A multi-station coordinated automatic assembly processing system for electric toys

Abstract

The application relates to the technical field of electric toy assembly, in particular to a multi-station cooperative automatic assembly processing system for electric toys, which comprises a data acquisition module, which is used for determining the procedures of a plurality of stations for assembling electric toys, setting procedure nodes corresponding to the plurality of stations, configuring a management point and an acquisition point group for each procedure node, and acquiring a plurality of initial images of target accessories corresponding to the procedures based on the acquisition point group; a defect analysis module, which is used for determining the three-dimensional images of the target accessories corresponding to each procedure node based on the plurality of initial images, and simultaneously, unfolding the three-dimensional images of the target accessories corresponding to each procedure node to obtain the planar images corresponding to each procedure node; and a toy assembly module, which is used for acquiring the standard images corresponding to each procedure node, can reduce the single-procedure data operation amount, improve the defect detection response speed and real-time performance, and improve the yield of electric toy assembly.

Description

Technical Field

[0001] This invention relates to the field of electric toy assembly technology, specifically to a multi-station collaborative automated assembly and processing system for electric toys. Background Technology

[0002] With the rapid development of the electric toy market and the increasing demands of consumers for product diversification and refinement, the structure of electric toys is becoming increasingly complex. Their assembly involves numerous parts and intricate processes, placing higher demands on production efficiency and assembly precision. Currently, during the assembly of electric toys, each functional station (such as clamping and positioning, parts transfer, and bolt fastening) is largely independently controlled and operates sequentially. Each station often requires defect detection of the entire structure of the electric toy, resulting in a massive amount of data processing per process, high inspection load, and insufficient real-time response. Traditional methods can only remove defective products after assembly, failing to identify and sort abnormal parts at the station where defects occur, making it difficult to prevent the spread of defects at their source. Furthermore, they cannot optimize and adjust the equipment operating parameters of the corresponding workstations in real time based on the actual type of assembly defect, easily leading to the recurrence of similar assembly problems and severely impacting assembly production efficiency and overall product quality.

[0003] To address these issues, we propose a multi-station collaborative automated assembly and processing system for electric toys. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-station collaborative automated assembly and processing system for electric toys to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-station collaborative automated assembly and processing system for electric toys, comprising: The data acquisition module is used to determine the processes of multiple workstations used for assembling electric toys, and to set process nodes for multiple workstations, and to configure management points and data acquisition point groups for each process node. The defect analysis module is used to collect multiple initial images of the target parts for the corresponding process based on the collection point group, determine the 3D image of the target parts corresponding to each process node based on the multiple initial images, and unfold the 3D image of the target parts corresponding to each process node to obtain the 2D image corresponding to each process node. The toy assembly module is used to acquire standard images corresponding to each process node, set a registration body between the planar image and the standard image, determine the abnormal offset of the planar image relative to the standard image based on the registration body, determine the defect type and corresponding control parameters of the target part based on the abnormal offset, adjust the corresponding process based on the control parameters, and assemble the target part using the assembly equipment of the adjusted process node.

[0006] Preferably, the steps of determining the processes for assembling electric toys at multiple workstations, setting process nodes for each workstation, and configuring management points and data collection point groups for each process node include: Acquire and identify multiple workstations in the assembly process of electric toys and the process content of each workstation; According to the process content of each workstation, corresponding process nodes are generated in a preset assembly sequence, and a node identifier is assigned to each process node. Each process node is configured with a management point and multiple data collection points placed in different locations. The communication between the multiple data collection points is then connected to form a data collection point group.

[0007] Preferably, the step of acquiring multiple initial images of the target component corresponding to the corresponding process based on the acquisition point group, and determining the stereoscopic image of the target component corresponding to each process node based on the multiple initial images includes: Identify the feature points of the target part in the current process and mark multiple feature points; Multiple initial images of the target accessory are acquired from different angles based on the collection point group, wherein each initial image contains feature points captured from the corresponding angle; Determine the initial stereoscopic image corresponding to each initial image, and overlap multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process; Based on the 3D diagram of the current process, determine the 3D diagram of the target component corresponding to each process node.

[0008] Preferably, the step of determining the initial stereoscopic image corresponding to each initial image and superimposing multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process includes: Based on the initial image, the various acquisition surfaces of the target part are restored to obtain multiple local stereoscopic images of the target part. The same feature points in each local stereoscopic image are extracted. The same feature points refer to feature points that have the same positional identification in each local stereoscopic image. Based on the same feature points, the multiple local stereoscopic images are stitched together to generate a stereoscopic image of the target part corresponding to the current process. Calculate the spatial overlap of the same feature points between each surface; based on the comparison between the spatial overlap and the preset overlap threshold, determine whether there are any anomalies in the feature point data corresponding to each collection point; If it is determined that the feature point data of a certain collection point is abnormal, then the collection point is marked as an abnormal collection point, and the verification result is obtained.

