Method and system for detecting defects in a mechanical arm fitting after casting

Abstract

The application discloses a kind of mechanical arm accessory after casting defect detection method and system, and the application relates to the technical field of visual inspection method, according to the current position of each abnormal area and the current posture of mechanical arm accessory Determine defect area and visual error area, the overall control of the casting process and detection process of mechanical arm accessory, and compatible machine vision head targeted control.Based on visual error area and multiple side images trigger autonomous regulation and control of machine vision head, to gradually avoid visual error area;Based on the defect type corresponding to defect area and corresponding surface determine multiple to be traced process, and determine the optimized casting process table of mechanical arm accessory according to the previous data corresponding to multiple to be traced process and defect area, realize the optimization of the casting process and detection process of mechanical arm accessory, and ensure the accuracy of defect detection of mechanical arm accessory after casting, avoid the subsequent influence of visual error area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of visual inspection method, and particularly relates to a mechanical arm accessory post-casting defect detection method and system. BACKGROUND

[0002] With the development of science and technology, a mechanical arm accessory is formed after a casting process and subsequent processing. The mechanical arm accessory is a key component of a mechanical arm and needs high surface quality. In the prior art, a camera is used to detect the surface of the mechanical arm accessory, and defects of the mechanical arm accessory are determined according to images collected by the camera. However, the shooting parameters of the camera are not considered, and the casting process and detection process of the mechanical arm accessory cannot be controlled as a whole. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art, and provides a mechanical arm accessory post-casting defect detection method and system.

[0004] The present application provides a mechanical arm accessory post-casting defect detection method, which comprises the following steps: collecting multiple side images of a mechanical arm accessory based on a machine vision head; determining a three-dimensional model of the mechanical arm accessory according to the multiple side images and scanning data corresponding to the mechanical arm accessory, and marking abnormal areas of the three-dimensional model; determining defect areas and visual error areas according to current positions of the abnormal areas and a current posture of the mechanical arm accessory; triggering autonomous regulation and control of the machine vision head based on the visual error areas and the multiple side images, so as to gradually avoid the visual error areas; determining multiple to-be-traced processes based on defect types corresponding to the defect areas and corresponding surfaces, and determining a casting process table of an optimized mechanical arm accessory according to previous data corresponding to the multiple to-be-traced processes and the defect areas.

[0005] The present application provides a mechanical arm accessory post-casting defect detection system, which is applied to the mechanical arm accessory post-casting defect detection method described above. The mechanical arm accessory post-casting defect detection system comprises the following parts:

[0006] A side image module is configured to collect multiple side images of a mechanical arm accessory based on a machine vision head.

[0007] An abnormal area module is configured to determine a three-dimensional model of the mechanical arm accessory according to the multiple side images and scanning data corresponding to the mechanical arm accessory, and mark abnormal areas of the three-dimensional model.

[0008] A multiple area module is configured to determine defect areas and visual error areas according to current positions of the abnormal areas and a current posture of the mechanical arm accessory.

[0009] An autonomous regulation module is configured to trigger autonomous regulation of the machine vision head based on the visual error area and the plurality of side images to gradually avoid the visual error area.

[0010] A casting process table module is configured to determine a plurality of to-be-traced processes based on the defect type corresponding to the defect area and the corresponding surface, and determine an optimized casting process table of the mechanical arm accessory according to the previous data corresponding to the plurality of to-be-traced processes and the defect area.

[0011] Compared with the prior art, the present application has the following advantages:

[0012] In the embodiment of the present application, the method in the embodiment of the present application is used to collect a plurality of side images of the mechanical arm accessory based on the machine vision head; a three-dimensional model of the mechanical arm accessory is determined according to the plurality of side images and the scanning data corresponding to the mechanical arm accessory, and an abnormal area of the three-dimensional model is marked; and a defect area and a visual error area are determined according to the current position of each abnormal area and the current posture of the mechanical arm accessory. The introduction of the defect area and the visual error area enables the overall control of the casting process and the detection process of the mechanical arm accessory, and is compatible with the targeted control of the machine vision head.

[0013] Therefore, the autonomous regulation of the machine vision head is triggered based on the visual error area and the plurality of side images to gradually avoid the visual error area; the plurality of to-be-traced processes are determined based on the defect type corresponding to the defect area and the corresponding surface, and the optimized casting process table of the mechanical arm accessory is determined according to the previous data corresponding to the plurality of to-be-traced processes and the defect area, thereby realizing the optimization of the casting process and the detection process of the mechanical arm accessory, ensuring the accuracy of the defect detection of the mechanical arm accessory after casting, and avoiding the subsequent influence of the visual error area. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is a flowchart of the defect detection method of the mechanical arm accessory after casting in the embodiment of the present application;

[0015] Figure 2 is a flowchart of step S11 in the defect detection method of the mechanical arm accessory after casting in the embodiment of the present application;

[0016] Figure 3 is a flowchart of step S12 in the defect detection method of the mechanical arm accessory after casting in the embodiment of the present application;

[0017] Figure 4 is a flowchart of step S13 in the defect detection method of the mechanical arm accessory after casting in the embodiment of the present application;

[0018] Figure 5is a flowchart of step S14 in the defect detection method for the mechanical arm accessory after casting in the embodiment of the present application;

[0019] Figure 6 is a flowchart of step S15 in the defect detection method for the mechanical arm accessory after casting in the embodiment of the present application;

[0020] Figure 7 is a structural composition schematic diagram of the defect detection system for the mechanical arm accessory after casting in the embodiment of the present application. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0022] Please refer to Figures 1 to 7 A defect detection method for a mechanical arm accessory after casting, the mechanical arm accessory comprising a mechanical arm shell, a mechanical arm connecting seat, a rotating part, etc.; the defect detection method for the mechanical arm accessory after casting comprises:

[0023] Step S11: acquiring multiple side surface images of the mechanical arm accessory based on a machine vision head;

[0024] Step S12: determining a three-dimensional model of the mechanical arm accessory according to the multiple side surface images and scanning data corresponding to the mechanical arm accessory, and marking abnormal areas of the three-dimensional model;

[0025] Step S13: determining defect areas and visual error areas according to current positions of the abnormal areas and a current posture of the mechanical arm accessory;

[0026] Step S14: triggering autonomous regulation and control of the machine vision head based on the visual error areas and the multiple side surface images, so as to gradually avoid the visual error areas;

[0027] Step S15: determining multiple to-be-traced processes based on defect categories corresponding to the defect areas and corresponding surfaces, and determining an optimized casting process table of the mechanical arm accessory according to previous data corresponding to the multiple to-be-traced processes and the defect areas;

[0028] Reference Figure 2 In step S11, multiple side surface images of the mechanical arm accessory are acquired based on a machine vision head;

[0029] In the specific implementation process of the present application, the specific steps are as follows:

[0030] S111: after the mechanical arm accessory is cast, a shooting space is determined according to a current position of the mechanical arm accessory and a form of the mechanical arm accessory;

[0031] S112: Trigger multiple machine vision heads around the shooting space according to the shooting space, wherein the multiple machine vision heads are arranged along the shooting space and shoot the mechanical arm accessory in different directions;

[0032] S113: Collect multiple images based on the shooting of the mechanical arm accessory by the multiple machine vision heads, determine multiple side surface images according to the synthesis of the multiple images, and present the state of multiple side surfaces in the mechanical arm accessory according to the multiple side surface images.

[0033] In the embodiments of the present application, after the mechanical arm accessory is cast, the shooting space is determined according to the current position of the mechanical arm accessory and the form of the mechanical arm accessory, which comprehensively considers the current position of the mechanical arm accessory and the form of the mechanical arm accessory, and guarantees the accuracy of the shooting space.

[0034] At this time, after the mechanical arm accessory is cast, the current position of the mechanical arm accessory after casting is determined using a position sensor (such as a laser range finder, an RFID tag reader, etc.) or machine vision technology (such as template matching, feature point detection, etc.); the current position usually includes the coordinates (X, Y, Z) and attitude information (such as the rotation angle) of the mechanical arm accessory in three-dimensional space. At the same time, the form of the mechanical arm accessory is obtained through the pre-known drawing; the form includes the size, shape, curved surface change, etc. of the mechanical arm accessory.

[0035] Further, combined with the current position and form information of the mechanical arm accessory, a suitable shooting space is calculated using a geometric algorithm or simulation software; the shooting space should be able to completely contain the mechanical arm accessory, and take into account the shooting error, the slight movement of the mechanical arm accessory and the field of view of the machine vision head; the shooting space is usually represented as a three-dimensional cube or a more complex geometric shape, the size and direction of which are adjusted according to the specific form and position of the mechanical arm accessory; optionally, a suitable shooting space is calculated according to the length and diameter of the mechanical arm accessory and its position on the workbench; assuming that the shooting space is a cuboid, its length is 1.2 meters (considering leaving a certain margin at both ends), its width is 0.6 meters (considering the width and rotation of the mechanical arm accessory), and its height is 0.8 meters (considering the height of the mechanical arm accessory and the field of view of the machine vision head).

