Burr data acquisition system and image processing method suitable for automobile stamping part

Abstract

The invention provides a burr data acquisition system and an image processing method suitable for an automobile stamping part, and relates to the technical field of image processing, and the method comprises the following steps: obtaining preset angle images of a to-be-detected stamping part at a plurality of different preset visual angles; extracting a target area image of the to-be-detected edge segment from each preset angle image, and generating a target area image sequence; for any physical position point on the edge segment to be detected, traversing all target area images in the target area image sequence, and respectively calculating local image feature saliency of the physical position point in each target area image; selecting a target area image with the highest local image feature saliency as an analysis source image for carrying out burr identification on the physical position point; and based on the analysis source image allocated to each physical position point on the to-be-detected edge section, identifying a burr region on the to-be-detected edge section, performing geometric dimensioning, and outputting a burr data result. The method can improve the detection efficiency and precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a burr data acquisition and image processing method suitable for automobile stamping parts. BACKGROUND

[0002] In the automobile manufacturing industry, stamping parts are the main components of the body-in-white, and the quality of their edge burrs directly affects the process precision of subsequent welding and assembly, as well as the safety and aesthetics of the vehicle. Therefore, high-precision burr detection of stamping parts is a crucial link in the production process.

[0003] Traditional detection methods mainly rely on manual visual inspection and contact measurement. This method is not only inefficient and labor-intensive, but also subject to the subjective experience of the detector, which can easily lead to missed detection and misjudgment, resulting in poor consistency of the detection results and difficulty in meeting the stringent requirements of modern intelligent manufacturing for quality control.

[0004] In order to improve the level of automation of detection, some detection devices based on machine vision have been proposed in the prior art. For example, the utility model patent with publication number CN222125112U discloses a stamping part burr detection device, which includes a workbench, a rotating mechanism, a fixing mechanism, and a CCD detection camera. The workpiece is rotated by a servo motor driving a rotating disc, allowing the CCD camera to capture images of the workpiece from different angles, thereby preliminarily replacing manual detection and improving work efficiency.

[0005] However, such devices only solve the problem of image acquisition automation, and their detection algorithms are usually simple, often based on threshold segmentation or edge extraction of images from a single perspective. Due to the complex structure of stamping parts and the variable edge direction, as well as the small size and irregular shape of burrs, under a single perspective, it is easy to cause the detection algorithm to fail due to poor lighting conditions, insufficient contrast between burrs and background, or obstruction by the workpiece structure itself, resulting in missed detection or measurement accuracy not meeting standards. SUMMARY

[0006] In view of the above-mentioned defects or shortcomings in the prior art, the present application aims to provide a burr data acquisition and image processing method suitable for automobile stamping parts to improve the detection accuracy of stamping part burr data. The method comprises the following steps: Obtaining pre-set angle images of a stamping part to be detected under multiple different pre-set perspectives; Extracting target region images of the edge segment to be measured of the stamping part to be detected from each pre-set angle image to generate a target region image sequence; wherein the edge segment to be measured is formed by connecting a plurality of physical position points; For any of the physical position points on the edge segment to be tested, all the target region images in the target region image sequence are traversed, and the local image feature saliency of the physical position point in each target region image is calculated respectively; The target region image with the highest local image feature saliency is selected as the analysis source image for burr identification of the physical position point; Based on the analysis source image assigned for each of the physical position points on the edge segment to be tested, the burr region on the edge segment to be tested is identified; The burr region is quantified in geometric size, and burr data results are output.

[0007] According to the technical scheme provided in the present application, after identifying the burr region on the edge segment to be tested based on the analysis source image assigned for each of the physical position points on the edge segment to be tested, the following steps are further included: According to the size and position information of all the identified burr regions, a burr distribution map corresponding to the stamping part to be detected is generated; The historical defect mode library is retrieved and traversed to obtain a defect cause corresponding to the burr distribution map corresponding to the stamping part to be detected; wherein the historical defect mode library includes a plurality of burr distribution maps and defect causes corresponding thereto, and the defect cause includes a die wear type or a die failure type; Based on the defect cause corresponding to the burr distribution map corresponding to the stamping part to be detected, corresponding die maintenance prompt information is generated, and the die maintenance prompt information is sent to a production management terminal.

