Visual detection method and system for mobile phone battery by intelligent edge device

Abstract

The invention discloses a visual detection method and system for a mobile phone battery by an intelligent edge device, and relates to the technical field of visual detection, and the method comprises the steps: determining a corresponding surface defect twin model based on the feature positions of a plurality of surface defect features, the corresponding feature forms and the overall form of the mobile phone battery; visual detection of the mobile phone is realized, and the accuracy of the surface defect twin model is improved. The multiple surface defect events of the mobile phone battery are determined according to detection of the surface defect twinborn model, the multi-dimensional defect system is determined according to the multiple surface defect events, the corresponding surface defect areas and the corresponding temperature data, and the accuracy of the multi-dimensional defect system is improved. Determining a defect maintenance route of the mobile phone battery according to the defect distribution diagram of each surface defect area, the corresponding area priority and the previous maintenance event of the mobile phone battery; and a plurality of defect maintenance nodes are determined based on the identification of the defect maintenance route, so that the accuracy of the dynamic maintenance event is further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual detection, and in particular to a visual detection method and system for a mobile phone battery by using an intelligent edge device. BACKGROUND

[0002] With the development of science and technology, the intelligent edge device is configured with multiple cameras as a smart device, and based on the multiple cameras, the mobile phone battery is photographed, and multiple images of the mobile phone battery are collected, and a three-dimensional model of the mobile phone battery is constructed according to the multiple images. In the prior art, the corresponding defect features are determined based on the recognition of the multiple images, and the defect content of the mobile phone is determined along a single defect feature, and the feature morphology of the surface defect feature and the overall morphology of the mobile phone battery are ignored, which affects the accuracy of the surface defect twin model, and leads to low accuracy of the dynamic maintenance event. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art, and provides a visual detection method and system for a mobile phone battery by using an intelligent edge device.

[0004] The present application provides a visual detection method for a mobile phone battery by using an intelligent edge device, which comprises the following steps: When the mobile phone battery enters the shooting space of the intelligent edge device, multiple cameras of the intelligent edge device perform visual detection on the mobile phone battery, and multiple surface images of the mobile phone battery are determined in the visual detection; Based on the image recognition of each surface image, multiple surface defect features are determined, and based on the feature position, corresponding feature morphology and overall morphology of the mobile phone battery, a corresponding surface defect twin model is determined. The surface defect twin model not only contains the geometric information of the mobile phone battery, but also contains a group of discrete digital twins with space-time attributes and physical characteristics; According to the detection of the surface defect twin model, multiple surface defect events of the mobile phone battery are determined, and according to the multiple surface defect events, corresponding surface defect regions and corresponding temperature data, a multi-dimensional defect system is determined; According to the multi-dimensional defect system, the use history of the mobile phone battery and the corresponding damaged behavior, a defect distribution map of each surface defect region is determined, and according to the defect distribution map of each surface defect region, corresponding regional priority and past maintenance events of the mobile phone battery, a defect maintenance route of the mobile phone battery is determined. The defect maintenance route contains a time axis, an operation sequence, a tool switching instruction and an emergency plan; Based on the recognition of the defect maintenance route, multiple defect maintenance nodes are determined, according to the node position, corresponding node priority and service life of each defect maintenance node, a dynamic maintenance event is determined, and a corresponding dynamic maintenance table is output.

[0005] The embodiment of the present application provides a visual detection system for a mobile phone battery of a smart edge device, which is applied to the visual detection method for the mobile phone battery of the smart edge device.

[0006] Compared with the prior art, the present application has the following beneficial effects: (1) When the mobile phone battery enters the shooting space of the smart edge device, the multiple cameras of the smart edge device perform visual detection on the mobile phone battery, and multiple surface images of the mobile phone battery are determined in the visual detection; multiple surface defect features are determined based on image recognition of each surface image, and corresponding surface defect twin models are determined based on feature positions, corresponding feature morphologies and overall morphologies of the mobile phone battery, multiple surface images are introduced, visual detection of the mobile phone is realized, and the accuracy of the surface defect twin model is improved.

[0007] (2) Multiple surface defect events of the mobile phone battery are determined according to the detection of the surface defect twin model, and a multi-dimensional defect system is determined according to the multiple surface defect events, corresponding surface defect regions and corresponding temperature data, the surface defect events are further controlled, the overall consideration of the multiple surface defect events, the corresponding surface defect regions and the corresponding temperature data is realized, and the accuracy of the multi-dimensional defect system is improved.

[0008] (3) The defect distribution diagram of each surface defect region is determined according to the multi-dimensional defect system, the use history of the mobile phone battery and the corresponding damaged behavior, the defect maintenance route of the mobile phone battery is determined according to the defect distribution diagram of each surface defect region, the corresponding region priority and the past maintenance events of the mobile phone battery; multiple defect maintenance nodes are determined based on the recognition of the defect maintenance route, dynamic maintenance events are determined according to the node positions of each defect maintenance node, the corresponding node priority and the service life of the mobile phone battery, the defect maintenance route of the mobile phone battery is further controlled, each defect maintenance node is strictly controlled, the accuracy of the dynamic maintenance event is improved, and the corresponding dynamic maintenance table is output. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a flowchart of the visual detection method for a mobile phone battery of a smart edge device in the embodiment of the present application; Figure 2 is a flowchart of step S11 in the visual detection method for a mobile phone battery of a smart edge device in the embodiment of the present application; Figure 3 is a flowchart of step S12 in the visual detection method for a mobile phone battery of a smart edge device in the embodiment of the present application; Figure 4is a flowchart of step S13 in the visual detection method for mobile phone batteries by the intelligent edge device in the embodiment of the application; Figure 5 is a flowchart of step S14 in the visual detection method for mobile phone batteries by the intelligent edge device in the embodiment of the application; Figure 6 is a flowchart of step S15 in the visual detection method for mobile phone batteries by the intelligent edge device in the embodiment of the application; Figure 7 is a structural composition diagram of the visual detection system for mobile phone batteries by the intelligent edge device in the embodiment of the application. DETAILED DESCRIPTION

[0010] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application.

[0011] Please refer to Figures 1 to 7 A visual detection method for mobile phone batteries by an intelligent edge device, applied to a visual detection scene; the visual detection method for mobile phone batteries by the intelligent edge device comprises the following steps. Step S11: When a mobile phone battery enters a shooting space of the intelligent edge device, multiple cameras of the intelligent edge device perform visual detection on the mobile phone battery, and multiple surface images of the mobile phone battery are determined in the visual detection; Step S12: Multiple surface defect features are determined based on image recognition of each surface image, and a corresponding surface defect twin model is determined based on feature positions, corresponding feature morphologies of the multiple surface defect features, and an overall morphology of the mobile phone battery; Step S13: Multiple surface defect events of the mobile phone battery are determined according to detection of the surface defect twin model, and a multi-dimensional defect system is determined according to the multiple surface defect events, corresponding surface defect regions, and corresponding temperature data; Step S14: A defect distribution map of each surface defect region is determined according to the multi-dimensional defect system, a use history of the mobile phone battery, and corresponding damaged behaviors, and a defect maintenance route of the mobile phone battery is determined according to the defect distribution map of each surface defect region, corresponding region priorities, and past maintenance events of the mobile phone battery; Step S15: Multiple defect maintenance nodes are determined based on recognition of the defect maintenance route, a dynamic maintenance event is determined according to node positions, corresponding node priorities of each defect maintenance node, and a service life of the mobile phone battery, and a corresponding dynamic maintenance table is output.

[0012] Reference Figure 2 In step S11, the specific steps are as follows: S111: detecting the intelligent edge device, marking the shooting space of the intelligent edge device during the detection process, determining the corresponding shooting range based on the shooting space of the intelligent edge device, and outputting the corresponding shooting signal by the intelligent edge device when the mobile phone battery enters the intelligent edge device, determining the visual detection system according to the shooting signal, the spatial position of the multiple cameras and the mobile phone battery relative to the shooting range, and realizing visual detection of the mobile phone battery; S112: in the visual detection system, determining multiple visual detection dimensions based on the identification of the visual detection system, and determining multiple surface images of the mobile phone battery according to each visual detection dimension, the overall shape of the mobile phone battery and each surface of the mobile phone battery.

[0013] In the embodiment of the application, the intelligent edge device is detected, and the shooting space of the intelligent edge device is marked during the detection process. The corresponding shooting range is determined based on the shooting space of the intelligent edge device. At the same time, when the mobile phone battery enters the intelligent edge device, the intelligent edge device outputs the corresponding shooting signal. The visual detection system is determined according to the shooting signal, the spatial position of the multiple cameras and the mobile phone battery relative to the shooting range, and the visual detection of the mobile phone battery is realized. The overall consideration of the shooting signal, the spatial position of the multiple cameras and the mobile phone battery relative to the shooting range is compatible, and the accuracy of the visual detection system is ensured.