[0036] Before actual shooting, trial shooting is performed using the machine vision head to observe whether each side surface of the mechanical arm accessory can be completely shot; if it is found that the field of view of the machine vision head is insufficient, the size of the shooting space or the arrangement position of the machine vision head is appropriately adjusted. Through the above steps, a suitable shooting space is determined so that the subsequent machine vision head can accurately shoot each side surface of the mechanical arm accessory, providing reliable data support for subsequent defect detection and process optimization.

[0037] Further, according to the shooting space, a plurality of machine vision heads are triggered, at this time, the plurality of machine vision heads are arranged along the shooting space and shoot the mechanical arm accessory in different directions, realizing the shooting of the mechanical arm accessory in multiple directions.

[0038] At this time, according to the size, shape of the shooting space and the shape of the mechanical arm accessory, the number of machine vision heads required is determined; the machine vision heads should be arranged along the edge or key position of the shooting space to ensure that each side of the mechanical arm accessory can be covered; considering the field of view, overlapping area and shooting angle of the machine vision head, the optimal arrangement position is determined.

[0039] According to the surface material, color and lighting conditions of the mechanical arm accessory, the shooting parameters such as exposure time, gain and white balance of the machine vision head are adjusted; the resolution and frame rate of the machine vision head are ensured to meet the shooting requirements; if necessary, the triggering mode (such as software triggering, hardware triggering) and shooting mode (such as continuous shooting, single shooting) of the machine vision head are also configured.

[0040] When the arrangement position of the shooting space and the machine vision head is determined, the control system sends a trigger signal to the machine vision head; after receiving the trigger signal, the machine vision head starts shooting according to the preset parameters and mode; the control system records the shooting time and state of each machine vision head for subsequent data processing and analysis.

[0041] During the shooting process, the shooting picture and state information of the machine vision head are viewed in real time through the monitoring system; if the shooting quality is found to be poor or the position of the mechanical arm accessory changes, the parameters or arrangement position of the machine vision head are adjusted in time; it is ensured that each machine vision head can shoot the relevant side of the mechanical arm accessory as expected.

[0042] Optionally, assuming that there is a shooting space with a cuboid shape, the size is 2m x 1m x 1m; the mechanical arm accessory is located in the center of the shooting space, and the shape is a complex metal structural part; according to the size of the shooting space and the shape of the mechanical arm accessory, it is decided to use 4 machine vision heads for shooting; the machine vision heads are arranged at the four corners of the shooting space to ensure that each side of the mechanical arm accessory can be covered; considering the field of view and overlapping area of the machine vision head, the specific position and angle of the machine vision head are adjusted.

[0043] Since the surface of the mechanical arm accessory is metal material with strong reflectivity, the exposure time and gain of the machine vision head are adjusted to reduce the reflection and overexposure phenomenon; at the same time, the white balance of the machine vision head is configured to ensure the color accuracy of the shooting picture; the machine vision head is set to single shooting mode and configured with software triggering mode.

[0044] When the mechanical arm accessory is placed in the center of the shooting space, the control system sends a trigger signal to the four machine vision heads; after receiving the trigger signal, the machine vision heads start shooting according to the preset parameters; the control system records the shooting time and state information of each machine vision head.

[0045] During the shooting process, the monitoring system real-time views the shooting pictures of the four machine vision heads; finds that the shooting picture of one of the machine vision heads has overexposure phenomenon, and adjusts the exposure time of the machine vision head in time; after ensuring that each machine vision head can shoot the relevant side of the mechanical arm accessory as expected, the shooting process is ended; through the above steps, multiple machine vision heads are arranged along the shooting space, and the mechanical arm accessory is shot in different directions, which provides high-quality data support for subsequent image processing and defect detection.

[0046] Therefore, based on the shooting of the mechanical arm accessory by multiple machine vision heads, multiple images are collected, multiple side images are determined according to the synthesis of the multiple images, and the state of multiple sides in the mechanical arm accessory is presented according to the multiple side images, and multiple side images are introduced.

[0047] At this time, when the multiple machine vision heads shoot the mechanical arm accessory according to the preset position and parameters, a series of images will be generated, which contain different sides, different angles of views of the mechanical arm accessory, and background information; it is necessary to ensure that each machine vision head can collect clear and non-blurry images for subsequent image processing and synthesis. Optionally, before image synthesis, the images collected by each machine vision head need to be preprocessed; the preprocessing steps include image denoising, image enhancement (such as contrast adjustment, sharpening, etc.), and image cropping (removing unnecessary background information); the purpose of preprocessing is to improve image quality and reduce the difficulty of subsequent image synthesis and processing.

[0048] Image registration is the process of aligning the same feature points in multiple images to ensure that they can be seamlessly spliced when synthesized; feature point matching algorithms (such as SIFT, SURF, etc.) are used to find the same feature points in different images, and the transformation relationship between images is calculated through these feature points; once the images are correctly registered, image synthesis techniques (such as image stitching, image fusion, etc.) are used to generate a complete synthesized image containing all sides of the mechanical arm accessory.

[0049] After the synthesis image is generated, it is further analyzed and processed to present the state of multiple sides of the mechanical arm accessory, which includes using image processing algorithms to detect defects on the surface of the mechanical arm accessory, measure dimensions, analyze shapes, etc.; finally, these analysis results are presented in a visual way, such as generating a three-dimensional model, generating a report, etc.

[0050] Specifically, assuming that there are 4 machine vision heads, which have captured 4 different sides of the robotic arm accessory; the 4 machine vision heads have collected images of the top, bottom, left side and right side of the robotic arm accessory, which contain different views of the robotic arm accessory and background information of the workbench; each image has been denoised to reduce noise and interference in the image; the contrast of the image has been enhanced to make the outline of the robotic arm accessory clearer; the image has been cropped to remove unnecessary background information and only keep part of the robotic arm accessory.

[0051] The SIFT algorithm is used to find feature points in each image, and the transformation relationship between different images is calculated; according to these transformation relationships, the four images are registered and aligned; image stitching technology is used to combine the four images into a complete composite image containing all sides of the robotic arm accessory.

[0052] The composite image is further analyzed, and image processing algorithms are used to detect defects on the surface of the robotic arm accessory (such as cracks, scratches, etc.); the size of the robotic arm accessory is measured to verify whether it meets the design requirements; the shape of the robotic arm accessory is analyzed to ensure that it meets the expected geometric characteristics; finally, the analysis results are presented in the form of a three-dimensional model, and a detailed inspection report is generated; through the above steps, the images collected by multiple machine vision heads are used to synthesize a complete robotic arm accessory side image, and the status of multiple sides is presented, providing reliable data support for subsequent defect detection and process optimization.

[0053] Reference Figure 3 In step S12, a three-dimensional model of the robotic arm accessory is determined according to the plurality of side images and the scanning data corresponding to the robotic arm accessory, and an abnormal area of the three-dimensional model is marked;

[0054] In the specific implementation process of the present application, the specific steps are as follows:

[0055] S121: based on the scanning of the scanning device on the robotic arm accessory, scanning data is determined, which is presented in the form of point cloud; the outer contour of the robotic arm accessory is determined according to the synthesis of the scanning data;

[0056] S122: a three-dimensional model of the robotic arm accessory is determined according to the synthesis of the plurality of side images and the outer contour of the robotic arm accessory, at this time, the plurality of side images and the outer contour of the robotic arm accessory are synthesized under the same position dimension, and the corresponding three-dimensional model is gradually constructed along the outer contour of the robotic arm accessory;

[0057] S123: based on the traversal of the three-dimensional model of the robotic arm accessory, a plurality of abnormal positions are determined, and according to the plurality of abnormal positions and the corresponding side, the abnormal area of the three-dimensional model is determined to mark the abnormal area of the three-dimensional model.

[0058] In the embodiments of the present application, the scanning data is determined based on the scanning of the mechanical arm accessory by the scanning device, and the scanning data is presented in the form of a point cloud; the outer contour of the mechanical arm accessory is determined according to the synthesis of the scanning data, thereby ensuring the outer contour of the mechanical arm accessory.

[0059] At this time, a three-dimensional scanning device suitable for scanning the mechanical arm accessory is selected, such as a laser scanner, a structured light scanner, or a stereo vision scanner; according to the size, shape, and material of the mechanical arm accessory, the parameters of the scanning device are adjusted, such as scanning speed, resolution, light source intensity, etc.; the relative position between the scanning device and the scanned mechanical arm accessory is ensured to be stable to avoid motion artifacts in the scanning process.

[0060] The scanning device is started to begin scanning the mechanical arm accessory; during the scanning process, the device emits light or patterns to the surface of the mechanical arm accessory and captures the reflected information, which is converted into a series of three-dimensional coordinate points, i.e., point cloud data; according to the complexity of the mechanical arm accessory, scanning is required from multiple angles to ensure complete surface information is obtained.