[0008] According to the technical scheme provided in the present application, the method further includes constructing a historical defect mode library; The construction of the historical defect mode library includes the following steps: The burr data of the stamping part and the die identity data of the die used to produce the collected stamping part are associated; the burr data includes the burr distribution map; The die identity data of the die and its maintenance record data are associated; Based on the time sequence association relationship between the burr data and the maintenance record data, a corresponding relationship between the burr distribution map and the die wear type or the die failure type is obtained to obtain the historical defect mode library.

[0009] According to the technical scheme provided in the present application, the corresponding relationship between the burr distribution map and the die wear type or the die failure type based on the time sequence association relationship between the burr data and the maintenance record data includes the following steps: determine a set of stamping parts produced before and after the maintenance based on the maintenance time in the maintenance record data; contrastively analyze the burr distribution patterns corresponding to each of the stamping parts in the set of stamping parts produced before and after the maintenance; determine whether the frequency or severity of the burr distribution corresponding to any of the burr distribution patterns decreases in the set of stamping parts produced after the maintenance; if yes, establish a corresponding relationship between the burr distribution pattern corresponding to the burr distribution pattern and the die wear type or the die failure type solved by the current maintenance operation.

[0010] According to the technical scheme provided in the present application, the determination of whether the frequency or severity of the burr distribution corresponding to any of the burr distribution patterns decreases in the set of stamping parts produced after the maintenance includes the following steps: calculate a first frequency of occurrence and a first average severity of a first burr distribution pattern in the set of stamping parts before the maintenance; calculate a second frequency of occurrence and a second average severity of the first burr distribution pattern in the set of stamping parts after the maintenance; if the decline rate of the second frequency of occurrence relative to the first frequency of occurrence is greater than a first preset threshold value, and / or the decline rate of the second average severity relative to the first average severity is greater than a second preset threshold value, it is determined that the frequency or severity of the burr distribution corresponding to any of the burr distribution patterns decreases.

[0011] According to the technical scheme provided in the present application, the acquisition of the preset angle image of the to-be-detected stamping part under a plurality of different preset viewing angles includes the following steps: acquire the type of the to-be-detected stamping part, and take it as a target stamping part type; retrieve and traverse a collection angle database to obtain a set of preset viewing angles corresponding to the target stamping part type, the set of preset viewing angles including a plurality of preset viewing angles; the collection angle database including a plurality of stamping part types and a set of preset viewing angles corresponding to each stamping part type; control a servo motor to drive a rotating disc and the to-be-detected stamping part fixed thereon to rotate to each preset viewing angle, and after stopping at each angle, trigger a CCD detection camera to collect an image as the preset angle image under the preset viewing angle.

[0012] According to the technical scheme provided in the present application, the method further includes constructing the collection angle database; the construction of the collection angle database includes the following steps: Selecting a plurality of test samples corresponding to the type of the stamping part, ensuring that the test samples cover the common burr states of the type of the stamping part; Fixing a test sample, controlling it to rotate uniformly around an axis, and continuously shooting by the CCD detection camera to obtain a continuous video stream of a week of the test sample; Frame processing the continuous video stream to obtain a plurality of high-definition images corresponding to uniformly distributed angles, constituting a full-view image set of the test sample; Obtaining a pixel-level segmentation mask of the burr region in each high-definition image in the full-view image set; Based on the segmentation mask, calculating the average contrast of the burr region in each high-definition image; Filtering out the high-definition images with an average contrast higher than a third preset threshold, and taking the corresponding rotation angle as an effective candidate view angle of the test sample; Based on the effective candidate view angle of each test sample corresponding to the type of the stamping part, obtaining the preset view angle set of the type of the stamping part; Cycling the above steps to obtain the preset view angle set corresponding to all types of stamping parts respectively, to obtain the collection angle database.