[0014] At this time, the intelligent edge device performs a bottom-layer hardware self-checking protocol to confirm that the camera, the I / O control unit and the image processing unit are in a ready state. Then, the intrinsic matrix and the extrinsic matrix of the camera are solved by a calibration algorithm to establish an imaging model. The system defines the three-dimensional boundary of the effective field of view of the camera in the physical world according to these parameters, maps it as a spatial mask in the device coordinate system, thereby clarifying the mapping relationship between the image sensor pixels and the physical space coordinates, and establishing the final “shooting range”.

[0015] When the target object enters the detection area, the edge device senses its arrival through a photoelectric sensor or a laser ranging sensor, triggers an interrupt request by level jump, and calculates the exposure delay by combining the real-time speed vector of the production line feedback by the encoder; the system sends a hard trigger signal containing accurate timing synchronization information to the light source controller and the camera when the target object moves to the preset optimal imaging plane, ensuring that the strobe of the light source and the opening of the camera shutter are synchronized at the microsecond level.

[0016] The original image data is acquired by using a multi-camera system, and the six-degree-of-freedom pose of the target object in the device coordinate system is solved by a stereo vision matching algorithm or a multi-view geometry constraint; at this time, the system combines the previously determined shooting range to construct a visual detection system that defines the topological relationship of each camera and the detection coverage, and if the target space position exceeds the preset tolerance, the ROI of the image processing is dynamically adjusted or an abnormal pose marker is output; the multi-camera of the edge device acquires images according to the pre-set light source strategy in parallel or time-sharing, controls the lighting module, and transmits the data to the memory buffer through a high-bandwidth bus, and completes the conversion from the physical space to the digital image space through preprocessing such as denoising and white balance correction, to prepare standardized data flow for subsequent steps.

[0017] Specifically, the intelligent edge device initializes the multi-angle shooting of the calibration board, calculates the intrinsic parameters of cameras A, B and C and the extrinsic parameters in the intelligent edge device base coordinate system, and strictly sets the camera A to cover the area of 100mm×60mm on the top surface of the mobile phone battery, and the camera C to cover the 20mm×20mm tab area on the right side, and marks these areas in the three-dimensional space coordinates as the shooting range.

[0018] When the mobile phone battery moves to the entrance of the intelligent edge device and blocks the light beam of the photoelectric sensor, the sensor generates a TTL level jump, and the edge computing board card of the intelligent edge device reads the encoder pulse of the conveyor belt, calculates that the battery moves at a speed of 0.5m / s, and after a preset 200ms delay, the system sends a shooting signal to all cameras and light source controllers.

[0019] The three cameras of the intelligent edge device respond to the shooting signal at the same time, and after the original image is collected, the edge device uses the projection coordinates of the feature points in the multi-camera view to solve the current pose of the mobile phone battery by PnP algorithm; for the case that the solved mobile phone battery rotates clockwise by 2°, the visual detection system of the intelligent edge device dynamically corrects the logical ROI of each camera according to the spatial position information, instructs camera A to translate the region of interest to the left and rotate by a corresponding angle, and ensures that the tab area after rotation still falls completely in the center of the field of view of camera C.

[0020] According to the adjusted visual detection system, camera A collects the image of the insulating film on the top surface of the mobile phone battery under the dome light source to detect scratches, and camera C collects the high-contrast image of the tab under the blue coaxial light illumination, and all image data is transmitted to the video memory of the intelligent edge device in real time, completing a high-precision, dynamic adaptive visual detection data acquisition for the mobile phone battery.

[0021] Further, in the visual detection system, a plurality of visual detection dimensions are determined based on the identification of the visual detection system, and a plurality of surface images of the mobile phone battery are determined according to each visual detection dimension, the overall shape of the mobile phone battery, and each surface of the mobile phone battery, which comprehensively considers each visual detection dimension, the overall shape of the mobile phone battery, and each surface of the mobile phone battery, and guarantees the accuracy of the plurality of surface images of the mobile phone battery.

[0022] At this time, the edge device reads the visual detection system parameters constructed in the S111 stage, including camera calibration data, target object pose, and topological relationship, and decomposes the overall visual detection task into a plurality of orthogonal visual detection dimensions through a task-driven algorithm according to the imaging characteristics, lighting conditions, and resolution of different cameras; these dimensions correspond to specific physical property observation directions or defect type detection modes, such as macro geometric shape dimension, micro surface texture dimension, or specific component assembly dimension, thereby defining the basis of the feature space and determining the granularity and direction of subsequent data extraction.

[0023] The system uses the CAD model parameters of the target object or the prior overall shape constraint to perform projection transformation combined with the real-time six-degree-of-freedom pose; the edge device generates a dynamic space mask in the image plane to digitally segment the background environment and the effective area of the target object, and further subdivides the effective image area into different surface sub-areas according to the geometric topological structure, such as the top main area, the side edge area, and the tab welding area, and each sub-area is assigned a corresponding attribute label for matching with the detection dimension.

[0024] The system extracts corresponding image blocks from the original high-resolution acquisition image based on the mapping relationship between the dimensions and the surface areas; the system performs preprocessing and standardization operations on these image blocks, including radiation correction, geometric correction, and resampling, and finally integrates the image data under different dimensions into standardized data packets, and outputs a set of surface image collections containing position, size, shape information, and pixel data matrix, which completely and non-overlappingly covers all features to be detected of the target object.

[0025] Specifically, the intelligent edge device knows that the mobile phone battery has a slight counterclockwise rotation on the conveyor belt; in the detection dimension decomposition stage, the algorithm module of the intelligent edge device analyzes the current visual detection system and defines two core visual detection dimensions: the "global topography dimension" based on large field of view and low magnification imaging, which is used to focus on the overall contour integrity, size consistency, and adhesion of the insulating film edge of the mobile phone battery; and the "local defect dimension" based on high magnification and narrow field of view imaging, which is used to focus on whether there are welding slag, virtual welding, and fine scratches on the tab area at the bottom of the mobile phone battery.

[0026] The intelligent edge device reads the standard size of the mobile phone battery, generates a parallelogram mask corresponding to the angle in the original large field of view image based on the rotation angle calculated in S111, and accurately removes the black background of the conveyor belt; at the same time, the system delimits the boundary in the mask according to the topological structure of the battery, and marks the area occupying 90% of the mask area as surface area A (top surface) for "global topography dimension", and marks the rectangular area of a certain size at the bottom of the mask as surface area B (tab area) for "local defect dimension".

[0027] For surface area A, the intelligent edge device extracts the corresponding pixel matrix, applies a perspective transformation algorithm to eliminate geometric distortion caused by slight warping of the battery or camera angle, generates a corrected "standard front view of the top surface of the mobile phone battery", and marks it as belonging to "global topography dimension"; for surface area B, the system extracts the tab area from the original image of the dedicated close-up camera, selects the layer with the clearest image gradient, generates a high-resolution "microscopic image of the tab of the mobile phone battery", and marks it as belonging to "local defect dimension"; the intelligent edge device outputs these two surface images with clear dimensional and regional attributes, ensuring that the feature analysis in the subsequent S12 step can be based on accurate geometric benchmarks.

[0028] Reference Figure 3 In step S12, the specific steps are as follows: S121: In each surface image, a plurality of sub-surface regions are determined based on the detection of the surface image, and corresponding surface defect features are determined according to the recognition of each sub-surface region, so as to collect a plurality of surface defect features; S122: The feature position of each surface defect feature is determined based on the recognition of the surface defect feature, and the feature morphology of the surface defect feature is marked, and a first repositioning model is determined according to the feature position of the plurality of surface defect features and the overall morphology of the mobile phone battery. S123: A first re-morphology model is determined according to the feature morphology of the plurality of surface defect features and the overall morphology of the mobile phone battery, and a corresponding surface defect twin model is determined based on the multi-dimensional synthesis of the first repositioning model and the first re-morphology model.

[0029] In the embodiments of the present application, in each surface image, a plurality of sub-surface regions are determined based on the detection of the surface image, and corresponding surface defect features are determined according to the recognition of each sub-surface region, so as to collect a plurality of surface defect features, which is compatible with the overall consideration of the recognition of each sub-surface region, and ensures the accuracy of the corresponding surface defect features.

[0030] At this time, the edge device performs preprocessing such as denoising, contrast enhancement, and edge sharpening on the input surface image, and then divides the image at the pixel level using an image segmentation algorithm; the system uses a semantic segmentation network based on deep learning to classify the pixels in areas with complex texture backgrounds, and uses traditional edge detection and threshold segmentation algorithms for areas with obvious geometric features, and finally divides the image into several non-overlapping or specifically nested connected domains, which are defined as sub-surface regions.

[0031] The system extracts a high-dimensional feature vector for each segmented sub-surface region, covering geometric features, texture features, color features, and depth features; the system uses a pre-trained classifier to analyze the feature vector, and determines whether the region meets the process specification based on the decision boundary. If the feature vector deviates from the normal sample cluster or falls into the defect sample cluster, it is identified and labeled as a specific type of surface defect feature.