[0061] The point cloud data obtained by scanning is preprocessed, including denoising, filtering, registration, etc.; denoising is to remove noise points generated during scanning, which are caused by device errors, environmental interference, etc.; filtering is to smooth the point cloud data and reduce high-frequency fluctuations in the data to improve data accuracy; registration is the process of merging point cloud data obtained from different angles into a whole, which requires finding overlapping areas between different point clouds and aligning them through algorithms.

[0062] The point cloud data after preprocessing is a complete and accurate three-dimensional representation of the mechanical arm accessory; algorithms are used to extract the outer contour of the mechanical arm accessory from the point cloud data, which usually involves steps such as boundary detection and surface fitting of the point cloud data; the outer contour is the boundary representation of the mechanical arm accessory in three-dimensional space, which reflects the shape and size information of the accessory.

[0063] Specifically, suppose there is a complex mechanical arm accessory that needs to be three-dimensionally reconstructed for quality detection; a high-precision laser scanner is selected, which can capture three-dimensional information of the mechanical arm accessory at high speed and high resolution; according to the size and shape of the mechanical arm accessory, the scanning speed and resolution of the scanner are adjusted to ensure that all details of the accessory can be captured.

[0064] Place the robotic arm accessory on the scanner's worktable and start the scanner to begin scanning; the scanner emits a laser beam onto the surface of the robotic arm accessory and captures the reflected laser information; during the scanning process, the robotic arm accessory is scanned from multiple angles to ensure that the complete surface information of the accessory is obtained.

[0065] After the scanning is complete, point cloud data containing millions of points is obtained; special software is used to denoise and filter the point cloud data, removing noise points and high-frequency fluctuations; then, a registration algorithm is used to merge the point cloud data obtained from different angles into a whole.

[0066] The preprocessed point cloud data is already a complete three-dimensional representation of the robotic arm accessory; a boundary detection algorithm is used to extract the outer contour of the robotic arm accessory from the point cloud data; the outer contour clearly shows the shape and size information of the robotic arm accessory, providing a foundation for subsequent quality detection and analysis; through the S121 step, the point cloud data of the robotic arm accessory is successfully obtained from the scanning device, and its outer contour is determined, which provides important basic data for the subsequent three-dimensional reconstruction and anomaly detection steps.

[0067] Further, the three-dimensional model of the robotic arm accessory is determined according to the synthesis of the plurality of side images and the outer contour of the robotic arm accessory, at this time, the plurality of side images and the outer contour of the robotic arm accessory are synthesized under the same position dimension, and the corresponding three-dimensional model is gradually constructed along the outer contour of the robotic arm accessory, ensuring the accuracy of the three-dimensional model.

[0068] At this time, it is ensured that the plurality of side images and the outer contour of the robotic arm accessory have been obtained; the side images should contain views of the robotic arm accessory at different angles to cover all key surface features; the outer contour should be obtained from a three-dimensional scanning device and accurately represent the three-dimensional boundary of the robotic arm accessory; image registration or point cloud registration techniques are used to align the side images and the outer contour to the same position dimension, which usually involves steps such as feature point matching and transformation matrix calculation. Optionally, the robotic arm accessory is the outer shell of the robotic arm, the cover of the robotic arm, or the surface decoration of the robotic arm.

[0069] Optionally, assuming there is a robotic arm accessory that needs to be three-dimensionally reconstructed for virtual simulation and display; four side images of the robotic arm accessory have been obtained, as well as an outer contour obtained from a three-dimensional scanning device; image registration techniques are used to align the side images and the outer contour to the same position dimension, which involves finding feature points in the images and corresponding points in the outer contour, and calculating a transformation matrix to align them.

[0070] Texture mapping is the process of mapping two-dimensional image information (such as color, texture) onto a three-dimensional model surface; for each side image, determine its correspondence with the mechanical arm accessory outer contour; use projection, unfolding or other algorithms to map the texture information in the side image onto the three-dimensional surface represented by the outer contour, this step needs to deal with image distortion, perspective effect, etc. to ensure the accuracy of texture mapping.

[0071] Optionally, for each side image, its correspondence with the mechanical arm accessory outer contour is determined; using projection algorithm, the texture information in the side image is mapped onto the three-dimensional surface represented by the outer contour; in the mapping process, image distortion and perspective effect are handled to ensure the accuracy of texture mapping.

[0072] Along the outer contour of the mechanical arm accessory, its three-dimensional model is gradually constructed, which usually involves extracting geometric information (such as vertices, edges, faces) from the outer contour and using these information to construct a three-dimensional mesh; in the process of constructing the mesh, it is necessary to ensure that the topological structure of the mesh is correct, that is, the connection relationship between vertices, edges and faces conforms to the actual shape of the mechanical arm accessory; at the same time, texture information is applied to the mesh to generate a three-dimensional model with realistic feeling.

[0073] Optionally, along the outer contour of the mechanical arm accessory, its three-dimensional model is gradually constructed; geometric information is extracted from the outer contour and used to construct a three-dimensional mesh; in the process of constructing the mesh, it is ensured that the topological structure of the mesh is correct, that is, the connection relationship between vertices, edges and faces conforms to the actual shape of the mechanical arm accessory; at the same time, texture information is applied to the mesh to generate a three-dimensional model with realistic feeling.

[0074] The generated three-dimensional model is optimized to improve its geometric precision and texture quality, which involves steps such as smoothing, removing redundant vertices, adjusting texture resolution, etc.; verification techniques (such as comparing the difference between the model and the original scanning data, visual inspection, etc.) are used to evaluate the accuracy of the model; according to the verification result, the model is adjusted as necessary to ensure that it meets the requirements of subsequent applications.

[0075] Optionally, the generated three-dimensional model has been optimized by smoothing and removing redundant vertices to improve its geometric precision and texture quality; the accuracy of the model is verified by comparing the difference between the model and the original scanning data and by visual inspection; according to the verification result, the model is adjusted as necessary to ensure that it meets the requirements of subsequent virtual simulation and display; through S122 step, multiple side images and outer contour are successfully integrated into the three-dimensional model of the mechanical arm accessory; the three-dimensional model has accurate geometric shape and realistic texture information, which can be used for subsequent virtual simulation, display and analysis applications.

[0076] Therefore, based on the traversal of the three-dimensional model of the robotic arm accessory, a plurality of abnormal positions are determined, and an abnormal area of the three-dimensional model is determined according to the plurality of abnormal positions and corresponding sides, so as to mark the abnormal area of the three-dimensional model. The overall consideration of the plurality of abnormal positions and corresponding sides is compatible, and the accuracy of the abnormal area of the three-dimensional model is guaranteed.

[0077] At this time, the three-dimensional model of the robotic arm accessory is traversed using an algorithm or a tool, which usually involves checking each vertex, edge or face of the model; During the traversal process, data about the model geometry, texture, color, etc. are collected; According to the preset standard or threshold, it is judged which part of the model is inconsistent with the expectation, so as to preliminarily determine the abnormal position.

[0078] During the traversal process, if it is found that a part of the model has a significant difference from the expectation (such as geometric shape deviation, texture loss or color anomaly, etc.), it will be marked as an abnormal position; The determination of the abnormal position involves comparison with the original design data, standard model or historical data; Record the specific information of each abnormal position, including its position, type (such as geometric anomaly, texture anomaly, etc.) and severity.

[0079] According to the spatial distribution and corresponding relationship of the plurality of abnormal positions, the abnormal area in the three-dimensional model is determined; The abnormal area is a continuous curved surface, a group of connected edges or a group of discrete points; Connect the abnormal positions to form the complete boundary of the abnormal area; Consider the side or perspective of the abnormal position to ensure the accuracy and integrity of the abnormal area.

[0080] The determined abnormal area is marked for subsequent analysis and processing; The marking is in the form of direct color addition, highlight display or annotation addition on the model; The marking information should include the type, position, severity and repair suggestion of the abnormal area; Generate a detailed abnormal report listing all detected abnormal areas and related information.

[0081] Specifically, assuming that there is a three-dimensional model of a robotic arm accessory that needs to be quality detected to find potential manufacturing defects; A special software tool is used to traverse the three-dimensional model of the robotic arm accessory; The tool can check each vertex, edge and face of the model; During the traversal process, data about the model geometry, texture and color are collected; According to the preset geometric tolerance, texture integrity and color consistency standard, several abnormal positions are preliminarily determined.

[0082] After the traversal is completed, it is found that several vertex positions on one side of the model have a significant difference from the expectation, which is manifested as geometric shape deviation; These vertices are marked as abnormal positions, and their specific information including position, type and severity are recorded.

