have been used for a long time. During use, they are easily subjected to various mechanical stresses and environmental factors. In addition to possible battery bulging, they usually have slight deformation. This slight deformation often manifests as overall unevenness on the surface or irregular displacement of local points. If these deformations are ignored, it is easy to cause deviations in subsequent analysis results.

[0019] Step S2 involves denoising the battery surface point cloud data, filtering and removing outliers to obtain standard surface point cloud data. Step S2 includes the following sub-steps: For step S201, please refer to... Figure 2 As shown, a spatial rectangular coordinate system is established, and the horizontal axis, vertical axis and vertical axis of the spatial rectangular coordinate system are respectively denoted as the x-axis, y-axis and z-axis; the point clouds of surface 11 and surface 12 are placed in the spatial rectangular coordinate system according to their relative relationship, and are placed vertically and horizontally parallel to the xy plane of the spatial rectangular coordinate system, and are respectively denoted as the upper point cloud and the lower point cloud according to their vertical relationship; If the original point cloud contains background point clouds or other irrelevant point clouds, the battery body region can be accurately separated from the original point cloud before placing it in the coordinate system to eliminate irrelevant environmental background interference. When placing it, it can be placed horizontally or vertically, and the relative position should be taken into account during subsequent processing.

[0020] Step S202: Obtain the z-axis coordinates of all points in the point cloud above, and calculate the median MA and median absolute deviation MD of all z-axis coordinates; denote any point in the point cloud above as the first point, and mark the z-axis coordinates of the first point as AZ; the distance between the z-axis after placement represents the thickness of the two surfaces. Step S203: If AZ is not located in [MA-3*MD, MA+3*MD], then mark the first point as a location error point; repeat the filtering of all location error points in the point cloud above and remove them to obtain the corresponding standard point cloud 1; use the median and median absolute deviation as statistical benchmarks, which are less sensitive to extreme outliers than the mean and standard deviation, and can effectively adapt to and remove obvious error points in the point cloud caused by scanning reflection, occlusion, isolated noise or acquisition anomalies, avoid a small number of outliers from skewing the overall judgment standard, and improve the reliability of the judgment.

[0021] Step S204: Perform statistical filtering on the standard point cloud 1 above, filter out the corresponding outliers, and remove them to obtain the corresponding standard point cloud 2; denoted as the initial point cloud above; repeat the processing on the point cloud below to obtain the corresponding standard point cloud 2, denoted as the initial point cloud below. The core of statistical filtering is to calculate the average distance from each point to its neighboring points and determine whether it is an outlier through global statistical analysis. It can effectively remove sparsely distributed isolated noise points, which are usually caused by airborne dust, small scratches on the battery surface, or laser speckle. The number of all neighboring points used in statistical filtering and the corresponding judgment threshold can be set according to the distribution characteristics of the point cloud in actual use. For example, for each point, calculate the average distance to its 50 nearest neighbors, denoted as the neighborhood average distance. Then calculate the median and absolute deviation of the median of all neighborhood average distances, resulting in a median of 0.08 mm and an absolute deviation of 0.02 mm. Set the threshold to 0.08 + 2 * 0.02 = 0.12 mm. Mark points with a neighborhood average distance greater than 0.12 mm as outliers and remove them.

[0022] Step S205: Obtain the projections of the upper and lower preliminary screening point clouds onto the xy plane, and divide the area where the projections are located on the xy plane into multiple grids of size a1*a1; the grid that simultaneously has the projections of the upper and lower preliminary screening point clouds is denoted as a non-empty grid; where a1 is the set grid side length; in this embodiment, a1=3mm, which can be flexibly set according to the actual application scenario; Step S206: Denote any non-empty grid as the first grid; considering only the x-axis and y-axis coordinates; obtain the upper and lower preliminary point clouds located in the first grid respectively, and denote them as the upper point set and lower point set of the first grid. By dividing the battery into grids, the thickness of the upper and lower surfaces can be analyzed grid by grid. This makes it easier to capture local anomalies and avoids treating the entire battery as a single unit, thereby improving the ability to identify point clouds near local deformation, scratched areas and small bulges. Step S207: Calculate the median of the z-axis coordinates of the upper and lower point sets respectively, and denote them as MZD and MZU respectively. Calculate MZD-MZU and denote it as the reference thickness TB of the first grid. Repeat the collection of reference thicknesses for all non-empty grids.