[0013] According to the technical scheme provided by the present application, the effective candidate view angle of each test sample corresponding to the type of the stamping part is obtained, and the preset view angle set of the type of the stamping part is obtained, including the following steps: Summarizing the effective candidate view angles corresponding to all test samples of the type of the stamping part to form an initial set containing a plurality of effective candidate view angles; Statistically analyzing the frequency of occurrence of each effective candidate view angle in the initial set; Filtering out the effective candidate view angles with a frequency of occurrence higher than a preset frequency threshold as the preset view angle corresponding to the type of the stamping part, to constitute the preset view angle set of the type of the stamping part.

[0014] According to the technical scheme provided by the present application, the local image feature saliency of the physical location point in each target region image is calculated respectively, including the following steps: Respectively extracting multi-dimensional image features corresponding to the region where the physical location point is located in each target region image; Fusion computing the multi-dimensional image features to generate a quantitative score value representing the visual saliency of the physical location point in each target region image; According to the quantitative score value of the visual saliency degree of the physical position point in each of the target region images, a local image feature saliency degree in each of the target region images is obtained.

[0015] According to the technical scheme provided in the application, the multi-dimension image features corresponding to the region where the physical position point is located in each of the target region images are extracted respectively, and the method comprises the following steps: For each of the target region images, a local image block of a predetermined size is extracted with the physical position point as the center; A gradient magnitude map and a gradient direction histogram are calculated as gradient features of the local image block; A local binary pattern feature is extracted by an LBP operator as a texture feature of the local image block; The gradient features and the texture features are taken as the multi-dimension image features.

[0016] Compared with the prior art, the application has the following beneficial effects: I. The detection rate and reliability of burr detection are improved: multi-view image sequences and intelligent analysis source image selection strategies are adopted. For each physical position point on the edge segment to be detected, the system calculates the local image feature saliency degree of the point in different view images, and matches a best view (analysis source image) that can best present the features of the point. This method effectively overcomes the problem of missed detection caused by occlusion, uneven illumination and insufficient contrast. Even if a burr is difficult to observe at most angles, as long as there is an angle that can make its features appear, the system can capture it, thereby achieving a high detection rate, and the reliability and consistency of the detection results are improved.

[0017] II. High-precision quantification of burr geometric size is achieved: in a single view, due to perspective distortion and edge pixel blur, there is often a large error in the measurement of burr height and length. The application selects the image with the highest local image feature saliency degree for each point to be measured for analysis, which means that the pixel-level positioning and measurement are performed under the condition that the edge is the clearest and the contrast is the strongest. This is equivalent to automatically selecting the best angle for each small measurement task, thereby greatly improving the precision of geometric size quantification and providing an extremely accurate data basis for process improvement and quality evaluation.

[0018] Third, the adaptive ability and intelligent level of the detection system are improved: compared with the device of simple rotation shooting in the background technology, the technical scheme of the present application embodies a higher order of intelligence. The system is no longer passively collecting images, but actively evaluating image quality and making decisions. It can dynamically select the most suitable analysis data source according to the specific characteristics of different position points. This "point-to-point" optimization strategy enables the system to cope with various complex structure stamping parts and variable field conditions, and the generalization ability and robustness are far superior to traditional visual detection systems with fixed algorithms.

[0019] Fourth, while improving detection quality, efficiency is maintained: although the present application processes multiple images, the entire analysis process is effectively controlled due to its focus on key areas through target area image extraction and the use of intelligent algorithms to quickly calculate saliency rather than processing the entire image. Compared with the subsequent repair costs soaring due to missed detection or repeated detection due to insufficient accuracy, the additional computational cost of this solution is negligible. It successfully balances "detection accuracy" and "efficiency", and truly realizes high-quality, high-efficiency automated intelligent detection. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A step flowchart of the burr data acquisition and image processing method for automobile stamping parts provided by the present application. DETAILED DESCRIPTION

[0021] The present application will be further described in detail below in conjunction with the drawings and examples. It should be understood that the specific examples described herein are only intended to explain the related invention, and not to limit the invention. In addition, it should be noted that only parts related to the invention are shown in the drawings for ease of description.