[0009] Preferably, the step of determining the 3D view of the target component corresponding to each process node based on the 3D view of the current process includes: Obtain the 3D image of the target part in the current process and its multiple initial images, and transmit them to the management point corresponding to the next process node as the initial 3D image of the next process node; The initial image of the target part is collected based on the collection point group corresponding to the next process node. The initial image of the next process node is compared with multiple initial images transmitted in the current process. The initial images with different feature points are extracted. The local stereo images of the initial stereo images corresponding to the current process with different feature points are removed from the initial stereo images in the initial stereo images to obtain the effective initial stereo images. The initial image with different feature points corresponding to the next process node is restored into a valid local stereoscopic image. Based on the same feature points, the valid local stereoscopic images are stitched onto the valid initial stereoscopic image to obtain the stereoscopic image of the target part of the next process node. This process is repeated to obtain the stereoscopic images of the target parts corresponding to each process node.

[0010] Preferably, the step of acquiring the standard image corresponding to each process node and setting a registration body between the planar image and the standard image includes: Store the standard image of the target part corresponding to the process node in each management point, and obtain the feature points in the standard image as reference points; Obtain the planar images corresponding to each process node and multiple feature points in the planar images. Configure a selection circle for each feature point. The selection circle contains multiple points that are connected, and the feature point is located at the center of multiple points. Connect each point to the feature point, and set multiple sub-points for the feature point; Positioning points are set for both the standard image and the planar image. Connecting lines are established between the positioning points located at the same positions on the standard image and the planar image, respectively. The positioning points located on the standard image and the planar image connected by the connecting lines are used as registration bodies.

[0011] Preferably, the step of determining the anomalous offset of the planar image relative to the standard image based on the registration body includes: Obtain the standard image of the target part and its corresponding reference point for the current process node. Obtain the positioning points of the planar image and the standard image. Based on the connecting line between the positioning points located on the planar image and the standard image respectively, make the two positioning points coincide, thereby aligning the planar image and the standard image. Obtain feature points from the planar image and distribute sub-points to multiple points within the selected circle; Obtain the point position in the planar image corresponding to the reference point as the target point position, and determine the relative displacement between the sub-points and feature points corresponding to the target point position; The feature points corresponding to the relative displacements that do not meet the preset conditions are taken as abnormal feature points. Multiple abnormal feature points are connected to form an abnormal feature point network. The reference points corresponding to the abnormal feature points are connected to form a reference point network. The relative displacement between the abnormal feature point network and the reference point network is taken as the abnormal offset.

[0012] Preferably, the steps of determining the defect type and corresponding control parameters of the target component based on the abnormal offset, adjusting the corresponding process based on the control parameters, and assembling the target component using the assembly equipment of the adjusted process node include: Obtain the abnormal offset of the target part in the current process. Based on the abnormal offset, match and determine the defect type corresponding to the target part in the preset defect type mapping relationship. Based on the defect type, the control parameters corresponding to the defect type are retrieved from the preset control parameter library; The control parameters are sent to the assembly equipment of the corresponding process to adjust the process parameters of the process. The assembly equipment at the adjusted process node is controlled, and the target component is assembled using the control parameters.

[0013] Compared with the prior art, the beneficial effects of the present invention are: Multiple acquisition points collect initial images of target components from different perspectives for corresponding process nodes. The reconstructed positions of feature points in the 3D model are cross-verified based on shared feature points among the initial images, improving the accuracy of feature point location determination. The 3D model across multiple process nodes is dynamically and adaptively updated, eliminating the need for re-analysis and verification of multiple initial images at each process node. Targeted detection and judgment are performed at the corresponding process stage, reducing the data processing load per process stage, effectively lowering equipment inspection load, improving defect detection response speed and real-time performance, and increasing the analysis efficiency of the initial images corresponding to the target components at each process node. This improves the efficiency of defect identification for target components at each process node, allowing defective target components to be eliminated. Simultaneously, control parameters for each process stage are determined based on defect type, thereby improving the quality of subsequent assembly of target components for electric toys. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a system structure block diagram of the present invention. Detailed Implementation