[0023] Step S208: Calculate the median TA and median absolute deviation TD of all reference thicknesses, and mark non-empty grids whose reference thickness is not located in [TA-3*TD, TA+3*TD] as abnormal grids; abnormal grids are areas that are obviously inconsistent with the main body state or have abnormal points, and need to be further filtered. Step S209: If the first grid is an anomalous grid, then based on the z-axis coordinate, use the interquartile range method to filter and remove anomalous points in the upper and lower point sets respectively; repeat the process for all anomalous grids, filter and remove all anomalous points, and obtain the upper standard point cloud and the lower standard point cloud, which are recorded as the standard point cloud data of the corresponding surface group 1; the interquartile range method has a strong ability to identify anomalous points that are local outliers. For example, if the first grid is an abnormal grid, for the point set above, obtain the z-axis coordinates of all points in the point set above, and obtain the first quartile Q1, the third quartile Q3, and the interquartile range IQR of all z-axis coordinates. Mark the points whose z-axis coordinates are not in [Q1-1.5*IQR, Q3+1.5*IQR] as abnormal points and remove them to obtain the corresponding standard point cloud above. Step S210: Repeatedly acquire standard point cloud data for all corresponding surface groups to obtain standard surface point cloud data; In the specific implementation process, the initial screening is performed by using the median and the absolute deviation of the median. Then, statistical filtering, grid division, and abnormal grid identification are used to further refine the removal of points in the abnormal grid. This can effectively remove abnormal points in the point cloud while preserving as much of the battery's true slight deformation and local morphological differences as possible, thus providing high-quality standard surface point cloud data for subsequent analysis and processing.

[0024] Step S3 involves obtaining actual thickness information based on standard surface point cloud data and setting an adaptive threshold for surface bulging analysis to obtain battery surface bulging information. Step S3 includes the following sub-steps: For step S301, please refer to... Figure 3 As shown, ignoring the z-axis coordinate, the position of each point in the standard point cloud above is obtained and recorded as the thickness point. Any thickness point is recorded as the second point. That is, the height value of the point is not considered first, but only the positional relationship of the point on the plane is retained. The subsequent thickness calculation is transformed into the calculation of the thickness at a certain plane position. Step S302: The distribution is based on the lower standard point cloud and the upper standard point cloud to perform surface fitting, and the lower fitting surface and the upper fitting surface are obtained respectively; the vertical distance between the second point in the lower fitting surface and the upper fitting surface is obtained and denoted as the actual thickness AF of the second point. Surface fitting can transform discrete point clouds into continuous surfaces, reducing the impact of single-point fluctuations on thickness judgment; the actual thickness is determined by both the upper and lower surfaces, and can more accurately reflect the true thickness of the battery casing at that location.

[0025] Step S303: Obtain all thickness points in the fitted surface and record them as upper thickness points; if the actual thickness AF of the second point is greater than 1.2*B0 or ​​less than A0, mark the upper thickness point at the corresponding position as an abnormal deformation point; less than A0 indicates that there is a high probability of a depression at the location of the point, and greater than 1.2*B0 indicates that there is a high probability of a bulge at the location of the point; 1.2*B0 is to allow a certain upward deviation considering the slight deformation of the battery after long-term use; Step S304: Repeatedly filter all abnormal deformation points in the upper thickness points and remove them. Use the remaining upper thickness points to perform surface fitting to obtain the trend surface corresponding to the upper fitting surface, which is denoted as the upper trend surface. Based on the lower fitting surface, repeatedly obtain the trend surface corresponding to the lower fitting surface, which is denoted as the lower trend surface. The upper and lower trend surfaces represent the trend patterns presented by ignoring the protrusions and depressions on the battery surface, rather than the actual surface morphology that is skewed by local protrusions and depressions. After removing abnormal deformation points and refitting, a trend surface that ignores local bulges can be obtained, providing a benchmark for subsequent residual analysis and improving the detection rate of micro bulges.