[0022] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the drawings and in conjunction with the embodiments.

[0023] Example 1 As mentioned in the background art, to solve the problems in the prior art, the present application proposes a burr data acquisition and image processing method for automobile stamping parts, as shown in Figure 1 The method comprises the following steps: S1, obtaining pre-set angle images of a stamping part to be detected at different pre-set viewing angles; S2, extracting a target area image of an edge segment to be detected of the stamping part to be detected from each of the pre-set angle images to generate a target area image sequence; wherein the edge segment to be detected is formed by connecting a plurality of physical position points; S3. For any physical location point on the edge segment to be tested, traverse all target region images in the target region image sequence and calculate the local image feature saliency of the physical location point in each target region image. S4. Select the target region image with the highest local image feature saliency as the source image for spur identification of the physical location point; S5. Based on the analysis source image assigned to each physical location point on the edge segment to be tested, identify the burr region on the edge segment to be tested; S6. Quantize the geometric dimensions of the burr area and output the burr data results.

[0024] Specifically, the stamping parts to be detected refer to automobile metal parts stamped by a die, such as a door, an engine cover, a bracket, etc. The stamping parts on and off the production line or samples for sampling inspection are placed on a detection tool for fixation. The preset viewing angles refer to specific camera angles for shooting the stamping parts, which are not randomly selected but determined through a large number of experiments and analysis, and are the observation angles that can best highlight the edge features of a certain type of stamping part, especially burrs. The implementation is to accurately rotate the workpiece to these preset angles through a programmable rotating platform, such as a servo motor driven turntable. The preset angle image: the high-definition digital image collected by an industrial camera, such as a CCD camera, at each preset viewing angle. When implemented, the lighting environment needs to be stable and uniform to avoid shadows and reflections that interfere with image quality. The edge segment to be detected refers to the profile edge part of the stamping part that needs to be detected for burrs. It can be a whole continuous edge or a segmented edge. The edge segment is physically connected by countless continuous physical position points. In digital image processing, these physical position points correspond to a set of pixel coordinates in the image. When implemented, the edge to be detected can be located from the original image through image segmentation and edge extraction algorithms, such as the Canny operator. The target area image: In order to reduce the amount of calculation and focus on the key area, the entire preset angle image is not processed. Instead, a strip-shaped area image of an appropriate width is extracted centered on the extracted edge segment to be detected. This series of images containing the same physical edge extracted from different angle images constitutes the target area image sequence. Physical position point: refers to a specific, tiny point on the edge segment to be detected. In implementation, the extracted edge line can be densely sampled to obtain a series of discrete coordinate points to represent the edge. Local image feature saliency: a core quantitative indicator for evaluating whether the features near a certain physical position point in a certain image at a certain angle (i.e., a local image block) are clear, obvious, and easy to identify by an algorithm. The higher the saliency, the more prominent the edge, texture, gradient, and other features at that point in the image, which is more conducive to determining whether there is a burr. Analysis source image: for each point on the edge, the saliency in different images is calculated by traversing all angle target area images, and the image with the highest saliency score is finally selected as the best evidence image for determining whether there is a burr at that point. Burr area recognition: after assigning the best analysis source image to each point, burr recognition algorithms (such as based on edge anomaly protrusion detection, template matching, machine learning classifier, etc.) can be executed on these images to accurately find the area where the burr is located. Geometric dimension quantification: the actual physical dimensions of the identified burr pixel area, such as length, height, and area, are calculated through camera calibration and pixel equivalent conversion.Burr data result: The final output of structured data, usually including the location, size, number of burrs, etc. It can be used to generate a detection report or statistical analysis.