[0083] According to the spatial distribution of the abnormal positions, the abnormal positions are connected to form a continuous abnormal area boundary; considering the side where the abnormal positions are located, the accuracy and integrity of the abnormal area are ensured; through analysis, it is determined that the abnormal area is a geometric shape deviation area caused by uneven material in the manufacturing process. Optionally, the abnormal area is an abnormal surface of the side of the three-dimensional model of the robot arm accessory, an abnormal surface with color difference, or an abnormal surface with disordered lines.

[0084] The determined abnormal area is marked, and red highlighting is used to represent its severity; annotations are added to the model to explain the type, location, and repair suggestions of the abnormal area; a detailed abnormal report is generated, listing all detected abnormal areas and their related information, including abnormal position, type, severity, and repair suggestions.

[0085] In an embodiment of the present application, according to the influence degree of different parts of the robot arm accessory on the overall performance, the weight of each position or feature is set; the weight is a value between 0 and 1, representing the importance of the position or feature. For each position, according to the actual situation of its geometric shape, texture, color and other features, the score is calculated; the score is a value between 0 and 100, representing the qualification degree of the position; the score calculation formula is: score = weight1 x feature1 score + weight2 x feature2 score +… A score calculation table is collected, as shown in Table 1:

[0086] Table 1 Score calculation table

[0087] Serial number Position Geometry score Texture score Color score Total score 1 Vertex A 80 (deviation 0.2 mm) 100 (complete) 100 (normal) 90 2 Edge B 100 (correct length) 60 (missing) 100 (normal) 87 3 Face C 100 (good flatness) 100 (complete) 80 (abnormal) 93.3

[0088] According to the total score, a threshold value (e.g. 90 points) is set, and positions with scores below the threshold value are determined as abnormal positions; according to the distribution of abnormal positions in the model, the abnormal area is determined and marked; the total score of vertex A, edge B and face C is less than 90 points, so they are determined as abnormal positions; in the model, the area where these abnormal positions are located is marked as red highlighting or annotations are added to indicate that the area has abnormalities.

[0089] Reference Figure 4 In step S13, the defect area and the visual error area are determined according to the current position of each abnormal area and the current pose of the robot arm accessory.

[0090] In the specific implementation process of the present application, the specific steps are as follows:

[0091] S131: determining the current position of each abnormal area based on the position detection of each abnormal area, and determining the corresponding abnormal feature according to the current position of each abnormal area and the area form of each abnormal area;

[0092] S132: Determine an anomaly detection route according to the plurality of anomaly features and the current pose of the robot arm accessory, and perform autonomous detection of the robot arm accessory along the anomaly detection route; at this time, the plurality of anomaly regions are autonomously detected under the control of the anomaly detection route;

[0093] S133: Determine a first anomaly parameter according to the autonomous detection of the plurality of anomaly regions, and determine a defect region and a visual error region according to the first anomaly parameter, the corresponding anomaly feature, and the anomaly mapping relationship.

[0094] In the embodiments of the present application, the current position of each anomaly region is determined based on the position detection of each anomaly region, and the corresponding anomaly feature is determined according to the current position of each anomaly region and the region form of each anomaly region, which comprehensively considers the current position of each anomaly region and the region form of each anomaly region, and ensures the accuracy of the corresponding anomaly feature.

[0095] At this time, the robot arm accessory is scanned or imaged using a high-precision sensor (such as a laser range finder, radar, vision sensor, etc.); through image processing or signal processing technology, the edge, contour or feature point of the anomaly region is extracted from the scanning or imaging data; according to the extracted feature information, combined with the geometric model and coordinate system of the robot arm accessory, the accurate position of the anomaly region is calculated.

[0096] Optionally, assuming that there is a suspected abnormal region on the robot arm accessory, detailed detection is needed to determine its position and anomaly feature; a high-resolution vision sensor is used to take pictures of the robot arm accessory to obtain clear images; through image processing technology, the edge and contour of the suspected abnormal region are extracted; combined with the CAD model of the robot arm accessory, the accurate position of the anomaly region in three-dimensional space is determined (such as located near the elbow joint of the robot arm, X = 100 mm, Y = 50 mm, Z = 20 mm away from a certain reference point).

[0097] Based on the current position and morphology of the abnormal area, determine its abnormal features, at this time, analyze the shape, size, boundary definition and other morphological characteristics of the abnormal area; compare the differences between the abnormal area and the surrounding normal area, identify the abnormal type (such as geometric shape deviation, texture abnormality, color change, etc.); According to the morphological characteristics and difference degree of the abnormal area, determine its corresponding abnormal feature description (such as "circular depression", "striped texture loss" and the like). Optionally, observe the morphology of the abnormal area, find that it presents a obvious circular depression with a diameter of about 10mm; compare the texture and color of the abnormal area and the surrounding normal area, find that the texture of the abnormal area is missing and the color is dark; According to these morphological characteristics, it is determined that the abnormal feature of the abnormal area is "circular depression, texture loss, color dark"; Through the above steps, the current position and abnormal features of the abnormal area on the mechanical arm accessory are successfully determined, which provides an important basis for subsequent analysis and processing.

[0098] Further, determine the abnormal detection route according to the multiple abnormal features and the current posture of the mechanical arm accessory, and perform autonomous detection on the mechanical arm accessory along the abnormal detection route; At this time, multiple abnormal areas are under the control of the abnormal detection route for autonomous detection, which takes into account the overall consideration of multiple abnormal features and the current posture of the mechanical arm accessory, and ensures the accuracy of the abnormal detection route.

[0099] At this time, an efficient and accurate abnormal detection route is planned to cover all abnormal areas, and all determined abnormal features and their position information, as well as the current posture information of the mechanical arm accessory (such as joint angle, end effector position, etc.) are summarized; Based on the position distribution of the abnormal area and the kinematic constraints of the mechanical arm (such as joint range of motion, obstacle avoidance requirements, etc.), analyze the detection path; Considering the detection efficiency (such as shortest path, least time) and accuracy (such as detection angle, lighting conditions, etc.), use path planning algorithm (such as A* algorithm, Dijkstra algorithm or heuristic search algorithm) to optimize the detection route; Finally generate one or more abnormal detection routes to ensure that each abnormal area can be effectively covered.

[0100] Optionally, assuming that there are three abnormal areas A, B and C on the mechanical arm accessory, which are located at different positions and heights of the mechanical arm; Collect the position information and feature description (such as shape, size, color, etc.) of the abnormal areas A, B and C; Analyze the kinematic constraints of the mechanical arm and the surrounding environment to determine the detection path; Use path planning algorithm to optimize the detection route to ensure that the mechanical arm can cover all abnormal areas with the shortest path; The final generated detection route is a path from the base of the mechanical arm, first to abnormal area A, then to B and C in turn, and finally returns to the base.

[0101] According to the planned detection route, each abnormal area is autonomously and accurately detected; at this time, the robot arm is moved to the starting point of the detection route, and the required sensors or tools for detection are prepared; the robot arm is controlled to move along the detection route, and sensors such as visual sensors, force sensors, temperature sensors, etc. are used to conduct detailed detection on each abnormal area; during the detection process, key data such as the geometry, texture, color, and temperature of the abnormal area are collected in real time; the collected data are associated with the position, characteristics, etc. of the abnormal area and stored in a database for subsequent analysis.

[0102] During the entire detection process, the position, speed, and acceleration, etc. of the robot arm are monitored in real time by a closed-loop control system to ensure that the robot arm can accurately and stably move along the detection route; at the same time, the detection strategy of the robot arm (such as detection speed, sensor configuration, etc.) is dynamically adjusted according to the characteristics of the abnormal area and the detection requirements to improve the detection efficiency and accuracy.

[0103] Optionally, the robot arm is moved to the starting point of the detection route (i.e. near the robot arm base); the robot arm is controlled to move along the planned detection route, and a visual sensor is used to conduct detailed detection on the abnormal area A to collect data such as its geometry, texture, and color; then, the robot arm continues to move to the abnormal areas B and C for detailed detection and data collection; during the entire detection process, the state information of the robot arm is monitored in real time by a closed-loop control system to ensure the safety and accuracy of the detection process; finally, the collected data are associated with the position, characteristics, etc. of the abnormal area and stored in a database to provide data support for subsequent analysis and processing; through the above steps, an efficient abnormal detection route is successfully planned and executed, and multiple abnormal areas on the robot arm accessory are autonomously and accurately detected.

[0104] Therefore, the first abnormal parameter is determined according to the autonomous detection of the multiple abnormal areas, and the defect area and the visual error area are determined according to the first abnormal parameter, the corresponding abnormal morphology abnormal feature, and the abnormal mapping relationship, which comprehensively considers the first abnormal parameter, the corresponding abnormal morphology abnormal feature, and the abnormal mapping relationship, guarantees the accuracy of the defect area and the visual error area, introduces the defect area and the visual error area, and comprehensively controls the casting process and the detection process of the robot arm accessory, and is compatible with the targeted control of the machine vision head.