[0026] Step S305: obtaining the corresponding point of the second point position on the upper trend surface, taking the corresponding point as an end point, obtaining the vertical displacement from the corresponding point to the upper fitting surface, and recording the vertical displacement as the upper residual DC of the second point position; the vertical displacement has direction, that is, DC has a sign, a positive sign represents a protrusion, and a negative sign represents a depression; the upper residual essentially reflects the protrusion degree of the points on the upper surface relative to the overall trend, so it can reveal local bulge features better than a pure thickness value; the overall deformation and local abnormal protrusion can be separated by the residual, so that tiny bulges can be identified more easily; Step S306: obtaining the corresponding point of the second point position on the lower trend surface, taking the corresponding point as an end point, obtaining the vertical displacement from the corresponding point to the lower fitting surface, and recording the vertical displacement as the lower residual UC of the second point position; the lower residual UC also has a sign; the lower residual reflects the protrusion degree of the points on the lower surface relative to the overall trend; Step S307: repeatedly obtaining the upper residual and lower residual of all thickness point positions, and obtaining an upper residual set and a lower residual set respectively; Step S308: arranging three circular windows with radii of e1, e2 and e3 respectively, which are sequentially recorded as a small window, a medium window and a large window, wherein e1, e2 and e3 are set radii, and e1<e2<e3; in this embodiment, e1=2mm, e2=3mm, e3=5mm, which can be flexibly set according to actual application scenarios; The advantage of arranging multiple circular windows is that it can not only capture fine bulges in a small range, but also cover continuous deformation areas in a large range, thereby avoiding missed detection or false detection caused by a single scale.

[0027] Step S309: based on the upper residual set, taking the second point position as the center of the small window, calculating the median RA and median absolute deviation RD of the upper residuals of all thickness point positions in the small window; if DC is greater than RA+k1*RD, marking the second point position as a suspected point position, wherein k1 is a set coefficient, in this embodiment, k1=2, which can be flexibly set according to actual application scenarios, and different coefficients can be set corresponding to different windows; The threshold RA+k1*RD can be adjusted along with the change of local residual distribution, so it can adapt to the difference of surface in different areas; when a certain area has slight deformation itself, it is not easy to misjudge normal deformation fluctuation as a bulge, and at the same time, it can mark points significantly higher than the local background as suspected point positions, which improves the accuracy of bulge detection; Step S310: repeatedly performing judgment by taking the second point position as the center of the medium window and the large window, if the second point position is marked as a suspected point position under any circular window, marking the second point position as an upper bulge point position; Step S311: Based on the lower residual set, repeat the judgment with the second point as the center of each circular window. If the second point is marked as a suspected point under any circular window, then mark the second point as the lower bulge point; ensure that the judgment criteria for bulges on the upper and lower sides are consistent, and reduce missed detections caused by unilateral deviation. Step S312: Repeatedly obtain all upper and lower drum point positions to obtain upper drum set and lower drum set respectively.

[0028] Step S313: Based on the upper bulge set, perform connectivity analysis, merge adjacent and continuous upper bulge points to obtain upper bulge regions, that is, further connect and aggregate the scattered upper bulge points into regions, and calculate the area of ​​each upper bulge region, denoted as bulge area, and the average value of the upper residual of each upper bulge point in the upper bulge region, denoted as bulge height; bulge area and bulge height can describe the scale and severity of bulges, and can also obtain the maximum value of the upper residual of the upper bulge point, i.e., the maximum bulge height, for subsequent sorting; Step S314: Based on the repeated processing of the lower bulge set, the lower bulge region, as well as the corresponding bulge area and bulge height are obtained; Step S315: Record all upper and lower bulge areas, as well as the corresponding bulge area and bulge height, as bulge detection information for the corresponding surface group 1; repeat the acquisition of all corresponding surface groups to obtain battery surface bulge information; In the actual implementation process, bulges may occur on all surfaces of a square lithium battery. By establishing upper and lower residuals simultaneously, abnormal states on both sides of the battery can be symmetrically identified, avoiding focusing only on one side of the surface and missing local bulges on the other side, which helps to improve the comprehensiveness of the detection.