[0032] For sub-surface regions identified as defects, the system calculates key parameters such as the centroid coordinates, the circumscribed rectangle, the defect intensity, and the direction angle; the edge device converts unstructured pixel data into structured feature descriptors, forming standardized defect records containing unique IDs, defect category labels, location information, and quantitative morphological parameters, and aggregates all identified structured records in the current image, outputting a set containing multiple surface defect features.

[0033] Specifically, the intelligent edge device is analyzing a silver metal wire-drawing texture image on the top surface of a mobile phone battery, which contains an obvious black foreign particle and a slight indentation; the intelligent edge device applies anisotropic diffusion filtering to preserve metal edges and remove noise, and then uses a bottom-hat transform algorithm based on morphological reconstruction to enhance dark areas in the image; through adaptive threshold segmentation, the system marks two independent connected domains in the image: RegionA, located in the upper right corner of the battery, with an extremely irregular shape and extremely low pixel grayscale, and RegionB, located in the center of the battery, with a long strip shape and a blurred edge.

[0034] For RegionA, the system extracts its small area but high edge gradient, and a length-to-width ratio close to 1, and determines that it does not meet the metal wire-drawing texture feature, and identifies it as a "foreign matter" category of surface defect feature; for RegionB, the system calculates its local binary pattern and finds that the texture uniformity has changed and has directionality, and determines it to be a "indentation" category of surface defect feature, while the remaining large-area wire-drawing texture area that meets the preset template is identified as a normal sub-surface region.

[0035] The intelligent edge device quantifies the "foreign matter" feature, calculates its centroid coordinates as (0, 0), and the size of the circumscribed rectangle is 2 mm x 2 mm, and the defect intensity level is set to "high"; the "indentation" feature is quantified, the centroid coordinates are calculated as (0, 0), the length is 10 mm, the width is 1 mm, the direction angle is 45°, and the defect intensity level is set to "medium"; the intelligent edge device outputs a surface defect feature set containing the two records, and accurately describes the physical properties of the defects existing on the top surface of the mobile phone battery in the form of structured data.

[0036] Further, based on the recognition of each surface defect feature, the feature position of the surface defect feature is determined, and the feature morphology of the surface defect feature is marked, and a first repositioning model is determined according to the feature positions of the plurality of surface defect features and the overall morphology of the mobile phone battery, which is compatible with the overall consideration of the feature positions of the plurality of surface defect features and the overall morphology of the mobile phone battery, and ensures the accuracy of the first repositioning model.

[0037] At this time, the edge device obtains the image pixel coordinates of the surface defect feature, calls the camera calibration parameters and the hand-eye calibration parameters in the visual detection system, and performs coordinate transformation in combination with the six-degree-of-freedom pose of the target object calculated in real time in the S111 stage; for monocular vision, the system converts the pixel coordinates into three-dimensional coordinates in the camera coordinate system by using the inverse transformation of the perspective projection matrix, and then converts to the world coordinate system of the battery through rigid body transformation; for binocular or structured light vision, the spatial three-dimensional coordinates of the feature points are directly obtained by using the triangulation principle or depth map mapping, so as to determine the accurate physical feature position of the defect on the surface of the battery entity.

[0038] The system performs geometric morphology analysis on the pixel region of the surface defect feature, extracts Fourier descriptors, invariant moments and topological properties based on the defect edge contour; the system matches these descriptors in the pre-defined morphology semantic library to qualitatively mark the defects, for example, marking high aspect ratio regions as linear scratches, and marking circular-like regions as point-like pits, while calculating the direction vector and size attribute of the defect, combining the quantitative geometric parameters with the qualitative label to form a complete feature morphology label.

[0039] The system reads the overall morphology model of the mobile phone battery, and maps the three-dimensional spatial coordinate point set of all the defect features calculated to the three-dimensional grid vertices of the overall morphology model; the system calculates the relative position relationship of each defect feature point and the overall structure of the battery, including the distance from the edge, the topological surface to which it belongs, and the relative spatial distance between the feature points, and finally generates a data structure containing the accurate position information of all the defect features in the overall coordinate system of the battery, forming a topological mapping diagram of the defects in the geometric space of the battery.

[0040] Specifically, it is known that the mobile phone battery is a cuboid structure with dimensions of 150 mm x 70 mm x 5 mm. In the image, a top surface scratch and a side surface R corner pit are detected. The smart edge device obtains the image pixel coordinates of the center of the top surface scratch, combines the camera calibration parameters and the current yaw angle of the mobile phone battery, and uses inverse perspective transformation to calculate the coordinates of the point in the physical coordinate system of the mobile phone battery as (x = 40 mm, y = 20 mm, z = 5 mm), determining the feature position of the scratch. For the side surface R corner pit, since it is located on a curved surface and may have perspective distortion, the system calculates the depth using the disparity value of the binocular camera and solves its three-dimensional coordinates as (x = 148 mm, y = 10 mm, z = 2.5 mm), accurately positioning it on the R corner transition surface of the right edge of the battery.

[0041] For the top surface feature, the system calculates its minimum circumscribed rectangle, finds that the long axis is 15 mm and the short axis is 0.2 mm, with an aspect ratio of 75:1, and marks it as "linear scratch" according to the morphological semantic library, and calculates its principal axis direction vector as (1, 0, 0); For the side surface feature, the system calculates its area and circumscribed circle diameter, finds that the shape is close to a circle and the diameter is about 0.8 mm, and marks it as "point pit", and records its morphology as "non-penetrating depression".

[0042] The smart edge device loads the overall morphology grid model of the mobile phone battery, maps the scratch coordinates to the "top surface 01" sub-area of the grid model, and maps the pit coordinates to the "right side surface R corner" sub-area; The system further analyzes the spatial relationship to determine that the scratch is 10 mm away from the top edge of the battery and the pit is located at the end of the length direction of the battery; The smart edge device generates a first position model, in which the three-dimensional grid of the mobile phone battery is assigned attributes: the top surface specific grid node is marked as "scratch attachment area", and the side surface R corner specific node is marked as "pit attachment area", clearly showing the absolute and relative position distribution of the defects in the battery physical space.

[0043] Therefore, the first position model is determined according to the feature morphology of the plurality of surface defect features and the overall morphology of the mobile phone battery, the corresponding surface defect twin model is determined based on the multi-dimensional synthesis of the first position model and the first morphology model, which is compatible with the overall consideration of the multi-dimensional synthesis of the first position model and the first morphology model, ensuring the accuracy of the corresponding surface defect twin model, and at the same time, multiple surface images are introduced to realize visual detection of the mobile phone, improving the accuracy of the surface defect twin model.

[0044] At this time, the edge device mathematically models the geometric properties of each surface defect feature collected, and uses corresponding geometric primitive fitting algorithms for defects of different morphologies; for linear defects, the system uses least squares method to fit a straight line or a spline curve in three-dimensional space; for area-type defects, polynomial surface fitting or sphere / ellipsoid fitting is used; for texture-type defects, a texture mapping matrix is used for description; the system performs Boolean operations on these fitted local defect geometries and the overall morphology of the mobile phone battery, performs Boolean difference set operation for concave defects, and performs Boolean union set operation for convex defects, thereby generating a first-morphology model containing defect geometric shape, size and depth information.

[0045] The system performs multi-dimensional data fusion on the first-morphology model containing defect geometric structure and the first-position model containing defect three-dimensional coordinates and topological index; the system accelerates the search process through a spatial index structure, instantiates each geometric defect body in the morphology model and transforms it to the precise physical pose defined in the position model, while establishing a bidirectional mapping linked list to bind the geometric properties of the defect with its position properties, ensuring that the "shape" and "position" of the defect are strictly consistent in logic and mathematics.

[0046] The edge device holographically packages the model and calculates relevant physical property mappings based on defect features, such as stress concentration coefficient based on defect depth or risk probability based on defect area; the final surface defect twin model is a highly accurate digital representation, which not only contains the geometric information of the mobile phone battery, but also contains a set of discrete digital twin bodies with space-time attributes and physical characteristics, maintaining micron-level consistency with the physical entity in geometry and marking the specific areas and extent of health state damage in semantics.

[0047] Specifically, it is known that the overall morphology of the mobile phone battery is a rectangular aluminum shell with dimensions 150mm x 70mm x 5mm, and there is a 15mm long scratch on the top surface and a 1mm diameter pit at the R corner of the side surface; in the first-morphology model construction stage, the intelligent edge device fits a three-dimensional straight line segment with a length of 15mm and a depth of 0.05mm for the "scratch" feature through pixel coordinate sequence fitting, and fits a half-sphere geometry with a diameter of 1mm and a depth of 0.3mm for the "pit" feature.