[0105] At this time, the key first abnormal parameters are extracted from the autonomously detected data for subsequent analysis; at the same time, the data collected during the autonomous detection process are sorted, including geometric shape data, texture data, color data, etc.; based on the sorted data, the feature parameters of each abnormal area are extracted, such as geometric deviation, texture loss degree, color difference value, etc.; according to the significance of the abnormal features and the influence degree on the performance of the robotic arm accessory, the most critical feature parameters are selected as the first abnormal parameters; the first abnormal parameters are usually determined by comparing the numerical size, coefficient of variation or correlation degree with other performance indicators of different feature parameters.

[0106] Optionally, assuming that there are two abnormal areas D and E on the robotic arm accessory, after autonomous detection, the following first abnormal parameters are extracted: abnormal area D: large geometric deviation, moderate texture loss degree, small color difference value; abnormal area E: small geometric deviation, severe texture loss degree, large color difference value. According to the data of autonomous detection, the feature parameters of each abnormal area are extracted, and the most critical feature parameters are selected as the first abnormal parameters; here, the geometric deviation and the texture loss degree are selected as the first abnormal parameters of areas D and E, because they usually have a high weight in the performance evaluation of the robotic arm accessory.

[0107] Based on the first abnormal parameters and the abnormal features, the defect areas and the visual error areas on the robotic arm accessory are identified; at this time, according to the design specifications, manufacturing standards and historical experience of the robotic arm accessory, a mapping relationship between the abnormal features and the potential defects or visual errors is established, which is usually an empirical process and needs to consider multiple factors; according to the first abnormal parameters and the abnormal morphological features, the mapping relationship of the abnormality is compared, and the defect areas on the robotic arm accessory are identified, which are usually caused by errors in the manufacturing process, material problems or wear and corrosion caused by long-term use; at the same time, the visual errors (such as sensor noise, changes in lighting conditions, etc.) that occur during the autonomous detection process are analyzed, and the visual error areas are identified, which are caused by the limitations of the detection equipment or environmental factors, resulting in false detection or missed detection; combined with the identification results of the defect areas and the visual error areas, a comprehensive judgment is made to ensure the accuracy of the identification.

[0108] Optionally, according to the design specifications and historical experience of the robotic arm accessory, a mapping relationship between the abnormal features and the potential defects or visual errors is established; for example, a large geometric deviation means that the size control in the manufacturing process is not proper, and a severe texture loss degree indicates that the material surface has been damaged or corroded.

[0109] According to the mapping relationship of the abnormality, it is identified that the abnormal area D is a defect area caused by improper size control in the manufacturing process, and the abnormal area E is a defect area caused by material surface damage or corrosion due to long-term use.

[0110] At the same time, the visual error in the autonomous detection process is analyzed, and the visual error area is identified; in this example, due to the influence of sensor noise, the color difference value of the abnormal area E has a certain error; however, due to the severity of the texture loss degree, it is still considered that the E area is a real defect area, and after comprehensive judgment, the defect areas D and E on the mechanical arm accessory are determined, and the visual error area is excluded, which provides accurate targets and directions for subsequent repair and improvement.

[0111] Reference Figure 5 In step S14, the autonomous regulation of the machine vision head is triggered based on the visual error area and the plurality of side images to gradually avoid the visual error area;

[0112] In the specific implementation process of the present application, the specific steps are:

[0113] S141: In the visual error area, a plurality of sides to be detected are determined according to the position of the visual error area, and the actual side of the mechanical arm accessory is marked according to the tracing of the plurality of sides to be detected; the shooting difference area is determined according to the comparison of the plurality of sides to be detected and the actual side;

[0114] S142: Determine the optimized shooting parameters according to the shooting difference area, the corresponding previous shooting parameters of the machine vision head, and the shooting parameter mapping relationship;

[0115] S143: Load the optimized shooting parameters into the shooting parameter database of the machine vision head to generate the best shooting parameter combination of the machine vision head for the mechanical arm accessory, and gradually avoid the visual error area.

[0116] In the embodiment of the present application, in the visual error area, a plurality of sides to be detected are determined according to the position of the visual error area, and the actual side of the mechanical arm accessory is marked according to the tracing of the plurality of sides to be detected; the shooting difference area is determined according to the comparison of the plurality of sides to be detected and the actual side, which is compatible with the overall consideration of the comparison of the plurality of sides to be detected and the actual side, and ensures the accuracy of the shooting difference area.

[0117] At this time, a plurality of sides of the mechanical arm accessory associated with the visual error area are identified, which are the source of the visual error or are affected by it; at the same time, the visual error area is analyzed accurately in position, which usually involves the determination of three-dimensional coordinates; according to the position information, in combination with the three-dimensional model or design drawing of the mechanical arm accessory, the sides directly related to the visual error area are identified, which are the external surface of the mechanical arm accessory and also part of the internal structure, and they are visually represented as errors or abnormalities; the identified sides are listed in a list for subsequent detection and analysis.

[0118] Mark the side to be detected on the physical object for subsequent photography and comparative analysis. At this time, trace the origin of each side to be detected, including its source, manufacturing process, and influencing factors, which helps to understand the causes of visual errors. Use a marker pen, stickers, or other methods to mark the side to be detected on the physical object of the robotic arm accessory. The mark should be clear and accurate to easily identify when taking pictures. At the same time of marking, record the relevant information of each side, such as position, size, shape, etc., for subsequent analysis.

[0119] Determine the shooting difference area by comparing the side to be detected with the actual photographed image. At this time, use a machine vision head or other imaging equipment to take pictures of the marked side, ensuring consistent shooting conditions (such as lighting, focal length, etc.). Compare the photographed image with the expected, error-free image, which is automatically done through image processing software or machine vision algorithms. In the comparison process, identify areas that do not match the expected image, i.e., the shooting difference area, which appears as inconsistencies in color, texture, shape, or size. Record the location, size, and features of the shooting difference area for subsequent analysis and processing.

[0120] Specifically, assume that there is a visual error area in the end effector part of the robotic arm accessory, which shows color abnormalities. Position analysis shows that the visual error area is located on side B of the end effector. In combination with the three-dimensional model, side B is identified as the side directly related to color abnormalities. Side B is included in the list of sides to be detected.

[0121] Conduct traceability analysis on side B and find that the side has been subjected to improper heat treatment during the manufacturing process, resulting in color changes. Use a marker pen to make a clear mark on side B on the physical object of the robotic arm accessory. Record the position, size, and shape information of side B.

[0122] Use a machine vision head to take pictures of the marked side B. Compare the photographed image with the expected, error-free image. In the comparison process, identify a clear color difference area that is consistent with the marked position on side B. Record the location, size, and features of the color difference area, such as the degree and range of color deviation. Through the above steps, the side to be detected associated with the visual error area is successfully determined and marked on the physical object. At the same time, by comparing the photographed image with the expected image, the shooting difference area is determined, providing a basis for subsequent processing and optimization.

[0123] Further, determine the optimized shooting parameters based on the shooting difference area, the previous shooting parameters corresponding to the machine vision head, and the mapping relationship of the shooting parameters, which takes into account the overall consideration of the shooting difference area, the previous shooting parameters corresponding to the machine vision head, and the mapping relationship of the shooting parameters, ensuring the accuracy of the optimized shooting parameters.

[0124] At this time, deeply understand the characteristics of the shooting difference area, which is the basis for optimizing the shooting parameters; at the same time, carefully analyze the image of the shooting difference area, identify the abnormalities in color, brightness, contrast, texture or shape, etc.; if the difference is quantified, such as measuring the numerical value of color deviation, the percentage of brightness change, etc.; based on the characteristics of the difference, infer the reasons leading to these differences, such as insufficient exposure, inaccurate white balance, improper focus, etc. Optionally, assuming that the shooting difference area of the robotic arm accessory mainly shows color deviation, and it is inferred that it is caused by insufficient exposure; analyze the shooting difference area, find that the color deviation is mainly concentrated in a certain specific area of the robotic arm accessory, and the deviation degree is obvious; quantify the color deviation, and measure the numerical value of the color deviation as ΔE = 5 (a commonly used color difference measurement standard); infer that insufficient exposure is the main reason for color deviation.

[0125] Understand the parameters used by the machine vision head when shooting the robotic arm accessory in the past, at this time, retrieve the parameter records used to shoot the robotic arm accessory in the past from the shooting parameter database of the machine vision head; organize these parameters, including exposure time, aperture size, ISO value, white balance setting, focal length, etc.; analyze the association between these parameters and the shooting difference area. Optionally, retrieve the parameters used to shoot the robotic arm accessory in the past from the shooting parameter database of the machine vision head, and find that the exposure time is set to 1 / 125 seconds; organize and analyze these parameters, especially the association between the exposure time and the color deviation.