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

[0017] Example: Please refer to Figure 1 This invention provides a technical solution for a multi-station collaborative automated assembly and processing system for electric toys: A multi-station collaborative automated assembly and processing system for electric toys includes: The data acquisition module is used to determine the processes of multiple workstations used for assembling electric toys, and to set process nodes for multiple workstations, and to configure management points and data acquisition point groups for each process node. The steps for determining the processes at multiple workstations used to assemble electric toys, setting process nodes for each workstation, and configuring management points and data collection point groups for each process node include: acquiring and identifying multiple workstations and their process content during the electric toy assembly process; generating corresponding process nodes according to the process content of each workstation in a preset assembly sequence, and assigning node identifiers to each process node; configuring management points and multiple data collection points placed in different locations for each process node, and connecting the communication between the multiple data collection points to obtain a data collection point group. Specifically, the process begins by retrieving a pre-stored workstation configuration table. This table records all workstation information required for assembling the electric toy and the corresponding process steps for each workstation. For example, for assembling an electric children's vehicle, the workstation configuration table would include at least: Workstation 1 (chassis assembly), Workstation 2 (motor installation), Workstation 3 (battery connection), Workstation 4 (wheel assembly), Workstation 5 (shell installation), and Workstation 6 (functional testing). Based on the process steps for each workstation, corresponding process nodes are generated according to a preset assembly sequence, and each process node is assigned a unique node identifier. The preset assembly sequence is determined by the assembly process route of the electric toy, for example, by determining the execution order of each workstation according to the assembly logic from chassis to complete vehicle, or from inside to outside. After associating the process content with the assembly sequence, process nodes are generated, and each process node is assigned a node number (e.g., N001, N002...N00n). Simultaneously, each process node is bound to the corresponding physical location of the workstation for location mapping during subsequent data acquisition point networking. A management point is configured for each process node to store potential defect types during assembly and their corresponding control parameters (each defect type corresponds to a device status information, and each status information corresponds to a control parameter. The control parameter refers to the control parameter that adjusts the device to the state corresponding to the defect type). The data acquisition point group is used to acquire images of the assembled electric toy from different angles, and the acquired images can be cross-validated for corresponding feature points. The consistency of the positional information of feature points acquired from different angles verifies the accuracy of each acquisition point, ensuring the accuracy of the acquired images.