[0048] The system loads the standard CAD mesh model of the mobile phone battery, performs mesh carving operations at the top surface coordinates to simulate the groove morphology of the scratch, and modifies the normal vectors of the mesh vertices at the R corner coordinates of the side surface to generate the spherical cap shape of the pit, at which time the model of the mobile phone battery becomes the first-morphology model containing the above physical deformation.

[0049] The intelligent edge device binds the node data in the first heavy position model with the geometric body in the first heavy shape model, verifies whether the direction vector of the scratch is perpendicular to the normal vector in the position model to verify the authenticity of the defect adhering to the surface, and associates the "R corner-node ID_5505" with the pit hemisphere, so that the defect becomes a "groove located at a specific coordinate on the top surface of the mobile phone battery and having a specific direction" and a "pit located at a specific coordinate of the R corner of the side surface of the mobile phone battery".

[0050] The intelligent edge device renders and encapsulates the fused model to generate a final surface defect twin model; a complete "health status layer" is included in the digital file of the model, and specific mesh areas on the top surface and the side surface are highlighted as red defect areas, with metadata attached: {defect type: scratch, risk level: medium, geometric size: 15mmx0.05mm}; at this time, the mobile phone battery in the memory of the intelligent edge device becomes a digital twin that is completely one-to-one corresponding to the physical entity on the conveyor belt in terms of shape and defect distribution, for direct reading and decision-making in subsequent steps.

[0051] Reference Figure 4 In step S13, the specific steps are as follows: S131: In the surface defect twin model, the multiple surface defect contents of the mobile phone battery on each surface are determined according to the identification of the surface defect twin model, and the surface defect features corresponding to each surface defect content are marked, and the multiple surface defect events of the mobile phone battery are determined according to each surface defect content, the corresponding surface defect feature, and the surface use event of the mobile phone battery; S132: Based on the identification of each surface defect event, the corresponding surface defect area is determined, and at the same time, temperature detection is performed along the surface defect area, and the corresponding temperature data is output, the first-dimensional defect content is determined according to the multiple surface defect events and the area shape of the corresponding surface defect area, the second-dimensional defect content is determined according to the multiple surface defect events and the corresponding temperature data, and the multi-dimensional defect system is determined based on the first-dimensional defect content and the second-dimensional defect content.

[0052] In the embodiments of the present application, in the surface defect twin model, the multiple surface defect contents of the mobile phone battery on each surface are determined according to the identification of the surface defect twin model, and the surface defect features corresponding to each surface defect content are marked, and the multiple surface defect events of the mobile phone battery are determined according to each surface defect content, the corresponding surface defect feature, and the surface use event of the mobile phone battery, which is compatible with the overall consideration of each surface defect content, the corresponding surface defect feature, and the surface use event of the mobile phone battery, and ensures the accuracy of the multiple surface defect events of the mobile phone battery.

[0053] At this time, the edge device traverses and deeply analyzes the surface defect twin model, uses three-dimensional geometric analysis algorithm and pattern recognition algorithm to perform semantic classification on the abnormal grid body in the model; the system calls a predefined defect semantic dictionary, maps the geometric body to a label with clear physical meaning according to the geometric topological characteristics of the abnormal area, such as volume, depth, aspect ratio and surface roughness, for example, defines a concave with a depth exceeding a threshold and an edge steep as a “corrosion pit”, and defines a strip-shaped protrusion as a “liquid injection residual protrusion”, so as to determine the surface defect content.

[0054] The system extracts and labels the corresponding surface defect features of each defect instance, which include geometric features, appearance features and relative position description; the system strongly binds the extracted feature vector and the unique ID of the defect to generate a feature-content association pair, realizing the conversion from qualitative cognition to quantitative representation.

[0055] The system introduces a time axis and a working condition background, reads the surface use event log of the mobile phone battery, analyzes the correlation between “surface defect content + features” and “surface use event” through association rule mining or logic reasoning engine; if the detected features and the specific use event are consistent in time and logic, the system determines that the defect is caused by the event, and finally encapsulates the defect content, defect features and triggering scene into a structured surface defect event, providing context information for subsequent risk assessment.

[0056] Specifically, the operation and maintenance log of the mobile phone battery shows that it has experienced a “mechanical arm grabbing action of the automatic assembly line” one hour ago, and the pressure sensor reading of the action is abnormally high; the intelligent edge device analyzes the surface defect twin model of the mobile phone battery, and detects a nonlinear, irregular edge local concave grid on the side of the battery; the system calculates the volume and maximum depth of the concave, finds that its depth exceeds 50% of the thickness tolerance of the battery shell, refers to the defect semantic dictionary, and determines the geometric anomaly as “shell structural concave”, and determines it as surface defect content_1.

[0057] For “shell structural concave”, the intelligent edge device extracts its corresponding surface defect features, including the position coordinates located at the center of the side of the battery, the morphological parameters of the maximum depth 0.3mm and the surface area 20mm^2, and the appearance description of the edge showing extrusion accumulation characteristics and the slight peeling of the surface coating, and the system binds these feature data with labels to complete the feature labeling.

[0058] The intelligent edge device reads the surface use event of the mobile phone battery and finds that there is a record of "mechanical arm grabbing overload" one hour ago; the logic engine of the system performs space-time matching and confirms that the indentation position corresponds to the usual grabbing contact point of the mechanical arm and is highly consistent with the grabbing event in time; the system determines that the indentation is not a material defect but a damage caused by external mechanical stress, and finally generates a surface defect event Event_A, which records that the shell structure is indented, the feature is described as 0.3mm deep and located at the grabbing point, and the cause / background is marked as caused by abnormally high pressure grabbing by the mechanical arm. The event is marked as high risk because it is an external damage and may have the risk of internal micro-crack expansion, thereby guiding the subsequent maintenance decision.

[0059] Further, based on the identification of each surface defect event, the corresponding surface defect region is determined, and at the same time, temperature detection is performed along the surface defect region, and corresponding temperature data is output, the first-dimensional defect content is determined according to the plurality of surface defect events and the region form of the corresponding surface defect region, the second-dimensional defect content is determined according to the plurality of surface defect events and the corresponding temperature data, and the multi-dimensional defect system is determined based on the first-dimensional defect content and the second-dimensional defect content. The overall consideration of the first-dimensional defect content and the second-dimensional defect content ensures the accuracy of the multi-dimensional defect system, and further controls the surface defect event, realizes the overall consideration of the plurality of surface defect events, the corresponding surface defect region and the corresponding temperature data, and improves the accuracy of the multi-dimensional defect system.

[0060] At this time, the edge device reads the surface defect event, extracts the accurate spatial position information of the surface defect feature contained therein, and locks the minimum bounding box or specific contour in the three-dimensional space of the mobile phone battery based on the coordinates, which is defined as the surface defect region to be detected; the system triggers the integrated thermal imaging acquisition module, performs coordinate remapping algorithm to convert the defect region coordinates in the visible light coordinate system to the thermal imaging sensor coordinate system, and controls the thermal sensor to align the region for high-precision thermal radiation energy acquisition.

[0061] The edge device obtains the original thermal image data through the infrared sensor, performs the emissivity correction and environmental temperature compensation algorithm to eliminate the influence of material emissivity difference and environmental reflection; the system collects the temperature matrix in the surface defect region, calculates the key statistics including the highest temperature, the lowest temperature, the average temperature and the temperature gradient, and performs space-time registration on the temperature data, which corresponds the thermal image pixels to the grid nodes of the defect twin model one by one, and outputs the temperature data with spatial attribute labels.

[0062] The system focuses on the integrity evaluation of the geometric structure, calculates the influence of the morphology on the structural strength of the battery according to the regional morphological parameters such as depth, area, aspect ratio and edge curvature change, uses a fracture mechanics model or a geometric damage evaluation algorithm, maps the quantitative morphological parameters to the structural risk level, and constructs the first dimension of defect content describing the "geometric damage degree".

[0063] The system focuses on the abnormality evaluation of the physical state, judges the temperature abnormal source according to the abnormal patterns in the temperature data such as local hot spots, overall temperature rise or uneven temperature distribution, and uses a thermal runaway early warning model in combination with the electrochemical characteristics of the battery to judge the temperature abnormal source, maps the amplitude and rate of the temperature abnormality to the thermal risk level, and constructs the second dimension of defect content describing the "thermal runaway possibility"; the system uses a multi-modal data fusion algorithm to synthesize the contents of the first and second dimensions, comprehensively considers the "whether the structure is damaged" and "whether the temperature is too high" two dimensions, and generates a multi-dimensional defect system containing composite risk factors.

[0064] Specifically, the intelligent edge device integrates a miniature infrared thermal imager, and the two defect events of "insulation layer damage" on the top surface and "virtual welding" at the tab have been identified in the previous steps; in the defect region dynamic locking based on event mapping and thermal sensing triggering stage, the intelligent edge device reads Event_01 (insulation layer damage) to obtain the center coordinates of the top surface (40mm, 30mm), the system calculates the corresponding coordinates in the field of view of the infrared thermal imager, locks a 10mm x 10mm area as the surface defect region_A; for Event_02 (tab virtual welding), the system locks the tab root area as the surface defect region_B in the field of view of the thermal imager, and then the thermal imager triggers to collect the thermal radiation signals of the two specific regions.