[0025] In a preferred embodiment, after identifying the burr area on the edge segment to be tested based on the analysis source image assigned to each physical position point on the edge segment to be tested, the following steps are further included: According to the size and position information of all identified burr areas, generate the burr distribution atlas corresponding to the stamping part to be detected; Retrieve and traverse the historical defect pattern library to obtain the defect cause corresponding to the burr distribution atlas corresponding to the stamping part to be detected; wherein the historical defect pattern library includes a plurality of burr distribution atlases and their corresponding defect causes, and the defect cause includes a mold wear type or a mold failure type; Based on the defect cause corresponding to the burr distribution atlas corresponding to the stamping part to be detected, generate corresponding mold maintenance prompt information, and send the mold maintenance prompt information to the production management terminal.

[0026] Specifically, the burr distribution atlas: is a structured data file or visual graph. It integrates the size (such as large, medium, small) and position information (for example, located at B segment of workpiece A edge C millimeter) of all identified burrs. In implementation, the position and severity level of burrs can be marked on a stamping part diagram with different colors and sizes of marker points, thereby forming an intuitive "lesion map", i.e. burr distribution atlas. Historical defect pattern library: This is a database that stores a large amount of historical case data. Each record contains a typical burr distribution atlas and a verified defect cause that leads to this atlas. In implementation, the library can be a relational database or a knowledge graph, supporting efficient atlas matching and query. Defect cause: refers to the problem type of the mold, mainly including: mold wear type: such as "upper die edge roundness wear", "lower die gap uniformity wear", etc., which is usually progressive. Mold failure type: such as "punch collapse", "guide pillar jam", "spring failure", etc., which is usually sudden. Mold maintenance prompt information: text information automatically generated based on matching results, for example: "Alert: the current workpiece burr distribution pattern (code: M-Pattern-07) matches the 'lower die corner severe wear' pattern in the historical database with a degree of 92%. It is recommended to immediately check and repair the lower die corner of mold D-052." The information is sent to the production management terminal, such as the MES (Manufacturing Execution System) screen, the engineer's mobile terminal or the Kanban system, through the factory network (such as MQTT, HTTP protocol).

[0027] The present embodiment realizes predictive maintenance: the detection result is improved from "post-judgment of workpiece qualification" to "pre-warning of mold health status", allowing intervention before the mold is completely damaged or a large number of waste products are produced, reducing unplanned downtime. It also realizes auxiliary decision-making: it provides clear fault clues for maintenance personnel, narrows down the scope of investigation, and improves maintenance efficiency and accuracy. This method digitizes, standardizes, and embeds the experience of experienced workers ("knowing where the mold is broken by looking at burrs") into the system, avoiding knowledge loss.

[0028] In a preferred embodiment, the method further comprises constructing a historical defect pattern library; The construction of the historical defect pattern library comprises the following steps: Correlate the burr data of the stamping part with the mold identity data of the mold used to produce the stamping part; the burr data includes the burr distribution atlas; Correlate the mold identity data of the mold with its maintenance record data; Based on the time sequence correlation between the burr data and the maintenance record data, the corresponding relationship between the burr distribution atlas and the mold wear type or the mold failure type is obtained to obtain the historical defect pattern library.

[0029] Specifically, the historical defect pattern library is constructed: this is an offline process of continuous learning, usually led by system administrators or data engineers. Mold identity data: information that uniquely identifies a mold, such as mold number, mold system number, production line, production part type, etc. In implementation, each mold should have an RFID tag or a two-dimensional code, and its identity data should be automatically scanned and collected at the production and detection links. Maintenance record data: electronic files recording each maintenance and repair of the mold. Including: maintenance time, maintenance personnel, maintenance operations taken (such as "grinding the upper die edge", "replacing the guide sleeve"), replaced parts, fault description found (such as "edge crack 3mm"), etc. These data usually come from the terminal input of maintenance personnel or are integrated with CMMS (computerized maintenance management system). Time sequence correlation: this is the key to building the knowledge base. In implementation, the system needs to establish a unified time axis: when a stamping part is produced, record its timestamp and the mold ID used. When this stamping part is detected, its burr data (including the burr distribution atlas) is bound with the mold ID and timestamp used to produce it. The maintenance record of the mold also has its timestamp. Through the mold ID and timestamp, the system can query "what changes have occurred in the burr of all workpieces produced by the mold before and after a certain maintenance".