[0018] The defect analysis module is used to collect multiple initial images of the target parts for the corresponding process based on the collection point group, determine the 3D image of the target parts corresponding to each process node based on the multiple initial images, and unfold the 3D image of the target parts corresponding to each process node to obtain the 2D image corresponding to each process node. The 3D model of the target component corresponding to each process node is unfolded to obtain the corresponding planar image of each process node: a cover film is set on the surface of the 3D model, and the cover film is completely attached to the 3D model. Each feature point of the 3D model is marked on the cover film and its corresponding attributes are deployed. After the film is covered, it is opened and laid flat. At the same time, the feature points and corresponding attributes of the 3D model are marked on the flat film at the corresponding positions to obtain the planar image of the corresponding 3D model. A unified unfolding reference plane for the 3D model of electric toy components is determined. The assembly reference plane of the component is used as the unfolding reference plane. A fixed projection unfolding direction and global unfolding scaling ratio are set to unify the unfolding standard of all process 3D models. The complete 3D model of the component generated in a single process is retrieved, and all curved surfaces, folded surfaces, splicing surfaces, and assembly docking surfaces inside the 3D model are separated and divided into independent unfolding planes. In the open unit area, based on the bending angle, assembly overlap relationship, and thickness dimension parameters of the three-dimensional structure, the curved surface, inclined surface, and irregularly shaped mating surface are flattened to eliminate the tilt angle and spatial difference in three-dimensional space. The original assembly hole positions, buckle slots, positioning protrusions, splicing edges, assembly limit edges and all structural outline dimensions and relative positional relationships in the three-dimensional drawing are retained without changing the structural layout spacing. At the same time, the process assembly points, feature marks, difference correction marks, and splicing and fusion traces marked on the three-dimensional drawing are simultaneously mapped and copied to the corresponding positions in the flattened planar image. Each process node completes the three-dimensional drawing planar unfolding according to the same set of benchmark rules, and generates a corresponding process-specific assembly planar image. The newly added structures and correction and adjustment areas of each process assembly are distinguished and marked, forming a full-process continuous planar unfolding atlas that can be directly used for process assembly comparison, station alignment reference, and assembly deviation verification. The steps of acquiring multiple initial images of the target component for the corresponding process based on a set of acquisition points, and determining the stereoscopic image of the target component corresponding to each process node based on the multiple initial images include: determining the feature points of the target component in the current process and marking multiple feature points; acquiring multiple initial images of the target component from different angles based on the set of acquisition points, wherein each initial image contains feature points captured at the corresponding angle; determining the initial stereoscopic image corresponding to each initial image, and superimposing multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process; and determining the stereoscopic image of the target component corresponding to each process node based on the stereoscopic image of the current process. The steps of determining the initial stereoscopic image corresponding to each initial image and overlaying multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process include: restoring each acquisition surface of the target part based on the initial image to obtain multiple local stereoscopic images of the target part; extracting the same feature points in each local stereoscopic image, where the same feature points refer to feature points with the same positional identifier in each local stereoscopic image; stitching the multiple local stereoscopic images based on the same feature points to generate the stereoscopic image of the target part corresponding to the current process; calculating the spatial overlap of the same feature points between each surface; judging whether there are any abnormalities in the feature point data corresponding to each acquisition point based on the comparison result of the spatial overlap with the preset overlap threshold; if it is determined that the feature point data of a certain acquisition point is abnormal, then the acquisition point is marked as an abnormal acquisition point, and the verification result is obtained. Specifically, "image acquisition of the target component from different angles through multiple acquisition points" includes: controlling multiple image acquisition devices to be deployed at different spatial acquisition points; each image acquisition device triggers acquisition commands synchronously according to a preset acquisition time synchronization strategy to obtain multi-angle images of the target component at the same time; "Extracting common feature points from each local stereoscopic image" specifically includes: obtaining feature points with preset geometric identifiers in each local stereoscopic image, including corner points, edge points, marker points, or texture feature points; determining feature points with the same physical location identifiers in each local stereoscopic image to form a correspondence between common feature points in each local stereoscopic image; The calculation of the spatial overlap of the same feature points between the local stereoscopic images specifically includes: selecting any common feature point in each local stereoscopic image as a reference feature point; obtaining the spatial coordinate values ​​of the reference feature point in its respective local stereoscopic image; and calculating the deviation distance or deviation direction between the spatial coordinate values ​​as the spatial overlap. "Based on the comparison results of spatial overlap degree with preset overlap degree threshold, determine whether there are abnormalities in the feature point data corresponding to each collection point" specifically includes: comparing the spatial overlap degree of the same feature points in each local stereoscopic image with the preset overlap degree threshold; if the spatial overlap degree of at least one of the same feature points in a local stereoscopic image is less than the preset overlap degree threshold, then it is determined that the feature point data of the collection point corresponding to that stereoscopic image is abnormal; or, if the spatial overlap degree of more than a preset number of the same feature points in a local stereoscopic image is less than the preset overlap degree threshold, then it is determined that the feature point data of the collection point corresponding to that local stereoscopic image is abnormal. If the feature point data of a certain collection point is determined to be abnormal, then the collection point is marked as an abnormal collection point. The process also includes: correcting the stereoscopic image or feature point data corresponding to the abnormal collection point based on the stereoscopic image feature point data of other normal collection points other than the abnormal collection point; and / or generating an abnormal prompt message, which includes at least the identification information of the abnormal collection point, the location of the abnormal feature point and the type of abnormality. Multiple acquisition devices in the acquisition point group corresponding to the current process are distributed at different angles to the target part, and images of the target part are taken from different angles to acquire image data of the target part at different angles. This image data contains multiple marked feature points on the target part. These marked feature points include not only feature points on the current face captured from the current angle, but also marked feature points on the side captured from the current angle. The images acquired by the acquisition devices are equivalent to the range of the target part projected from the current shooting angle. As long as the feature points within the shooting range are present in the image, and when acquisition devices at different angles acquire images at their respective angles, multiple images will contain the same feature points. In other words, the same feature points are captured by acquisition devices at different angles. The images acquired by each acquisition device are used to reconstruct a 3D image of the target part. The location information of the same feature points in the local 3D images is obtained by obtaining the location information of the same feature points in the local 3D images from different angles. The accuracy of the acquisition of each feature point is judged. When the spatial positions of the same feature points in different local 3D images can coincide, it means that the images of the target part acquired by each acquisition device are accurate. By acquiring the same feature points by multiple acquisition devices at different angles and reconstructing the corresponding spatial positions, the multiple feature points acquired by different acquisition devices are cross-verified, thereby improving the accuracy of feature point acquisition and ensuring the accuracy of identifying defects in the target part.