[0065] For region A (damage), the thermal imager detects that the temperature matrix center point value is 52.5°C, and the system records the temperature data of this point and calculates that it is 12°C higher than the surrounding normal area; for region B (virtual welding), the thermal imager detects that there is a small high temperature point of 68.2°C at the root of the tab, while the temperature of the battery body is only 38°C, and the system accurately binds these temperature values to Event_01 and Event_02; The intelligent edge device analyzes the regional morphology corresponding to Event_01 and finds that the insulation layer presents a peeling shape and the metal substrate is exposed, with an area of 5mm², and determines that the structure is damaged to a "medium" degree according to the geometric damage model, and the first dimension of defect content_01 is defined as "insulation failure"; for the regional morphology corresponding to Event_02, the system finds that there is irregular accumulation on the surface of the welding point, the area is small but located in the key current path, and determines that the structural risk is "high", and the first dimension of defect content_02 is defined as "geometric abnormality at key connection".

[0066] The intelligent edge device analyzes the temperature data, determines that the temperature rise of 12°C of Event_01 may exist a small leakage, and defines the thermal risk as "medium"; the temperature rise of more than 30°C of Event_02 belongs to the typical thermal accumulation caused by excessive contact resistance, and the thermal risk is defined as "extremely high", thereby constructing the second dimension defect content; the intelligent edge device performs multi-dimensional synthesis, fuses "high structural risk" and "extremely high thermal risk" for Event_02, and marks it as "emergency blocking level defect" in the multi-dimensional defect system; for Event_01, it is marked as "monitoring level defect", and the system clearly defines the maintenance priority: the dummy ear virtual welding must be processed first, and then the insulating layer is repaired.

[0067] Reference Figure 5 In step S14, the specific steps are: S141: determining the use history of the mobile phone battery based on the detection of the mobile phone battery, determining a plurality of damaged behaviors according to the identification of the use history of the mobile phone battery, determining a defect distribution framework according to the multi-dimensional defect system and the plurality of damaged behaviors, and determining a defect distribution map of each surface defect area according to the matching of the defect distribution framework and the use history of the mobile phone battery; S142: determining the corresponding priority coefficient based on the comparison of each surface defect area to output the area priority of each surface defect area, and collecting the past maintenance events of the mobile phone battery; S143: determining the primary maintenance route of the mobile phone battery according to the defect distribution map of each surface defect area and the past maintenance events of the mobile phone battery, and determining the defect maintenance route of the mobile phone battery according to the area priority of each surface defect area and the primary maintenance route of the mobile phone battery.

[0068] In the embodiment of the present application, the use history of the mobile phone battery is determined based on the detection of the mobile phone battery, a plurality of damaged behaviors are determined according to the identification of the use history of the mobile phone battery, a defect distribution framework is determined according to the multi-dimensional defect system and the plurality of damaged behaviors, and a defect distribution map of each surface defect area is determined according to the matching of the defect distribution framework and the use history of the mobile phone battery, which is compatible with the overall consideration of the matching of the defect distribution framework and the use history of the mobile phone battery, and ensures the accuracy of the defect distribution map of each surface defect area.

[0069] At this time, the edge device integrates multi-dimensional data streams such as the number of charge and discharge cycles, the charge and discharge rate curve, the historical working temperature fluctuation range, and the mechanical impact record by reading the control chip log built-in the mobile phone battery and the record of the external Internet of Things system; the system reconstructs these discrete data points into a continuous life cycle behavior trajectory using time series alignment technology, extracts key timestamp nodes, and forms a complete battery use history archive; then, the damaged behavior identification based on time series anomaly mining is carried out, the system scans the history data using an anomaly detection algorithm, identifies damaged behaviors that exceed the normal working condition range, including overvoltage charging, deep over-discharge, large current instantaneous impact, and other electrochemical damaged behaviors, as well as instantaneous acceleration mutation, long-term high temperature and high humidity environment, etc. physical and mechanical damaged behaviors, and calculates the strength and duration of the behavior, which is marked as a stress source that leads to potential defects.

[0070] The system performs causal correlation analysis on the multi-dimensional defect system and the identified damaged behaviors, and establishes a "behavior-defect" mapping relationship using knowledge graph reasoning or failure physical model; based on these mapping rules, the system constructs a descriptive spatial logic model that specifies which type of damaged behavior should lead to which type of defect in the three-dimensional space of the battery and the probability distribution pattern of the defect in space.

[0071] The system substitutes the accurate spatial coordinates of each surface defect area into the constructed defect distribution framework for matching verification; the system performs spatial superposition operation to project the specific defect position into the logical space of the framework, and according to the coincidence, enhances the confidence or marks it as an abnormal distribution, finally generates a visual defect distribution map, which not only marks the physical location of the defect, but also annotates the corresponding cause behavior of each defect area through color coding or layer data, realizes the fusion of spatial positioning and cause tracing of defects.

[0072] Specifically, the intelligent edge device reads the BMS chip record of the mobile phone battery, and knows that the mobile phone battery has two defects of top cover edge indentation and positive electrode tab root discoloration / ablation; the intelligent edge device reads the BMS log and production line MES system record of the mobile phone battery, and reconstructs the use history of the mobile phone battery: records an acceleration peak of 30G impact event on the 50th day after leaving the factory, and records a 4.5C super high rate discharge event lasting 10 minutes on the 60th day.

[0073] For the impact peak of 30G, the system reference mechanical safety threshold identifies the damaged behavior A: "mechanical impact / suspected drop"; for the high rate discharge of 4.5C, the system reference electrochemical safety window identifies the damaged behavior B: "electric overload / high current thermal impact"; in the distribution framework construction stage based on the association of multi-dimensional defect system and damaged behavior, the system calls the failure physical model library for logical judgment: damaged behavior A (mechanical impact) usually causes plastic deformation of rigid structures such as battery edges and corners; damaged behavior B (electric overload) usually causes joule heating effect in high impedance areas such as tab welding points and bus bars; based on this, the intelligent edge device constructs the defect distribution framework, defines the "top cover edge" of the battery as a high-risk area of mechanical stress, and defines the "positive tab root" as a high-risk area of electro-thermal stress.

[0074] The system substitutes the two actually detected surface defect areas into the framework, and the indentation position of the top cover edge perfectly matches the "mechanical stress high-risk area" in the framework, so the system marks it as "mechanical impact damage" in the distribution map; the discoloration / ablation position of the positive tab root perfectly matches the "electro-thermal stress high-risk area" in the framework, so the system marks it as "electric overload damage"; the intelligent edge device outputs the defect distribution map of the mobile phone battery, which not only contains the geometric contour of the defect, but also has two semantic labels:

edge indentation-cause: drop impact

tab ablation-cause: large current overload

[0075] At this time, the behavior-defect mapping table is as shown in Table 1:

[0076] Further, based on the comparison of each surface defect area, the corresponding priority coefficient is determined to output the area priority of each surface defect area, and the past maintenance events of the mobile phone battery are collected, which is compatible with the overall consideration of the comparison of each surface defect area, ensuring the accuracy of the corresponding priority coefficient.

[0077] At this time, the edge device performs horizontal comparison analysis on the identified each surface defect area, and calls the quantitative data in the multi-dimensional defect system to construct a priority evaluation function; the function comprehensively considers key parameters such as morphology severity, thermodynamic abnormality and key position weight, and the system adopts multi-criteria decision algorithms such as analytic hierarchy process or weighted summation model to linearly or nonlinearly weight and fuse the above parameters, and calculates the priority coefficient of each surface defect area, whose value range is usually normalized to [0, 1] or [0, 100].

[0078] After obtaining the priority coefficients of all surface defect areas, the system executes a sorting algorithm to arrange all defect areas in descending order according to the size of the coefficient values; the system sets one or more decision thresholds to divide the sorted areas into different risk levels, and generates an area priority list, which not only contains the sorting results, but also the priority labels of each area, converts the quantified coefficients into intuitive execution instructions, and clearly guides the allocation order of subsequent maintenance resources.

[0079] The edge device accesses the digital footprint or background maintenance database of the mobile phone battery through the data interface, executes historical data retrieval to collect the past maintenance events of the battery; the system collects detailed data including timestamps, operation types, maintenance results, and component replacements, and converts these unstructured or semi-structured historical data into standardized context feature vectors, which are loaded into the current evaluation memory to provide a basis for route correction in subsequent steps.