[0126] According to the characteristics of the shooting difference area and the past shooting parameters, the optimized shooting parameters are determined by using the predefined shooting parameter mapping relationship, at this time, the mapping rule matching the characteristics of the current shooting difference area is retrieved from the mapping relationship database; according to the retrieved mapping rule, the past shooting parameters are adjusted to reduce or eliminate the shooting difference; the adjusted parameters are further optimized to ensure that they meet the shooting needs of the robotic arm accessory and minimize visual errors. Optionally, apply the predefined shooting parameter mapping relationship to retrieve the mapping rule matching the characteristics of the current color deviation, which suggests increasing the exposure time to reduce the color deviation; according to the mapping rule, the exposure time is adjusted from 1 / 125 seconds to 1 / 60 seconds, and other parameters remain unchanged; further optimize the adjusted parameters to ensure that they meet the shooting needs of the robotic arm accessory and minimize color deviation.

[0127] Further, ensure that the optimized shooting parameters can reduce or eliminate the shooting differences in actual shooting; use the optimized shooting parameters for preliminary shooting test of the mechanical arm accessory; compare the preliminary test image with the expected image to evaluate the effect of the optimized parameters; if the preliminary test effect is not ideal, fine-tune the parameters according to the evaluation results and repeat the test process until the satisfactory shooting effect is achieved.

[0128] Optionally, use the optimized shooting parameters (exposure time 1 / 60 s) to perform preliminary shooting test of the mechanical arm accessory; compare the preliminary test image with the expected image, and find that the color deviation is significantly reduced, and the numerical value of the color deviation is reduced to ΔE = 2; according to the evaluation results, it is considered that the effect of the optimized parameters is ideal, and further fine-tuning is not needed; through the above steps, the optimized shooting parameters are successfully determined, and their effects are verified in actual shooting, and these optimized parameters will be used for subsequent shooting of the mechanical arm accessory to reduce or eliminate visual errors.

[0129] Therefore, the optimized shooting parameters are loaded into the shooting parameter database of the machine vision head to generate the best shooting parameter combination of the machine vision head for the mechanical arm accessory, and gradually avoid the visual error area, and the best shooting parameter combination of the machine vision head for the mechanical arm accessory is introduced.

[0130] At this time, the optimized and verified shooting parameters are loaded into the shooting parameter database of the machine vision head; at the same time, it is ensured that the optimized shooting parameters (such as exposure time, aperture size, ISO value, white balance setting, etc.) are ready and can reduce or eliminate visual errors; access the shooting parameter database of the machine vision head through the software interface or control system of the machine vision head; upload the optimized shooting parameters into the database to ensure that they can cover or replace the original parameters that cause visual errors.

[0131] Based on the loaded optimized parameters, the best shooting parameter combination for the mechanical arm accessory is generated, at this time, the optimized shooting parameters are integrated together to form a complete parameter combination; after integration, the effectiveness of the parameter combination is verified again to ensure that they can reduce or eliminate visual errors in actual shooting;

[0132] The verified best shooting parameter combination is stored in the control system of the machine vision head so as to be directly called in subsequent shooting.

[0133] With the optimized shooting parameter combination, gradually reduce or completely avoid the visual error area on the mechanical arm accessory; At this time, continuously monitor the image quality during subsequent shooting, especially the visual error area where color deviation occurred before; If visual errors still exist, fine-tune the shooting parameters according to the error type and characteristics, and repeat the verification process; Record the results of each shooting and the parameter adjustment process, analyze which parameters are most effective in reducing visual errors, in order to further optimize future shooting.

[0134] Specifically, assuming that the shooting parameters of the machine vision head have been optimized through step S142, these optimized parameters will now be loaded into the machine vision head and the best shooting parameter combination will be generated; The optimized shooting parameters include: exposure time 1 / 60s, aperture F / 5.6, ISO 200, white balance set to daylight mode; Through the software interface of the machine vision head, access its shooting parameter database; Upload the above-mentioned optimized parameters to the database, replacing the original exposure time 1 / 125s, aperture F / 8, ISO 400, automatic white balance setting.

[0135] Integrate the optimized parameters into a complete parameter combination: exposure time 1 / 60s, aperture F / 5.6I, SO 200, white balance daylight mode; After integration is complete, a test shooting of the mechanical arm accessory is performed using these parameters, and it is found that the visual error area is significantly reduced; Store the verified best shooting parameter combination in the control system of the machine vision head, and name it "mechanical arm accessory best shooting parameters".

[0136] During subsequent shooting, continuously monitor the image quality, especially the visual error area where color deviation occurred before; It is found that the image quality after shooting with optimized parameters has improved significantly, and color deviation has almost completely disappeared; Record the results of each shooting and the parameter adjustment process, and analyze that exposure time and white balance setting are most effective in reducing color deviation; In future shooting, these optimized parameters will continue to be used, and fine-tuning will be made according to actual situation to ensure continuous improvement of image quality. Through the above steps, the optimized shooting parameters are successfully loaded into the machine vision head, and the best shooting parameter combination for the mechanical arm accessory is generated. These optimized parameters will help to gradually reduce or completely avoid the visual error area in subsequent shooting, improve image quality and detection accuracy.

[0137] Reference Figure 6 In step S15, based on the defect type corresponding to the defect area and the corresponding surface, a plurality of to-be-traced processes are determined, and the casting process table of the optimized mechanical arm accessory is determined according to the past data corresponding to the plurality of to-be-traced processes and the defect area;

[0138] In the specific implementation process of the present application, the specific steps are:

[0139] S151: In the defect area, the defect type corresponding to the defect area is determined according to the identification of the defect area, and the corresponding surface is determined according to the position of the defect area, and the plurality of to-be-traced processes are determined according to the traceability of the defect type corresponding to the defect area and the corresponding surface, at this time, the plurality of to-be-traced processes correspond to the forming processes of the surface with the defect area respectively;

[0140] S152: The plurality of to-be-traced processes corresponding to the previous data are determined according to the plurality of to-be-traced processes and the casting database of the mechanical arm accessory; the optimized casting parameters are determined according to the previous data, the defect type corresponding to the defect area, and the casting parameter mapping relationship;

[0141] S153: The optimized casting parameters are loaded to the casting parameter database of the mechanical arm accessory to generate the best casting parameter combination of the mechanical arm accessory, and the optimized casting process table of the mechanical arm accessory is output.

[0142] In the embodiment of the application, in the defect area, the defect type corresponding to the defect area is determined according to the identification of the defect area, and the corresponding surface is determined according to the position of the defect area, and the plurality of to-be-traced processes are determined according to the traceability of the defect type corresponding to the defect area and the corresponding surface, at this time, the plurality of to-be-traced processes correspond to the forming processes of the surface with the defect area respectively, which is compatible with the overall consideration of the traceability of the defect type corresponding to the defect area and the corresponding surface, and ensures the accuracy of the plurality of to-be-traced processes.

[0143] At this time, in the visual detection process of the mechanical arm accessory, the area with defects is accurately identified, and at the same time, the mechanical arm accessory is scanned in all directions using a machine vision system; through image processing algorithms such as edge detection, texture analysis, color recognition, etc., the area with obvious differences from the normal area is identified as a defect area; the position, size, shape and other characteristics of the defect area are marked and recorded.

[0144] According to the characteristics of the defect area, the type of the defect is determined, at this time, a defect type database is established, which contains characteristic descriptions and picture examples of various defects (such as cracks, pores, inclusions, shrinkage holes, etc.); the characteristics of the defect area are compared with the defect types in the database to find the most matched defect type.

[0145] The specific surface where the defect is located is determined, at this time, the three-dimensional coordinates of the defect area are obtained by using the three-dimensional positioning function of the machine vision system; the three-dimensional coordinates of the defect area are mapped to the specific surface in combination with the three-dimensional model of the mechanical arm accessory; the surface name and position information where the defect is located are marked and recorded.

[0146] According to the defect type and the corresponding surface, trace the molding process that causes the defect, at this time, establish the association database of molding process and defect type, record the defect type caused by each molding process; according to the defect type and the corresponding surface, find the molding process that causes the defect in the association database; if there are multiple molding processes, further analysis is needed in combination with actual production situation, process parameters, equipment state and other factors.

[0147] Determine the molding process related to the defect as the process to be traced, at this time, based on the results of traceability analysis, select the molding process that most causes the defect as the process to be traced; if there are multiple processes to be traced, they need to be sorted according to priority for subsequent investigation and improvement.

[0148] Specifically, assuming that a significant crack defect is found in the visual inspection process of the mechanical arm accessory; the machine vision system scans the mechanical arm accessory, identifies the crack defect area, and marks its position, size and shape; compare the features of the crack defect area with the defect type database to determine that the defect is of the "crack" type.

[0149] Use the three-dimensional positioning function of the machine vision system to obtain the three-dimensional coordinates of the crack defect and map them onto the upper surface of the mechanical arm accessory; according to the crack defect type and the upper surface information, find the molding process that causes the crack in the association database of molding process and defect type; after analysis, it is found that uneven cooling in the casting process is the main reason for the crack; determine the cooling process as the process to be traced and prepare for further investigation and improvement; through the above steps, the defect area on the mechanical arm accessory is successfully identified, the defect type and the corresponding surface are determined, and the molding process that causes the defect is traced, which will provide an important basis for subsequent investigation and improvement.