[0019] The steps for determining the 3D images of target parts corresponding to each process node based on the 3D image of the current process include: acquiring the 3D image of the target part in the current process and its multiple initial images, and transmitting them to the management point corresponding to the next process node as the initial 3D image of the next process node; acquiring the initial image of the target part based on the acquisition point group corresponding to the next process node, comparing the initial image of the next process node with the multiple initial images transmitted from the current process, extracting the initial images with different feature points, and removing the local 3D images of the initial images with different feature points corresponding to the current process in the initial 3D image to obtain valid initial 3D images; restoring the initial images with different feature points corresponding to the next process node into valid local 3D images, and stitching the valid local 3D images onto the valid initial 3D image based on the same feature points to obtain the 3D image of the target part of the next process node; and so on, sequentially acquiring the 3D images of the target parts corresponding to each process node. Specifically, each process node is arranged sequentially according to the assembly timeline. Each process corresponds to only a single category of target parts / assemblies. All process acquisition point groups are pre-calibrated, unifying internal and external parameters, shooting height, light intensity, and shooting angle to eliminate image errors caused by differences in the acquisition environment. The stereoscopic image transmitted in the current process is a 3D point cloud stereoscopic model, and the initial image is a 2D calibration image acquired synchronously from multiple perspectives, corresponding one-to-one with each local area of ​​the stereoscopic image. The data packet transmitted in the current process is bound pairing data, containing a complete initial stereoscopic point cloud model, the corresponding initial image from the perspective, and a mapping table between image coordinates and stereoscopic areas. It is not transmitted independently. After the management point corresponding to the next process node receives it, it establishes a data binding index (specifically: first, the multiple perspective acquisition cameras are globally calibrated, unifying the world coordinate system of all acquisition cameras, determining the shooting position, field of view, and projection matrix corresponding to each initial image, assigning a unique perspective number to each acquisition device, dividing the initial stereoscopic image of the target part into spatial blocks according to each initial image, binding the local stereoscopic images in the initial stereoscopic image to each initial image one-to-one, and adding the corresponding local stereoscopic images to the initial stereoscopic image). The process involves marking and defining each local stereoscopic image, acquiring initial images of the target parts corresponding to the next process node, and restoring them into corresponding local stereoscopic images. The initial images of the next process node are compared with those of the current process node to check for differences in feature points. If differences exist, the initial image of the next process node is restored into a valid local stereoscopic image, replacing the corresponding local stereoscopic image in the initial stereoscopic image. The data binding index primarily divides the initial image according to its corresponding local stereoscopic image, associating and storing the local stereoscopic images and the initial image. The identifier corresponding to the initial image is bound one-to-one to the region code of the corresponding local stereoscopic image. This allows for the retrieval of the index relationship based on the local stereoscopic image indexed by the initial image. After the next process management point imports the index, it automatically matches the correspondence between the initial image and the local stereoscopic image, quickly locking its unique local stereoscopic image by retrieving the initial image (without secondary matching and positioning), ensuring that each initial image can be accurately mapped to a unique local 3D region in the stereoscopic image.The multiple acquisition points in each process node are interconnected. In the first process node, all acquisition points can be cross-verified in their corresponding initial images. In subsequent process nodes, for unchanged feature points, since cross-verification has already been completed in previous process nodes, it is only necessary to verify the feature points in the multiple initial images corresponding to different feature points. That is, only the feature points in the initial images acquired in the corresponding process node that have changed need to be verified. Based on the pixel coordinates and field-of-view parameters of the differing initial images, the 3D point cloud coordinate range of the initial stereo image is mapped back to using the camera calibration matrix, accurately locating abnormal local stereo regions. Only the local point cloud blocks corresponding to the deviation are removed, rather than the entire region being deleted. After each process completes the stereo image update, the final stereo image, the latest matched image, and the difference correction log of this process are automatically archived as the initial reference data for all subsequent processes, overwriting and replacing the original data of the previous process. All processes in the production line... The nodes repeatedly execute a closed-loop logic of "data reception, image acquisition, difference comparison, invalid region removal, reconstruction and stitching update" until all assembly processes are completed, ultimately generating a dynamic iterative 3D model adapted to each assembly step. The 3D images across multiple process nodes are dynamically and adaptively updated, eliminating the need to re-analyze and verify multiple initial images at each process node. This improves the efficiency of analyzing the initial images corresponding to the target parts at each process node, thereby increasing the efficiency of defect identification for the target parts at each process node and removing defective target parts. Simultaneously, the control parameters for each process are determined based on the defect type, improving the quality of subsequent assembly of target parts for electric toys. This achieves dynamic adaptive updating of the 3D vision model between processes, ensuring visual positioning accuracy in multi-process collaborative assembly. It eliminates the need for full-domain defect detection of the entire electric toy structure, performing targeted detection and judgment only at the corresponding process. This significantly reduces the data computation load per process, effectively lowering the equipment's detection load and improving defect detection response speed and real-time performance. Abnormal workpieces can be directly identified and removed at the current process where the defect occurs, preventing defective workpieces from flowing into subsequent assembly processes and avoiding the continuous accumulation of defects. It can also accurately match the corresponding workstation and optimize and adjust the equipment operating parameters in real time based on the actual type of assembly defects detected, effectively avoiding the recurrence of similar assembly abnormalities. This not only steadily improves the overall assembly production efficiency of electric toys but also effectively improves product assembly quality and significantly increases the yield rate of finished products.