[0080] Specifically, three surface defect areas are detected: a slight scratch on the top edge (area A), discoloration and a small bump at the positive tab root (area B), and a large area of depression on the bottom surface (area C); in the priority coefficient calculation stage based on multi-dimensional feature weighting, the intelligent edge device runs the priority evaluation algorithm; for area B (tab root), the system identifies that it is located in the "current transmission critical path" and has "severe thermal anomaly" (temperature 65°C), and calculates the priority coefficient Pc(B)=0.95 (extremely high); for area C (bottom surface depression), the system identifies that it has "deep structural damage" but no thermal risk and is not a critical functional area, and calculates Pc(C)=0.60 (medium-high); for area A (top surface scratch), the system identifies that it only affects the appearance and has no structural risk and thermal risk, and calculates Pc(A)=0.15 (low); The system sorts the three areas in descending order as B>C>A; the intelligent edge device outputs the area priority list: Rank1 is area B, with a level label of "urgent, thermal runaway risk"; Rank2 is area C, with a level label of "high risk, structural strength risk"; Rank3 is area A, with a level label of "observation, cosmetic flaw".

[0081] The intelligent edge device connects to the cloud database by scanning the two-dimensional code or RFID tag of the mobile phone battery to collect past maintenance events; the system obtains record 1 showing that area B (tab root) had a laser repair welding maintenance due to "virtual welding" three months ago, and the record after maintenance shows that "impedance decreased but rebounded recently"; record 2 shows that areas A and C were first discovered and had no maintenance history; the intelligent edge device loads the historical context of "area B has been repaired and the problem has recurred" into the current task, which provides a key decision basis for the subsequent step of "area B may need to be replaced rather than repaired again".

[0082] Therefore, the preliminary maintenance route of the mobile phone battery is determined according to the defect distribution map of each surface defect area and the previous maintenance event of the mobile phone battery, the defect maintenance route of the mobile phone battery is determined according to the area priority of each surface defect area and the preliminary maintenance route of the mobile phone battery, the overall consideration of the area priority of each surface defect area and the preliminary maintenance route of the mobile phone battery is compatible, and the accuracy of the defect maintenance route of the mobile phone battery is ensured.

[0083] At this time, the system generates a basic traversal path according to geometric space logic and historical experience; the edge device extracts the three-dimensional space coordinates of all surface defect areas based on the defect distribution map, and calls a path planning algorithm to calculate a space path capable of traversing all defect points with the shortest total movement path or the lowest mechanical arm adjustment energy consumption as the target.

[0084] Meanwhile, the system introduces previous maintenance events as a hard constraint condition, and if historical records show that a defect area has been repaired multiple times but failed, the system marks it as needing special station processing or recommends scrapping, and automatically skips these inefficient repair points or arranges them at the end of the path in path planning; the finally generated preliminary maintenance route mainly considers spatial accessibility and historical feasibility, and establishes a basic geometric framework for maintenance operation.

[0085] The system introduces the urgency dimension on the basis of the preliminary route, and introduces the area priority output by S142 as a weight factor into the path planning model; the system uses a weighted scheduling algorithm to correct the node order in the preliminary maintenance route, and the correction logic follows the principle of “high-risk priority blocking”; if a high-priority defect is placed in a later position in the preliminary route, the system will move it to the head of the queue, even if this will increase the movement distance of the mechanical arm or cause path backtracking; at the same time, the system dynamically plans pre-operation and post-operation according to the type of high-priority defect, and finally outputs the defect maintenance route, which includes time axis, operation sequence, tool switching instruction and emergency plan.

[0086] Specifically, the intelligent edge device is equipped with a three-axis mechanical arm and a laser repair head, the defect distribution map shows three defect points of the top left (point 1), the right of the tab (point 2) and the center of the bottom (point 3), and the previous maintenance event shows that point 2 belongs to “stubborn thermal failure”, and the area priority shows that point 2 is the highest (emergency); in the preliminary maintenance route generation stage based on spatial topology and historical constraints, the intelligent edge device plans based on geometric distance, finds that the total movement trajectory of the mechanical arm is shortest in the order of “top left (point 1) > right of tab (point 2) > center of bottom (point 3)”, which conforms to the preliminary maintenance route of “from top to bottom”.

[0087] The system found that point 2 had a record of "welding failure" in the analysis of past maintenance events, so it preset the logic in the primary route: this point is no longer suitable for regular "laser welding" operation, but is marked as a candidate point for "replacement of spare parts" or "deep disassembly", and to prevent interference with other operations, it is defaulted to the end of the route to facilitate isolation.

[0088] The system loads the area priority data and detects that point 2 has a very high thermal risk, which is an emergency failure that may cause thermal runaway; the system executes the scheduling algorithm, in order to eliminate the high temperature risk, breaks the "shortest path" principle in space, and raises the operation priority of point 2 to the highest; the final output of the defect maintenance route is corrected as: Step 1 directly moves the mechanical arm to the right side of the lug (point 2), the system no longer performs regular welding, but performs "emergency shutdown" or "heat insulation treatment", and simultaneously marks the point to be transferred to the scrap process; Step 2 moves to the center of the bottom surface (point 3) for structural reinforcement treatment; Step 3 moves to the left side of the top surface (point 1) for simple surface polishing repair; Although this route increases the movement distance of the mechanical arm, it ensures that the most dangerous defect is handled first, thereby avoiding the possibility of thermal explosion of the battery when handling low-risk scratches.

[0089] Reference Figure 6 In step S15, the specific steps are: S151: dynamically identify the defect maintenance route, and determine a plurality of defect maintenance areas in the identification process, determine a corresponding defect maintenance node according to the area position, corresponding area form and corresponding defect maintenance content of each defect maintenance area, to collect a plurality of defect maintenance nodes; S152: determine the node position of each defect maintenance node based on the identification of each defect maintenance node, and simultaneously determine the corresponding node priority according to the comparison of each defect maintenance node, determine the first dynamic maintenance content according to the node position of each defect maintenance node and the service life of the mobile phone battery; S153: determine the second dynamic maintenance content according to the node priority of each defect maintenance node and the service life of the mobile phone battery, determine the corresponding dynamic maintenance event based on the first dynamic maintenance content and the second dynamic maintenance content, determine a plurality of dynamic maintenance items according to the identification of the dynamic maintenance event, determine the corresponding dynamic maintenance table based on the plurality of dynamic maintenance items, the maintenance process of the mobile phone battery and the corresponding current maintenance progress;

[0090] In the embodiments of this application, the defect maintenance route is dynamically identified, and multiple defect maintenance areas are determined during the identification process. Based on the location, shape, and content of each defect maintenance area, a corresponding defect maintenance node is determined to collect multiple defect maintenance nodes. This approach takes into account the location, shape, and content of each defect maintenance area, ensuring the accuracy of the corresponding defect maintenance nodes.

[0091] At this point, the edge device performs real-time analysis and dynamic identification of the generated defect maintenance route. The system combines the kinematic model of the actuator and the physical properties of the work tool to perform spatial buffer analysis. When the virtual end of the actuator moves along the route, the system calculates the spatial envelope of the actual effect of the tool on the battery surface in each pose, and performs Boolean intersection operation in combination with the geometric boundary of the surface defect area to accurately cut out discrete spatial segments that need to be physically operated from the route. These spatial segments are defined as defect maintenance areas, each containing precise work boundaries and normal vectors to ensure that the tool's range of action completely covers the defect and does not damage the normal area.

[0092] The system performs multi-dimensional feature extraction for each defined defect maintenance area, and performs parametric modeling by integrating three types of attributes: area location, area morphology, and defect maintenance content. The system extracts the precise three-dimensional coordinates and surface normal vectors of the maintenance area in the battery world coordinate system, analyzes the surface curvature, flatness, and spatial geometric topology of the area, and matches the corresponding process parameter set according to the maintenance type. These attributes are encapsulated into structured data objects, namely defect maintenance nodes. The system traverses the entire route, collects and serializes all nodes, and forms the basic execution queue of maintenance tasks.

[0093] Specifically, the intelligent edge device is equipped with a precision laser welding robotic arm. It is known that mobile phone batteries have two defects: a poor weld at the root of the tab (requiring laser welding) and scratches on the top insulation layer (requiring laser cladding repair). In the maintenance area spatial calculation stage based on dynamic route scanning, the intelligent edge device analyzes the maintenance route and simulates the movement of the laser head along the planned path.

[0094] For the poorly welded area of ​​the electrode tab, the system calculates the range of effect that must be covered by combining the laser welding head spot diameter and the geometric distribution of the poorly welded points. In three-dimensional space, a cylindrical space segment including the connection at the root of the electrode tab is defined as the defect maintenance area_A, and a heat-affected zone margin is reserved. For the scratch on the top surface, the system generates a rectangular strip space that covers the entire length of the scratch and is slightly wider than the scratch on the scratch trajectory, based on the length trajectory of the scratch and the width of the cladding nozzle. This strip space is defined as the defect maintenance area_B.