[0150] Further, according to the multiple processes to be traced and the casting database of the mechanical arm accessory, determine the past data corresponding to the multiple processes to be traced; according to the past data, the defect type corresponding to the defect area and the casting parameter mapping relationship, determine the optimized casting parameters, compatible with the overall consideration of the past data, the defect type corresponding to the defect area and the casting parameter mapping relationship, to ensure the accuracy of the optimized casting parameters.

[0151] At this time, retrieve the past data related to the multiple processes to be traced from the casting database of the mechanical arm accessory, at the same time, determine the specific name and version of the process to be traced, as well as the related casting batch or production batch; access the casting database, retrieve the relevant past data according to the name, version and batch information of the process to be traced; the past data includes casting parameters (such as temperature, time, pressure, alloy composition, etc.), equipment state, environmental conditions, operator information, etc.; organize the retrieved past data into tables or reports for subsequent analysis.

[0152] According to the mapping relationship between the defect type and the casting parameter, it is determined which casting parameters are related to the generation of the defect; at this time, a mapping relationship database between the casting parameter and the defect type is established, which records the influence degree of different casting parameters on the generation of the defect; according to the past data searched in step one and the defect type corresponding to the defect area, the related casting parameters are searched in the mapping relationship database; the correlation between these casting parameters and the generation of the defect is analyzed, and it is determined which parameters are the key factors leading to the defect.

[0153] According to the mapping relationship between the past data and the casting parameter, the optimized casting parameter is determined; at this time, according to the analysis result, the casting parameter leading to the defect is adjusted; the principle of adjustment is to reduce or eliminate the generation of the defect, while ensuring the quality and performance of the casting; the methods such as test verification, simulation analysis, expert consultation and the like are adopted to determine the best combination of casting parameters; the optimized casting parameter is recorded and prepared to be updated to the casting database.

[0154] Specifically, it is assumed that in the S151 step, it is determined that the crack defect on the mechanical arm accessory is related to uneven cooling in the casting process, and two processes to be traced are involved: casting process A and casting process B; the past data related to the casting process A and the casting process B are searched from the casting database; the casting parameter records of these two processes in the last few batches are found, including the casting temperature, the cooling time, the alloy composition and the like.

[0155] The mapping relationship database between the casting parameter and the defect type is applied; it is found that the casting temperature being too high and the cooling time being too short both lead to the generation of the crack defect; in the searched past data, it is found that the casting temperature of the casting process A is generally too high, while the cooling time of the casting process B is generally too short. Based on the above analysis, the optimized casting parameter is determined; for the casting process A, it is decided to reduce the casting temperature by 50℃; for the casting process B, it is decided to extend the cooling time by 10 minutes, so that after the adjustment, it is expected to reduce or eliminate the generation of the crack defect. Through the above steps, the past data related to the process to be traced is successfully searched, the casting parameter mapping relationship is applied, and the optimized casting parameter is determined, which will provide important guidance for subsequent production improvement.

[0156] Therefore, the optimized casting parameter is loaded to the casting parameter database of the mechanical arm accessory to generate the best combination of casting parameters of the mechanical arm accessory, and the casting process table of the optimized mechanical arm accessory is output, the casting process table of the optimized mechanical arm accessory is introduced, the optimization of the casting process and the detection process of the mechanical arm accessory is realized, and the accuracy of the defect detection of the mechanical arm accessory after casting is ensured, and the subsequent influence of the visual error area is avoided.

[0157] At this time, the optimized casting parameters are loaded into the casting parameter database of the mechanical arm accessory to replace the original parameters that caused defects; ensure that the optimized casting parameters have been fully verified and tested to ensure that they can effectively reduce or eliminate defects while ensuring the quality and performance of the casting; access the casting parameter database to find the entry corresponding to the casting parameters to be optimized; input the optimized casting parameter values into the corresponding fields to replace the original parameter values; if the database supports version control, create a new version for the new parameter combination and record the time, reason, and modifier of the modification; ensure that other related data in the database (such as process description, equipment requirements, etc.) are consistent with the new casting parameters.

[0158] Combine other related casting parameters (such as mold design, alloy composition, equipment status, etc.) to generate the best casting parameter combination for the mechanical arm accessory; at this time, all factors affecting the quality of the casting are considered, including but not limited to casting temperature, cooling time, pressure, alloy composition, mold design, etc.; use data analysis, simulation, or expert experience to determine the best matching relationship between factors; combine the optimized casting parameters with other related parameters to form a complete set of best casting parameter combinations for the mechanical arm accessory; record and save the best casting parameter combination for subsequent production and use.

[0159] According to the best casting parameter combination, develop a detailed casting procedure table to guide production personnel to perform casting operations according to the optimized parameters; at this time, the best casting parameter combination is converted into specific operation steps and process requirements; develop a casting procedure table including the name of each process, operation steps, parameter settings, and detailed information such as matters needing attention; ensure that the content of the casting procedure table is clear, accurate, easy to understand, and meets the requirements of actual production operations; output the casting procedure table as a paper or electronic file and distribute it to production personnel and relevant personnel.

[0160] Specifically, suppose in the S152 step, the optimized casting parameters are determined as follows: the casting temperature is reduced by 50°C to 600°C, the cooling time is extended by 10 minutes to 60 minutes, and the silicon content in the alloy composition is increased to 12%; load these optimized parameters into the casting parameter database of the mechanical arm accessory; in the database, find the entry corresponding to the current casting batch and modify the casting temperature to 600°C, the cooling time to 60 minutes, and the silicon content in the alloy composition to 12%; at the same time, record the time, reason, and modifier of the modification.

[0161] The optimal casting parameter combination for the mechanical arm accessory is generated in combination with factors such as mold design (e.g., mold preheating temperature, mold material, etc.), equipment state (e.g., power of the melting furnace, performance of the cooling system, etc.), etc.; for example, the mold preheating temperature is determined to be 300°C, the power of the melting furnace is 150kW, and the working pressure of the cooling system is 0.6MPa, etc.

[0162] According to the optimal casting parameter combination, a detailed casting procedure table is formulated; for example, procedure one: melting the alloy, the operation steps include weighing the alloy raw materials, adding them to the melting furnace, adjusting the melting temperature to 600°C and keeping it at this temperature until the alloy is completely melted; procedure two: pouring, the operation steps include pouring the melted alloy liquid into the mold preheated to 300°C; procedure three: cooling, the operation steps include keeping the alloy liquid in the mold slowly cooled to room temperature within 60 minutes; procedure four: demolding and cleaning, the operation steps include opening the mold, taking out the casting, cleaning the impurities on the surface of the casting, etc.; these steps and parameter settings are recorded in the casting procedure table and distributed to the production personnel and relevant personnel; through the above steps, the optimized casting parameters are successfully loaded into the database, the optimal casting parameter combination is generated, and the detailed casting procedure table is output, which will provide important guidance and basis for subsequent production.

[0163] In an embodiment of the present application, it is assumed that there is a casting parameter matching table for recording the casting parameters of different mechanical arm accessories; the mechanical arm accessory matching table is shown in Table Two:

[0164] Table Two Mechanical Arm Accessory Matching Table

[0165]

[0166] For the mechanical arm accessory A002, the new casting parameters are determined after optimization: the casting temperature is 600°C, the cooling time is 60 minutes, and the silicon content in the alloy composition is 12%; these parameters are loaded into the database and the mechanical arm accessory matching table is updated to form the updated mechanical arm accessory matching table; the updated mechanical arm accessory matching table is shown in Table Three:

[0167] Table Three Updated Mechanical Arm Accessory Matching Table

[0168]

[0169] It is assumed that three key factors are considered: casting temperature (weight 0.4), cooling time (weight 0.3), and alloy composition (silicon content, weight 0.3); score different parameter values under each factor: casting temperature: 600°C (9 points), 650°C (7 points); cooling time: 60 minutes (8 points), 50 minutes (6 points); alloy composition (silicon content): 12% (10 points), 10% (8 points);

[0170] The comprehensive score of the casting temperature of 600 DEG C is 0.4*9=3.6, the comprehensive score of the cooling time of 650 DEG C is 0.4*7=2.8, the comprehensive score of the alloy composition of 60 minutes is 0.3*8=2.4, the comprehensive score of the alloy composition of 50 minutes is 0.3*6=1.8, the comprehensive score of the silicon content of 12% is 0.3*10=3.0, and the comprehensive score of the silicon content of 10% is 0.3*8=2.4. The combination of the parameter values with the highest comprehensive score is selected as the optimal combination of casting parameters, that is, the casting temperature of 600 DEG C (3.6 points), the cooling time of 60 minutes (2.4 points), and the silicon content of 12% (3.0 points) in the alloy composition; a detailed casting process table is formulated according to the optimal combination of casting parameters; and the casting process table is shown in Table 4:

[0171] Table 4 Casting process table

[0172]

[0173] Please refer to Figure 7 , Figure 7 is a structural composition schematic diagram of a defect detection system for a mechanical arm accessory after casting in an embodiment of the present application; the defect detection system for the mechanical arm accessory after casting comprises:

[0174] a side image module 21 configured to collect multiple side images of the mechanical arm accessory based on a machine vision head;

[0175] an abnormal area module 22 configured to determine a three-dimensional model of the mechanical arm accessory according to the multiple side images and scanning data corresponding to the mechanical arm accessory, and mark an abnormal area of the three-dimensional model;

[0176] a multiple area module 23 configured to determine a defect area and a visual error area according to a current position of each abnormal area and a current posture of the mechanical arm accessory;

[0177] an autonomous regulation module 24 configured to trigger autonomous regulation of the machine vision head based on the visual error area and the multiple side images, so as to gradually avoid the visual error area;

[0178] a casting process table module 25 configured to determine multiple to-be-traced processes based on a defect type corresponding to the defect area and a corresponding surface, and determine an optimized casting process table of the mechanical arm accessory according to previous data corresponding to the multiple to-be-traced processes and the defect area.

[0179] Any combination of the technical features of the above embodiments is possible, and 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 contradictory, they should be considered as the scope of the present application.

Claims

1. A method for detecting defects in a mechanical arm fitting after casting, characterized by, The method comprises the following steps: Collecting multiple side images of the mechanical arm accessory based on the machine vision head; Determining the three-dimensional model of the mechanical arm accessory according to the multiple side images and the scanning data corresponding to the mechanical arm accessory, and marking the abnormal area of the three-dimensional model; Determining the defect area and the visual error area according to the current position of each abnormal area and the current posture of the mechanical arm accessory, comprising: determining the current position of each abnormal area based on the position detection of each abnormal area, and determining the corresponding abnormal feature according to the current position of each abnormal area and the area form of each abnormal area; determining the abnormal detection route according to the multiple abnormal features and the current posture of the mechanical arm accessory, and performing autonomous detection on the mechanical arm accessory along the abnormal detection route; at this time, the multiple abnormal areas are autonomously detected under the control of the abnormal detection route; determining the first abnormal parameter according to the autonomous detection of the multiple abnormal areas, and determining the defect area and the visual error area according to the first abnormal parameter, the corresponding abnormal form abnormal feature and the abnormal mapping relationship; Triggering autonomous regulation and control of the machine vision head based on the visual error area and the multiple side images to gradually avoid the visual error area; Determining multiple to-be-traced processes based on the defect type corresponding to the defect area and the corresponding surface, and determining the optimized casting process table of the mechanical arm accessory according to the previous data corresponding to the multiple to-be-traced processes and the defect area.

2. The mechanical arm attachment casting defect detection method according to claim 1, characterized by, The method comprises the following steps: After the mechanical arm accessory is cast, the shooting space is determined according to the current position of the mechanical arm accessory and the form of the mechanical arm accessory; Triggering multiple machine vision heads in the surrounding according to the shooting space, at this time, the multiple machine vision heads are arranged along the shooting space and shoot the mechanical arm accessory in different directions; Collecting multiple images based on the shooting of the mechanical arm accessory by the multiple machine vision heads, determining multiple side images according to the synthesis of the multiple images, and presenting the state of multiple sides in the mechanical arm accessory according to the multiple side images.

3. The mechanical arm attachment casting defect detection method according to claim 1, characterized by, The method comprises the following steps: Determining the scanning data based on the scanning of the mechanical arm accessory by the scanning device, which is presented in the form of point cloud; determining the outer contour of the mechanical arm accessory according to the synthesis of the scanning data; Determining the three-dimensional model of the mechanical arm accessory according to the synthesis of the multiple side images and the outer contour of the mechanical arm accessory, at this time, the multiple side images and the outer contour of the mechanical arm accessory are synthesized in the same position dimension, and the corresponding three-dimensional model is gradually constructed along the outer contour of the mechanical arm accessory; Determining multiple abnormal positions based on the traversal of the three-dimensional model of the mechanical arm accessory, and determining the abnormal area of the three-dimensional model according to the multiple abnormal positions and the corresponding side to mark the abnormal area of the three-dimensional model; connecting the abnormal positions according to their spatial distribution to form a continuous abnormal area boundary; the abnormal area is a geometric shape deviation area caused by uneven material in the manufacturing process.

4. The mechanical arm attachment casting defect detection method according to claim 1, characterized by, The autonomous regulation of the machine vision head is triggered based on the visual error area and the plurality of side images to gradually avoid the visual error area, including: In the visual error area, a plurality of sides to be detected are determined according to the position of the visual error area, and the actual side of the mechanical arm accessory is marked according to the tracing of the plurality of sides to be detected; and a shooting difference area is determined according to the comparison between the plurality of sides to be detected and the actual side.

5. The mechanical arm attachment method for detecting defects after casting according to claim 4, wherein The autonomous regulation of the machine vision head is triggered based on the visual error area and the plurality of side images to gradually avoid the visual error area, and further includes: The optimized shooting parameters are determined according to the shooting difference area, the previous shooting parameters corresponding to the machine vision head, and a shooting parameter mapping relationship; The optimized shooting parameters are loaded into a shooting parameter database of the machine vision head to generate an optimal shooting parameter combination of the machine vision head for the mechanical arm accessory, and the visual error area is gradually avoided.

6. The mechanical arm attachment casting defect detection method according to claim 1, wherein A plurality of processes to be traced are determined based on the defect type corresponding to the defect area and the corresponding surface, and an optimized casting process table of the mechanical arm accessory is determined according to the previous data corresponding to the plurality of processes to be traced and the defect area, including: In the defect area, the defect type corresponding to the defect area is determined according to the identification of the defect area, and the corresponding surface is determined according to the position of the defect area; and the plurality of processes to be traced are determined according to the tracing of the defect type corresponding to the defect area and the corresponding surface, at this time, the plurality of processes to be traced respectively correspond to the forming process of the surface having the defect area.

7. The mechanical arm attachment method for detecting a defect after casting according to claim 6, wherein A plurality of processes to be traced are determined based on the defect type corresponding to the defect area and the corresponding surface, and an optimized casting process table of the mechanical arm accessory is determined according to the previous data corresponding to the plurality of processes to be traced and the defect area, and further includes: The previous data corresponding to the plurality of processes to be traced is determined according to the plurality of processes to be traced and a casting database of the mechanical arm accessory; and the optimized casting parameters are determined according to the previous data, the defect type corresponding to the defect area, and a casting parameter mapping relationship; The optimized casting parameters are loaded into a casting parameter database of the mechanical arm accessory to generate an optimal casting parameter combination of the mechanical arm accessory, and an optimized casting process table of the mechanical arm accessory is output.

8. A mechanical arm accessory in a post-casting defect detection system, characterized by, The mechanical arm accessory defect detection system after casting is applied to the mechanical arm accessory defect detection method after casting as claimed in any one of claims 1-7, and the mechanical arm accessory defect detection system after casting includes: A side image module configured to acquire a plurality of side images of the mechanical arm accessory based on a machine vision head; An abnormal area module configured to determine a three-dimensional model of the mechanical arm accessory according to the plurality of side images and scanning data corresponding to the mechanical arm accessory, and mark an abnormal area of the three-dimensional model; and An abnormal area module configured to determine a three-dimensional model of the mechanical arm accessory according to the plurality of side images and scanning data corresponding to the mechanical arm accessory, and mark an abnormal area of the three-dimensional model; and The multiple area module is used to determine the defect area and the visual error area according to the current position of each abnormal area and the current posture of the mechanical arm accessory, and comprises: determining the current position of each abnormal area based on the position detection of each abnormal area, and determining the corresponding abnormal feature according to the current position of each abnormal area and the area form of each abnormal area; determining the abnormal detection route according to the multiple abnormal features and the current posture of the mechanical arm accessory, and autonomously detecting the mechanical arm accessory along the abnormal detection route; at this time, the multiple abnormal areas are autonomously detected under the control of the abnormal detection route; determining the first abnormal parameter according to the autonomous detection of the multiple abnormal areas, and determining the defect area and the visual error area according to the first abnormal parameter, the corresponding abnormal form abnormal feature and the abnormal mapping relationship; The autonomous control module is used to trigger the autonomous control of the machine vision head based on the visual error area and the multiple side images, so as to gradually avoid the visual error area; The casting process table module is used to determine the multiple to-be-traced processes based on the defect type corresponding to the defect area and the corresponding surface, and to determine the optimized casting process table of the mechanical arm accessory according to the previous data corresponding to the multiple to-be-traced processes and the defect area.

Images