[0020] The toy assembly module is used to acquire standard images corresponding to each process node, set a registration body between the planar image and the standard image, determine the abnormal offset of the planar image relative to the standard image based on the registration body, determine the defect type and corresponding control parameters of the target part based on the abnormal offset, adjust the corresponding process based on the control parameters, and assemble the target part using the assembly equipment of the adjusted process node. The steps of acquiring standard images corresponding to each process node and setting a registration body between the planar image and the standard image include: storing the standard image of the target part corresponding to the process node in each management point, and acquiring feature points in the standard image as reference points; acquiring the planar image corresponding to each process node and multiple feature points in the planar image, configuring a selection circle for each feature point, wherein multiple points with connections are laid in the selection circle, and the feature point is located at the center of multiple points; connecting each point with the feature point, and setting multiple sub-points for the feature point; setting positioning points for both the standard image and the planar image, establishing connecting lines between positioning points located at the same position on the standard image and the planar image respectively, and using the positioning points located on the standard image and the planar image respectively connected by the connecting lines as the registration body; The steps for determining the abnormal offset of a planar image relative to a standard image based on a registration body include: acquiring the standard image of the target component of the current process node and its corresponding reference point; acquiring the positioning points of the planar image and the standard image; aligning the two positioning points by aligning them with the connecting line between the positioning points located on the planar image and the standard image; acquiring feature points in the planar image and distributing them to multiple points in the selection circle; acquiring the point in the planar image corresponding to the reference point as the target point and determining the relative displacement between the corresponding point and the feature point; identifying feature points whose relative displacement does not meet preset conditions as abnormal feature points; connecting multiple abnormal feature points to obtain an abnormal feature point network; connecting the reference points corresponding to the abnormal feature points to obtain a reference point network; and using the relative displacement between the abnormal feature point network and the reference point network as the abnormal offset. It should be noted that relative displacement refers to the offset between a feature point and its corresponding reference point. The abnormal offset is obtained by connecting multiple abnormal feature points to form a graph corresponding to the abnormal feature points (i.e., an abnormal feature point network). Then, the reference points corresponding to multiple abnormal feature points are connected to form a graph (i.e., a reference point network, which is composed of abnormal feature points, not all reference points). The reference point network and the abnormal feature point network are compared, and the reference points correspond one-to-one with the abnormal feature points. The relative displacement between each abnormal feature point and its corresponding reference point is an offset. Multiple relative displacements between multiple abnormal feature points and their corresponding reference points form an offset graph, which is the corresponding abnormal offset. Specifically, firstly, two positioning points located in the same position in different images (one in a planar image and the other in a standard image) are overlapped using connecting lines. After overlap, the planar image and the standard image are essentially aligned. In the planar image, a selection circle is configured for each feature point, which is equivalent to the offset range corresponding to the feature point (the position of the feature point will change when there are abnormalities or defects in the assembly). Multiple sub-points are configured for each feature point, and the sub-points are distributed across various points. The sub-points are connected to the feature points, and the sub-points are like virtual machines used to find the position of the reference point among multiple points. The point that coincides with the reference point in the planar image is determined as the target point, and the sub-points at the target point are taken as target sub-points. The relative displacement between the target sub-point and the feature point is taken as the offset of a feature point. By distributing multiple sub-points across different points, the target point is determined by analyzing the position of the reference point through multiple sub-points. Multiple sub-points analyze each point simultaneously, thereby enabling... It can quickly determine the position information of the reference point within the offset range corresponding to the feature point, and connect the feature point and the reference point through the target sub-point to quickly determine the offset of each feature point. After determination, it connects multiple abnormal feature points to obtain the abnormal offset of the entire target part's corresponding planar image. Based on the feature points with offsets and their offset information (including offset position and offset direction), it determines the defects of the target part in the assembly process, and calls the corresponding control parameters to adjust the assembly equipment of the corresponding process node according to the corresponding defects. This enables real-time identification of assembly defects of electric toys in each assembly process, quick location of defective workpieces, and automatic rejection of defective products. At the same time, it classifies and statistically analyzes the detected defects, accurately determines the defect type and cause, and reversely adjusts the assembly equipment parameters and operation actions of the corresponding process based on the defect data. From the equipment end, it avoids the recurrence of similar assembly abnormalities, effectively reduces the defect rate of electric toy assembly, and improves the batch assembly yield and production stability.