[0095] For region_A (the lug), the system extracts coordinates and normal vectors, identifies it as a special-shaped metal lap joint surface, matches the "laser deep penetration welding" process, generates a defect maintenance node_Node_01, and its instruction set includes positioning the TCP to the specified coordinates and adjusting the posture, calling the laser power parameter_P1 to perform spot welding; for region_B (top surface scratch), the system extracts the start and end points of the scratch track, identifies it as a linear groove on the plane, matches the "laser cladding" process, generates a defect maintenance node_Node_02, and its instruction set includes controlling the TCP to move along a straight line from the start point to the end point, setting the speed and calling the laser power parameter_P2 for continuous scanning cladding; the intelligent edge device has collected the serialized defect maintenance node set [Node_01, Node_02], ready to be sent to the motion controller for execution.

[0096] Further, based on the identification of each defect maintenance node, the node position of each defect maintenance node is determined, and according to the comparison of each defect maintenance node, the corresponding node priority is determined, and the first dynamic maintenance content is determined according to the node position of each defect maintenance node and the service life of the mobile phone battery, which is compatible with the overall consideration of the node position of each defect maintenance node and the service life of the mobile phone battery, and ensures the accuracy of the first dynamic maintenance content.

[0097] At this time, the edge device analyzes the generated defect maintenance node and extracts the geometric pose information therein; the system converts the region position defined in the node from the local coordinate system to the world coordinate system to eliminate errors caused by the placement angle of the battery or the deviation of the clamp, and combines the hand-eye calibration matrix to accurately map the two-dimensional image coordinates recognized by vision to the three-dimensional space coordinates of the end effector of the robot arm; the node position determined in this way includes three-dimensional values and attitude vectors, which ensures that the execution tool can cut into the work point at the correct angle.

[0098] The system compares the transverse characteristics of all analyzed defect maintenance nodes to determine the execution order; the system constructs a multi-parameter evaluation function, and the input variables include defect severity from the multi-dimensional defect system, region criticality of the node position, and operation dependency, the system arranges the nodes in descending order according to the weighted scores of these variables, calculates the real-time node priority, and the node with high priority will be marked as "immediate execution", while the node with low priority may be suspended or delayed.

[0099] The system reads the health status and remaining cycle life data of the mobile phone battery, executes a cost-benefit analysis algorithm, and compares the importance of the node position with the remaining life of the battery: if the battery is in the early life stage, the system determines the first heavy content to be "standard repair" or "enhanced repair"; if the battery is in the end-of-life stage, for non-critical node positions, the system may adjust the maintenance content to "low-cost stop-loss" or "simple plugging"; the system fuses "position characteristics" and "life status" to determine the first heavy dynamic maintenance content that describes the depth of physical work.

[0100] Specifically, two defect maintenance nodes are known to have been generated: Node_A (tab virtual welding) and Node_B (top surface scratch), and the health of the mobile phone battery is detected by card reading to be 88%; the intelligent edge device analyzes Node_A, extracts its pixel position in the visual camera field of view, converts it to coordinates in the base coordinate system of the mechanical arm and locks the normal vector vertically upward; Node_B is analyzed, and the trajectory center of the top surface scratch is extracted to obtain a three-dimensional coordinate sequence and lock the starting point.

[0101] The system compares the nodes, Node_A is located in the core current path and has a virtual welding risk, with high severity; Node_B is located in the top surface decoration area, only affecting the appearance and local insulation, with low severity; the system calculates the node priority, the priority coefficient of Node_A is 0.95 (highest level), and the priority coefficient of Node_B is 0.30 (ordinary level), and it is decided that Node_A must be executed before Node_B.

[0102] The system analyzes the service life of the mobile phone battery, given that the battery health is acceptable and has not reached the scrap threshold, Node_A is located in the key tab position, and the system determines to perform "standard precision repair" on Node_A, i.e. the first heavy dynamic maintenance content is defined as using high-power laser to perform complete penetration welding to ensure that the conductive cross-section is restored to 100% standard; for Node_B, although the position is relatively secondary, but considering that the battery still has a long life, a simple repair may fall off again in use, and the system determines to perform "durable repair" on it, i.e. the first heavy dynamic maintenance content is defined as performing double-layer insulation coating, rather than simple single-layer covering.

[0103] Therefore, the second dynamic maintenance content is determined according to the node priority of each defect maintenance node and the service life of the mobile phone battery, the corresponding dynamic maintenance event is determined based on the first dynamic maintenance content and the second dynamic maintenance content, the plurality of dynamic maintenance items is determined according to the identification of the dynamic maintenance event, the corresponding dynamic maintenance table is determined based on the plurality of dynamic maintenance items, the maintenance process of the mobile phone battery and the corresponding current maintenance progress, the overall consideration of the plurality of dynamic maintenance items, the maintenance process of the mobile phone battery and the corresponding current maintenance progress is compatible, the accuracy of the corresponding dynamic maintenance table is ensured, at the same time, the defect maintenance route of the mobile phone battery is further controlled, each defect maintenance node is strictly controlled, the accuracy of the dynamic maintenance event is improved, and the corresponding dynamic maintenance table is output.

[0104] At this time, the system makes logical decisions by comprehensively considering the node priority and the service life state of the mobile phone battery, and balances between job urgency and residual value retention; if it is high priority with high life, the content is defined as a standard execution mode; if it is high priority with low life, the content is defined as a fast loss mode; if it is low priority with low life, the content is defined as a skip or delay mode, so as to determine whether the maintenance job is full repair, emergency treatment or abandonment; then, the dynamic maintenance event instantiation based on double content fusion is carried out, the system performs semantic fusion and instantiation on the determined first dynamic maintenance content and the second dynamic maintenance content determined in this step; through the mapping relationship, the system generates specific dynamic maintenance events, each dynamic maintenance event contains event ID, target node, job type, execution mode, process parameter set and allowed time window, and abstract maintenance content is converted into events with execution conditions.

[0105] The system analyzes the dynamic maintenance event and calls the process knowledge base, and the event is disassembled into a discrete action sequence arranged in time sequence; each action is a dynamic maintenance item, and each item contains specific motion instructions and I / O control signals, and the high-level maintenance event is issued to the bottom controller; the system uses a real-time scheduling algorithm, combines the maintenance process of the mobile phone battery and the current maintenance progress, and generates a final dynamic maintenance table; the maintenance table is a dynamically updated time sequence instruction set, the system calculates the expected start time and the expected end time of each item according to the current system load, the kinematics state of the mechanical arm and the execution result of the previous item, forms a Gantt chart execution plan table accurate to milliseconds, and synchronizes the execution mechanism in real time.

[0106] Specifically, the first content of Node_A (the virtual welding of the gull wing) is known as "standard depth welding" and has the highest priority, the first content of Node_B (the top surface scratch) is "double-layer coating" and has a lower priority, and the battery SOH is 88%; for Node_A (high priority + SOH 88%), the system determines that this is a core fault and the battery is worth repairing, and determines that the second dynamic maintenance content is "full-precision closed-loop execution", which needs to be combined with real-time temperature monitoring; for Node_B (low priority + SOH 88%), since the risk of Node_A is very high, the system determines the second dynamic maintenance content as "conditional trigger execution" according to the principle of conservation of resources, that is, only when Node_A is processed and there is enough time, it will be executed; in the dynamic maintenance event instantiation stage based on the fusion of double content, the system fuses the first and second contents to generate dynamic maintenance events Event_A: <target: gull wing; mode: precision welding; need to start temperature control feedback; parameter: power P = Max> and Event_B: <target: top surface; mode: standby / standard coating; parameter: if T < limit, execute>.

[0107] The system decomposes Event_A into a dynamic maintenance project sequence including fast movement of the mechanical arm to a safe point above the gull wing, slow exploration to the welding focus plane, starting the laser and activating the real-time temperature control sensor, executing the welding trajectory scanning, turning off the laser and lifting the reset; decomposes Event_B into a preliminary project sequence, temporarily stored in the buffer queue; the intelligent edge device monitors the maintenance progress of 0% to start generating a dynamic maintenance table, allocates Project_A1 at T = 0ms, and then executes in sequence; when T = 2000ms, the system detects that Project_A4 (welding) is completed and the temperature control feedback is normal, the maintenance process is updated, the system calculates that the remaining time is sufficient, immediately changes the state of Event_B from "on hold" to "ready"; T = 2100ms, the dynamic maintenance table automatically inserts Project_B1, instructing the mechanical arm to turn to the top surface scratch; the final output of the dynamic maintenance table is a real-time instruction stream that dynamically adjusts subsequent projects according to the current maintenance progress as time progresses.