[0029] Step S4: Based on the bulging information on the surface of the square lithium batteries, sort the square lithium batteries for reuse. Step S4 includes the following sub-steps: Step S401: Set the sorting criteria corresponding to the number of bulges, the area of ​​the bulges, and the height of the bulges; Based on the bulge information on the surface of the first lithium battery, obtain the number of bulges and the corresponding total area of ​​the bulges; The sorting criteria can be set according to actual needs, and can be established from multiple dimensions such as the number of bulges, the range of bulges, and the degree of bulge protrusion. Step S402: Based on the set sorting criteria, determine the sorting criteria that the first lithium battery meets, and sort the first lithium battery for secondary use; repeatedly obtain the battery surface bulge information of the first type of battery, and perform the corresponding secondary use sorting. In the specific implementation process, by establishing sorting standards and sorting based on the bulging information on the battery surface, objective, uniform and batch tiered sorting of square lithium batteries can be achieved. This can reduce subjective differences and missed detection problems caused by manual visual inspection, reduce the poor adaptability of fixed threshold methods to slight deformation and local differences, improve the accuracy of bulging detection, and thus improve the consistency and reliability of tiered utilization sorting of retired lithium batteries.

[0030] Example 2, please refer to Figure 4 As shown, Figure 4 A schematic diagram of an electronic device is provided, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call these instructions. When the processor executes a computer-readable instruction, it performs steps such as those in a smart sorting method for the secondary utilization of lithium batteries using fusion vision inspection, to achieve the following functions: acquiring basic appearance parameters of square lithium batteries and collecting three-dimensional point clouds of the battery surface to obtain battery surface point cloud data; denoising the battery surface point cloud data, filtering and removing outliers to obtain standard surface point cloud data; obtaining actual thickness information based on the standard surface point cloud data and setting an adaptive threshold for surface bulging analysis to obtain battery surface bulging information; and sorting the square lithium batteries for secondary utilization based on the battery surface bulging information.

[0031] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0032] Example 3: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of a smart sorting method for the secondary utilization of lithium batteries that integrates visual inspection, to achieve the following functions: acquiring basic appearance parameters of square lithium batteries and collecting three-dimensional point clouds of the surface of the square lithium batteries to obtain battery surface point cloud data; performing noise reduction processing on the battery surface point cloud data, filtering and removing abnormal points in the point cloud to obtain standard surface point cloud data; acquiring actual thickness information based on the standard surface point cloud data, and setting an adaptive threshold for surface bulging analysis to obtain battery surface bulging information; and sorting the square lithium batteries for secondary utilization based on the battery surface bulging information.

[0033] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0034] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0035] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart sorting method for the cascade utilization of lithium batteries integrating visual inspection, characterized in that, Includes the following steps: The basic appearance parameters of the square lithium battery are obtained, and the three-dimensional point cloud of the surface of the square lithium battery is collected to obtain the point cloud data of the battery surface. The battery surface point cloud data is denoised, and outliers are filtered and removed to obtain standard surface point cloud data. The actual thickness information is obtained based on standard surface point cloud data, and surface bulging analysis is performed by setting an adaptive threshold to obtain battery surface bulging information. Based on the information about bulging on the surface of the square lithium batteries, the square lithium batteries are sorted for secondary use.

2. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection as described in claim 1, characterized in that, Obtaining the basic appearance parameters of the square lithium battery and acquiring the 3D point cloud of the battery surface to obtain the battery surface point cloud data includes the following sub-steps: Let any type of square lithium battery be referred to as the first type of battery; let any two parallel surfaces of the first type of battery be referred to as the corresponding surface group; let the corresponding surface groups to be tested be obtained and referred to as corresponding surface group 1 to corresponding surface group n in sequence; where n is the number of corresponding surface groups to be tested. For the corresponding surface group 1, the vertical straight-line distance between the two surfaces in the corresponding surface group 1 is recorded as the surface distance of the corresponding surface group 1, and marked as surface distance 1; the range in which the surface distance 1 of the qualified first type battery should be located is obtained, and recorded as the corresponding qualified range [A0, B0].

3. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection as described in claim 2, characterized in that, Obtaining the basic appearance parameters of the square lithium battery and acquiring the three-dimensional point cloud of the surface of the square lithium battery to obtain the battery surface point cloud data also includes the following sub-steps: Repeatedly obtain the qualified intervals corresponding to the surface distances of all corresponding surface groups, and record them as the basic appearance parameters of the first type of battery; The two surfaces in the corresponding surface group 1 are respectively designated as surface 11 and surface 12; any one of the first type batteries to be sorted in the tier is designated as the first lithium battery; Using a laser scanning device, the three-dimensional point cloud of the surface 11 and surface 12 of the first lithium battery is acquired simultaneously to obtain the point cloud data of the corresponding surface group 1 of the first lithium battery. Repeatedly collect point cloud data of all corresponding surface groups of the first lithium battery to obtain the battery surface point cloud data of the first lithium battery.

4. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 3, characterized in that, The process of denoising the battery surface point cloud data, filtering and removing outliers to obtain standard surface point cloud data, includes the following sub-steps: Establish a spatial rectangular coordinate system, and label the horizontal axis, vertical axis and vertical axis of the spatial rectangular coordinate system as the x-axis, y-axis and z-axis respectively; place the point clouds of surface 11 and surface 12 in the spatial rectangular coordinate system according to their relative relationship, and place them vertically and horizontally parallel to the xy plane of the spatial rectangular coordinate system, and label them as the upper point cloud and the lower point cloud respectively according to their vertical relationship; Obtain the z-axis coordinates of all points in the point cloud above, and calculate the median MA and median absolute deviation MD of all z-axis coordinates; denote any point in the point cloud above as the first point, and label the z-axis coordinates of the first point as AZ; If AZ is not located in [MA-3*MD, MA+3*MD], then mark the first point as a location error point; repeat the filtering of all location error points in the point cloud above and remove them to obtain the corresponding standard point cloud 1; Then, statistical filtering is performed on the standard point cloud 1 above to filter out the corresponding outliers and remove them, resulting in the corresponding standard point cloud 2; this is called the initial point cloud above. The same process is repeated on the point cloud below to obtain the corresponding standard point cloud 2, which is called the initial point cloud below.

5. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 4, characterized in that, Denoising the battery surface point cloud data, filtering and removing outliers to obtain standard surface point cloud data, also includes the following sub-steps: Obtain the projections of the upper and lower preliminary screening point clouds onto the xy plane, and divide the region where the projections are located on the xy plane into multiple grids of size a1*a1; the grid that simultaneously contains the projections of the upper and lower preliminary screening point clouds is denoted as a non-empty grid; where a1 is the set grid side length; Denote any non-empty grid as the first grid; considering only the x-axis and y-axis coordinates; obtain the upper and lower preliminary point clouds located in the first grid respectively, and denote them as the upper point set and lower point set of the first grid; Calculate the median z-axis coordinates of the upper and lower point sets respectively, denoted as MZD and MZU. Calculate MZD-MZU, denoted as the reference thickness TB of the first grid. Repeat the collection of reference thicknesses for all non-empty grids.

6. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 5, characterized in that, Denoising the battery surface point cloud data, filtering and removing outliers to obtain standard surface point cloud data, also includes the following sub-steps: Calculate the median TA and median absolute deviation TD for all reference thicknesses, and mark non-empty grids whose reference thicknesses are not located in [TA-3*TD, TA+3*TD] as anomalous grids; If the first grid is an abnormal grid, then based on the z-axis coordinate, the interquartile range method is used to filter and remove abnormal points in the upper and lower point sets respectively; repeat the process for all abnormal grids, filter and remove all abnormal points, and after completion, the upper standard point cloud and the lower standard point cloud are obtained, which are recorded as the standard point cloud data of the corresponding surface group 1. Repeatedly acquire standard point cloud data for all corresponding surface groups to obtain standard surface point cloud data.

7. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection as described in claim 6, characterized in that, The actual thickness information is obtained based on standard surface point cloud data, and surface bulging analysis is performed by setting an adaptive threshold. The process to obtain battery surface bulging information includes the following sub-steps: Ignore the z-axis coordinate, obtain the position of each point in the standard point cloud above, and record it as the thickness point. Record any thickness point as the second point. The distribution is based on the lower standard point cloud and the upper standard point cloud to perform surface fitting, and the lower fitting surface and the upper fitting surface are obtained respectively; the vertical distance between the second point in the lower fitting surface and the upper fitting surface is obtained, and denoted as the actual thickness AF of the second point; Obtain all thickness points in the fitted surface and record them as upper thickness points; if the actual thickness AF of the second point is greater than 1.2*B0 or ​​less than A0, mark the upper thickness point at the corresponding position as an abnormal deformation point. Repeatedly filter all abnormal deformation points in the upper thickness points and remove them. Then, use the remaining upper thickness points to perform surface fitting to obtain the trend surface corresponding to the upper fitted surface, which is denoted as the upper trend surface. Based on the lower fitted surface, repeatedly obtain the trend surface corresponding to the lower fitted surface, which is denoted as the lower trend surface.

8. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 7, characterized in that, The actual thickness information is obtained based on standard surface point cloud data, and surface bulging analysis is performed by setting an adaptive threshold. The process to obtain battery surface bulging information includes the following sub-steps: Obtain the corresponding point of the second point on the upper trend surface, and take the corresponding point as the endpoint to obtain the vertical displacement from the corresponding point to the upper fitting surface, which is denoted as the upper residual DC of the second point. Obtain the corresponding point of the second point on the lower trend surface, and take the corresponding point as the endpoint to obtain the vertical displacement from the corresponding point to the lower fitting surface, which is denoted as the lower residual UC of the second point. Repeatedly obtain the upper and lower residuals of all thickness points to obtain the upper residual set and the lower residual set respectively; Set three circular windows with radii e1, e2, and e3, respectively, and label them small window, medium window, and large window. Here, e1, e2, and e3 represent the set radii, e1... <e2<e3。 9. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 8, characterized in that, The actual thickness information is obtained based on standard surface point cloud data, and surface bulging analysis is performed by setting an adaptive threshold. The process to obtain battery surface bulging information includes the following sub-steps: Based on the upper residual set, with the second point as the center of the small window, calculate the median RA and median absolute deviation RD of the upper residuals of all thickness points within the small window; if DC is greater than RA+k1*RD, then mark the second point as a suspected point, where k1 is a set coefficient. Repeat the judgment with the second point as the center of the middle window and the large window. If the second point is marked as a suspected point under any circular window, then mark the second point as the upper bulge point. Based on the lower residual set, the judgment is repeatedly performed with the second point as the center of each circular window. If the second point is marked as a suspected point under any circular window, then the second point is marked as the lower bulge point. Repeatedly obtain all the upper and lower drum point locations to get the upper drum set and the lower drum set respectively; Connectivity analysis is performed based on the upper bulge set. Adjacent and continuous upper bulge points are merged to obtain upper bulge regions. The area of ​​each upper bulge region is calculated and denoted as the bulge area. The average value of the upper residual of each upper bulge point in the upper bulge region is denoted as the bulge height. Based on the repeated processing of the lower bulge set, the lower bulge region, as well as the corresponding bulge area and bulge height, are obtained; Record all upper and lower bulge areas, along with their corresponding bulge areas and heights, as the bulge detection information for the corresponding surface group 1; repeat this process for all corresponding surface groups to obtain the battery surface bulge information.

10. The intelligent sorting method for the cascade utilization of lithium batteries based on visual inspection according to claim 9, characterized in that, Based on the information about surface bulges in the square lithium batteries, the sorting of square lithium batteries for secondary use includes the following sub-steps: Set sorting criteria corresponding to the number of bulges, bulge area, and bulge height; obtain the number of bulges and the corresponding total bulge area based on the bulge information on the surface of the first lithium battery; Based on the set sorting criteria, the sorting criteria met by the first lithium battery are judged, and the first lithium battery is sorted for secondary use; the battery surface bulge information of the first type of battery is repeatedly obtained, and the corresponding secondary use sorting is performed.

Images