[0021] The steps of determining the defect type and corresponding control parameters of the target component based on the abnormal offset, adjusting the corresponding process based on the control parameters, and assembling the target component using the assembly equipment at the adjusted process node include: obtaining the abnormal offset of the target component in the current process; determining the corresponding defect type of the target component based on the abnormal offset by matching it in a preset defect type mapping relationship; calling the control parameters corresponding to the defect type from a preset control parameter library based on the defect type; sending the control parameters to the assembly equipment at the corresponding process to adjust the process parameters; and controlling the assembly equipment at the adjusted process node to perform the assembly operation on the target component using the control parameters. Specifically, the feature points of the target component are compared with the reference points to determine the deviation values ​​between each feature point and the reference points, which are used as abnormal offsets. The offset scalars of the abnormal offsets in at least two directions are obtained. The offset scalars in each direction are compared with preset defect offset thresholds. Based on the comparison results of the offset scalars in each direction and the defect offset thresholds, the defect type is matched and determined. The management point stores the defect type mapping relationship, which includes the correspondence between the threshold range of the abnormal offset and the defect type. When an abnormal offset is determined, it is transmitted to the management point. The management point determines the corresponding control parameters based on the abnormal offset and writes the control parameters into the executable instructions of the assembly equipment corresponding to the process node. The assembly equipment performs the assembly action according to the control parameters in the executable instructions to complete the assembly of the target component.

[0022] Multiple process nodes are set up for the assembly process, and multiple acquisition devices are deployed at each process node. Images of the target component corresponding to the current process are acquired from different angles by multiple acquisition devices. The feature points acquired by other acquisition devices are cross-verified based on the images acquired from different angles. At the same time, the 3D image of the target component corresponding to the current process is unfolded, and the position information of each feature point is determined in the unfolded image. When the position information of the feature point is abnormal, the corresponding defect type is determined based on the abnormal feature point, and the control parameters of the current process are determined based on the defect type. Different unfolded images are corresponding to different processes, and the unfolded images are layered and progressively increased. There is no need to reprocess the previous data at each process, which improves the efficiency of defect type detection and thus improves the assembly efficiency. A multi-station hierarchical collaborative architecture is adopted. Each process station is independently managed and operates independently, while the processes are superimposed and coordinated. Defect detection and abnormal workpiece rejection can be completed on-site in the current process without having to traverse and inspect the entire structure of the electric toy. This greatly reduces the amount of data processing and computing load in a single process and improves the real-time performance of process detection and production cycle. By independently identifying assembly defects at each stage and removing defective parts on-site, the system can prevent the accumulation of defects at the source, avoiding the introduction of flaws from previous assembly processes. Simultaneously, based on the type of defect generated in the current process, the system can adjust the assembly control parameters of the corresponding workstation equipment in real time, achieving adaptive correction and control at the workstation level, effectively reducing the recurrence of similar assembly anomalies. Through a multi-workstation independent operation and hierarchical collaborative quality control model, the system significantly reduces the incidence of assembly defects in electric toys, improving multi-workstation collaborative assembly accuracy, production stability, and finished product yield.

[0023] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0024] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-station collaborative automated assembly and processing system for electric toys, characterized in that, include: The data acquisition module is used to determine the processes of multiple workstations used for assembling electric toys, and to set process nodes for multiple workstations, and to configure management points and data acquisition point groups for each process node. The defect analysis module is used to collect multiple initial images of the target parts for the corresponding process based on the collection point group, determine the 3D image of the target parts corresponding to each process node based on the multiple initial images, and unfold the 3D image of the target parts corresponding to each process node to obtain the 2D image corresponding to each process node. The toy assembly module is used to acquire standard images corresponding to each process node, set a registration body between the planar image and the standard image, determine the abnormal offset of the planar image relative to the standard image based on the registration body, determine the defect type and corresponding control parameters of the target part based on the abnormal offset, adjust the corresponding process based on the control parameters, and assemble the target part using the assembly equipment of the adjusted process node.

2. The multi-station collaborative automated assembly and processing system for electric toys according to claim 1, characterized in that: The steps of determining the processes for assembling electric toys at multiple workstations, setting process nodes for each workstation, and configuring management points and data collection point groups for each process node include: Acquire and identify multiple workstations in the assembly process of electric toys and the process content of each workstation; According to the process content of each workstation, corresponding process nodes are generated in a preset assembly sequence, and a node identifier is assigned to each process node. Each process node is configured with a management point and multiple data collection points placed in different locations. The communication between the multiple data collection points is then connected to form a data collection point group.

3. The multi-station collaborative automated assembly and processing system for electric toys according to claim 1, characterized in that: The step of acquiring multiple initial images of the target component for the corresponding process based on the acquisition point group, and determining the stereoscopic image of the target component corresponding to each process node based on the multiple initial images includes: Identify the feature points of the target part in the current process and mark multiple feature points; Multiple initial images of the target accessory are acquired from different angles based on the collection point group, wherein each initial image contains feature points captured from the corresponding angle; Determine the initial stereoscopic image corresponding to each initial image, and overlap multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process; Based on the 3D diagram of the current process, determine the 3D diagram of the target component corresponding to each process node.