[0108] Please refer to Figure 7 , Figure 7 is a structural composition diagram of a visual detection system of a smart edge device for a mobile phone battery in an embodiment of the present application; the visual detection system of the smart edge device for the mobile phone battery comprises: a visual detection module 21, configured to perform visual detection on the mobile phone battery by a plurality of cameras of the smart edge device when the mobile phone battery enters a shooting space of the smart edge device, and determine a plurality of surface images of the mobile phone battery in the visual detection; a surface defect twin model module 22 configured to determine a plurality of surface defect features based on image recognition of each surface image, determine a corresponding surface defect twin model based on feature positions of the plurality of surface defect features, corresponding feature morphologies, and an overall morphology of the mobile phone battery; a multi-dimensional defect system module 23 configured to determine a plurality of surface defect events of the mobile phone battery based on detection of the surface defect twin model, determine a multi-dimensional defect system based on the plurality of surface defect events, corresponding surface defect regions, and corresponding temperature data; a defect maintenance route module 24 configured to determine a defect distribution map of each surface defect region based on the multi-dimensional defect system, a usage history of the mobile phone battery, and corresponding damaging behaviors, determine a defect maintenance route of the mobile phone battery based on the defect distribution map of each surface defect region, corresponding region priorities, and past maintenance events of the mobile phone battery; a dynamic maintenance module 25 configured to determine a plurality of defect maintenance nodes based on recognition of the defect maintenance route, determine a dynamic maintenance event based on node positions of each defect maintenance node, corresponding node priorities, and a service life of the mobile phone battery, and output a corresponding dynamic maintenance table.

[0109] Any combination of the technical features in the above embodiments is possible. In order to make the description concise, not all combinations of the technical features in the above embodiments are described, however, as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present disclosure.

Claims

1. A visual inspection method for mobile phone batteries using intelligent edge devices, characterized in that, The method comprises the following steps: When the mobile phone battery enters the shooting space of the intelligent edge device, the multiple cameras of the intelligent edge device perform visual detection on the mobile phone battery, and multiple surface images of the mobile phone battery are determined in the visual detection; Based on the image recognition of each surface image, multiple surface defect features are determined, and based on the feature position, corresponding feature morphology and overall morphology of the mobile phone battery, a corresponding surface defect twin model is determined; the surface defect twin model not only contains the geometric information of the mobile phone battery, but also contains a group of discrete digital twins with space-time attributes and physical characteristics; Based on the detection of the surface defect twin model, multiple surface defect events of the mobile phone battery are determined, and based on the multiple surface defect events, corresponding surface defect regions and corresponding temperature data, a multi-dimensional defect system is determined; Based on the multi-dimensional defect system, the use history of the mobile phone battery and the corresponding damaged behavior, a defect distribution map of each surface defect region is determined, and based on the defect distribution map of each surface defect region, the corresponding region priority and the past maintenance events of the mobile phone battery, a defect maintenance route of the mobile phone battery is determined, which includes a time axis, an operation sequence, a tool switching instruction and an emergency plan; Based on the recognition of the defect maintenance route, multiple defect maintenance nodes are determined, and based on the node position, corresponding node priority and service life of each defect maintenance node, a dynamic maintenance event is determined, and a corresponding dynamic maintenance table is output.

2. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 1 wherein, The method comprises the following steps: When the mobile phone battery enters the shooting space of the intelligent edge device, the multiple cameras of the intelligent edge device perform visual detection on the mobile phone battery, and multiple surface images of the mobile phone battery are determined in the visual detection; The method comprises the following steps:

3. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 1 wherein, In the visual detection system, based on the recognition of the visual detection system, multiple visual detection dimensions are determined, and based on each visual detection dimension, the overall morphology of the mobile phone battery and each surface of the mobile phone battery, multiple surface images of the mobile phone battery are determined. The method comprises the following steps: In each surface image, based on the detection of the surface image, multiple sub-surface regions are determined, and based on the recognition of each sub-surface region, a corresponding surface defect feature is determined to collect multiple surface defect features; Based on the recognition of each surface defect feature, the feature position of the surface defect feature is determined, and the feature morphology of the surface defect feature is marked, and based on the feature position of the multiple surface defect features and the overall morphology of the mobile phone battery, a first repositioning model is determined. According to the feature morphology of the plurality of surface defect features and the overall morphology of the mobile phone battery, a first heavy morphology model is determined, and a corresponding surface defect twin model is determined based on the multi-dimensional synthesis of the first heavy position model and the first heavy morphology model.

4. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 1, wherein, According to the detection of the surface defect twin model, a plurality of surface defect events of the mobile phone battery are determined, and a multi-dimensional defect system is determined based on the plurality of surface defect events, the corresponding surface defect area, and the corresponding temperature data, including: In the surface defect twin model, according to the identification of the surface defect twin model, a plurality of surface defect contents of the mobile phone battery on each surface are determined, and the corresponding surface defect features of each surface defect content are marked, and a plurality of surface defect events of the mobile phone battery are determined based on each surface defect content, the corresponding surface defect feature, and the surface use event of the mobile phone battery.

5. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 4, wherein, According to the detection of the surface defect twin model, a plurality of surface defect events of the mobile phone battery are determined, and a multi-dimensional defect system is determined based on the plurality of surface defect events, the corresponding surface defect area, and the corresponding temperature data, including: Based on the identification of each surface defect event, the corresponding surface defect area is determined, and at the same time, temperature detection is carried out along the surface defect area, and the corresponding temperature data is output, the first-dimensional defect content is determined based on the plurality of surface defect events and the area morphology of the corresponding surface defect area, the second-dimensional defect content is determined based on the plurality of surface defect events and the corresponding temperature data, and the multi-dimensional defect system is determined based on the first-dimensional defect content and the second-dimensional defect content.

6. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 1, wherein, According to the multi-dimensional defect system, the use history of the mobile phone battery, and the corresponding damaged behavior, the defect distribution diagram of each surface defect area is determined, and the defect maintenance route of the mobile phone battery is determined based on the defect distribution diagram of each surface defect area, the corresponding area priority, and the previous maintenance event of the mobile phone battery, including: Based on the detection of the mobile phone battery, the use history of the mobile phone battery is determined, a plurality of damaged behaviors are determined based on the identification of the use history of the mobile phone battery, a defect distribution framework is determined based on the multi-dimensional defect system and the plurality of damaged behaviors, and the defect distribution diagram of each surface defect area is determined based on the matching of the defect distribution framework and the use history of the mobile phone battery.

7. The method of visual inspection of mobile phone battery by the intelligent edge device as claimed in claim 6, wherein, According to the multi-dimensional defect system, the use history of the mobile phone battery, and the corresponding damaged behavior, the defect distribution diagram of each surface defect area is determined, and the defect maintenance route of the mobile phone battery is determined based on the defect distribution diagram of each surface defect area, the corresponding area priority, and the previous maintenance event of the mobile phone battery, including: Based on the comparison of each surface defect area, the corresponding priority coefficient is determined to output the area priority of each surface defect area, and the previous maintenance event of the mobile phone battery is collected; According to the defect distribution diagram of each surface defect area and the previous maintenance event of the mobile phone battery, the primary maintenance route of the mobile phone battery is determined, and the defect maintenance route of the mobile phone battery is determined based on the area priority of each surface defect area and the primary maintenance route of the mobile phone battery.

8. The method of visual inspection of mobile phone battery by intelligent edge device as claimed in claim 1, wherein, The method comprises the following steps: identifying the defect maintenance route, determining the defect maintenance nodes based on the identification of the defect maintenance route, determining the dynamic maintenance events according to the node positions of the defect maintenance nodes, the corresponding node priorities and the service life of the mobile phone battery, and outputting the corresponding dynamic maintenance table. The method comprises the following steps: dynamically identifying the defect maintenance route, determining a plurality of defect maintenance areas in the identification process, determining the corresponding defect maintenance nodes according to the area positions of the defect maintenance areas, the corresponding area forms and the corresponding defect maintenance contents, and collecting a plurality of defect maintenance nodes.

9. The method of visual inspection of mobile phone battery by the intelligent edge device according to claim 8, wherein, The method comprises the following steps: identifying the defect maintenance route, determining the defect maintenance nodes based on the identification of the defect maintenance route, determining the dynamic maintenance events according to the node positions of the defect maintenance nodes, the corresponding node priorities and the service life of the mobile phone battery, and outputting the corresponding dynamic maintenance table. The method comprises the following steps: determining the node positions of the defect maintenance nodes based on the identification of the defect maintenance nodes, determining the corresponding node priorities according to the comparison of the defect maintenance nodes, determining the first dynamic maintenance content according to the node positions of the defect maintenance nodes and the service life of the mobile phone battery; The method comprises the following steps: determining the second dynamic maintenance content according to the node priorities of the defect maintenance nodes and the service life of the mobile phone battery, determining the corresponding dynamic maintenance events based on the first dynamic maintenance content and the second dynamic maintenance content, determining a plurality of dynamic maintenance items according to the identification of the dynamic maintenance events, and determining the corresponding dynamic maintenance table based on the plurality of dynamic maintenance items, the maintenance process of the mobile phone battery and the corresponding current maintenance progress.

10. A visual inspection system for a mobile phone battery by a smart edge device, characterized in that, The visual detection system of the mobile phone battery applied to the intelligent edge device is used in the visual detection method of the mobile phone battery applied to the intelligent edge device.

Images