4. The multi-station collaborative automated assembly and processing system for electric toys according to claim 3, characterized in that: The step of determining the initial stereoscopic image corresponding to each initial image and superimposing multiple initial stereoscopic images to obtain the stereoscopic image corresponding to the current process includes: Based on the initial image, the various acquisition surfaces of the target part are restored to obtain multiple local stereoscopic images of the target part. The same feature points in each local stereoscopic image are extracted. The same feature points refer to feature points that have the same positional identification in each local stereoscopic image. Based on the same feature points, the multiple local stereoscopic images are stitched together to generate a stereoscopic image of the target part corresponding to the current process. Calculate the spatial overlap of the same feature points between each surface; based on the comparison between the spatial overlap and the preset overlap threshold, determine whether there are any anomalies in the feature point data corresponding to each collection point; If it is determined that the feature point data of a certain collection point is abnormal, then the collection point is marked as an abnormal collection point, and the verification result is obtained.

5. The multi-station collaborative automated assembly and processing system for electric toys according to claim 3, characterized in that: The step of determining the 3D view of the target component corresponding to each process node based on the 3D view of the current process includes: Obtain the 3D image of the target part in the current process and its multiple initial images, and transmit them to the management point corresponding to the next process node as the initial 3D image of the next process node; The initial image of the target part is collected based on the collection point group corresponding to the next process node. The initial image of the next process node is compared with multiple initial images transmitted in the current process. The initial images with different feature points are extracted. The local stereo images of the initial stereo images corresponding to the current process with different feature points are removed from the initial stereo images in the initial stereo images to obtain the effective initial stereo images. The initial image with different feature points corresponding to the next process node is restored into a valid local stereoscopic image. Based on the same feature points, the valid local stereoscopic images are stitched onto the valid initial stereoscopic image to obtain the stereoscopic image of the target part of the next process node. This process is repeated to obtain the stereoscopic images of the target parts corresponding to each process node.

6. The multi-station collaborative automated assembly and processing system for electric toys according to claim 1, characterized in that: The step of obtaining the standard image corresponding to each process node and setting a registration body between the planar image and the standard image includes: Store the standard image of the target part corresponding to the process node in each management point, and obtain the feature points in the standard image as reference points; Obtain the planar images corresponding to each process node and multiple feature points in the planar images. Configure a selection circle for each feature point. The selection circle contains multiple points that are connected, and the feature point is located at the center of multiple points. Connect each point to the feature point, and set multiple sub-points for the feature point; Positioning points are set for both the standard image and the planar image. Connecting lines are established between the positioning points located at the same positions on the standard image and the planar image, respectively. The positioning points located on the standard image and the planar image connected by the connecting lines are used as registration bodies.

7. The multi-station collaborative automated assembly and processing system for electric toys according to claim 6, characterized in that: The step of determining the anomalous offset of the planar image relative to the standard image based on the registration body includes: Obtain the standard image of the target part and its corresponding reference point for the current process node. Obtain the positioning points of the planar image and the standard image. Based on the connecting line between the positioning points located on the planar image and the standard image respectively, make the two positioning points coincide, thereby aligning the planar image and the standard image. Obtain feature points from the planar image and distribute sub-points to multiple points within the selected circle; Obtain the point position in the planar image corresponding to the reference point as the target point position, and determine the relative displacement between the sub-points and feature points corresponding to the target point position; The feature points corresponding to the relative displacements that do not meet the preset conditions are taken as abnormal feature points. Multiple abnormal feature points are connected to form an abnormal feature point network. The reference points corresponding to the abnormal feature points are connected to form a reference point network. The relative displacement between the abnormal feature point network and the reference point network is taken as the abnormal offset.

8. The multi-station collaborative automated assembly and processing system for electric toys according to claim 1, characterized in that: The steps of determining the defect type and corresponding control parameters of the target component based on the abnormal offset, adjusting the corresponding process based on the control parameters, and assembling the target component using the assembly equipment of the adjusted process node include: Obtain the abnormal offset of the target part in the current process. Based on the abnormal offset, match and determine the defect type corresponding to the target part in the preset defect type mapping relationship. Based on the defect type, the control parameters corresponding to the defect type are retrieved from the preset control parameter library; The control parameters are sent to the assembly equipment of the corresponding process to adjust the process parameters of the process. The assembly equipment at the adjusted process node is controlled, and the target component is assembled using the control parameters.

Images