[0030] Further, the corresponding relationship between the burr distribution atlas and the die wear type or the die failure type is obtained based on the time sequence correlation between the burr data and the maintenance record data, and includes the following steps: Based on the maintenance time in the maintenance record data, a set of stamping parts produced before and after the maintenance is determined; The burr distribution atlas corresponding to each stamping part in the set of stamping parts produced before and after the maintenance is compared and analyzed; It is judged whether the occurrence frequency or the severity of the burr distribution corresponding to any burr distribution atlas in the set of stamping parts produced after the maintenance is reduced; If yes, the corresponding relationship between the burr distribution atlas corresponding to the burr distribution mode and the die wear type or the die failure type solved by the current maintenance operation is established.

[0031] Specifically, the set of stamping parts is determined: according to the time point T_maintenance of a maintenance event in the maintenance record, the system automatically performs a query: the set before maintenance: all stamping parts produced by the die in the time period [T_maintenance-ΔT, T_maintenance). The set after maintenance: all stamping parts produced by the die in the time period [T_maintenance, T_maintenance +ΔT]. ΔT is a preset time window, which can be adjusted according to the production rhythm. The burr distribution atlas is compared and analyzed: the system respectively performs cluster analysis or pattern matching on the burr distribution atlas of all workpieces in the two sets before and after the maintenance, and counts the occurrence of each repeatedly occurring burr distribution mode (i.e. the first burr distribution mode). It is judged whether the mode is improved: for each burr distribution mode that occurs before the maintenance, it is checked whether the occurrence frequency or the severity is significantly reduced after the maintenance. The corresponding relationship is established: if a certain mode (for example, “long burr on the edge and in the middle”) almost disappears after the maintenance, and the maintenance record shows that the operation of this maintenance is “replace the broken pressure plate spring”, then the system can establish a corresponding relationship: the burr distribution atlas (long burr on the edge and in the middle) corresponds to the defect reason (die failure type: pressure plate spring fracture).

[0032] The embodiment realizes automatic knowledge base construction: manual and subjective input of failure mode corresponding relationship is avoided, and instead, the corresponding relationship is automatically mined and verified from historical data in a data-driven manner, so as to ensure the objectivity and accuracy of the knowledge base. Moreover, the knowledge base can be continuously evolved: as production data is continuously accumulated, the system can continuously discover new failure modes and enrich the mode library, so that the system becomes more and more perfect and intelligent.

[0033] Further, the judging whether the occurrence frequency or the severity of the burr distribution corresponding to any of the burr distribution patterns is reduced in the set of the stamping parts produced after the maintenance comprises the following steps: calculating a first occurrence frequency and a first average severity of the first burr distribution pattern in the set of the stamping parts before the maintenance; calculating a second occurrence frequency and a second average severity of the first burr distribution pattern in the set of the stamping parts after the maintenance; if a drop ratio of the second occurrence frequency relative to the first occurrence frequency is greater than a first preset threshold, and / or a drop ratio of the second average severity relative to the first average severity is greater than a second preset threshold, it is determined that the occurrence frequency or the severity of the burr distribution corresponding to any of the burr distribution patterns is reduced.

[0034] Specifically, the indicators before the maintenance are calculated: for a certain burr distribution pattern (the first burr distribution pattern) to be investigated, the first occurrence frequency = (the number of the workpieces of the pattern) / (the total number of the workpieces in the set before the maintenance), and the first average severity: the severity of the pattern on each workpiece needs to be quantified (for example, the total area of the burr under the pattern, the maximum height, etc. can be used as the indicators). Then the average severity of all the workpieces of the pattern is calculated. The indicators after the maintenance are calculated: the second occurrence frequency and the second average severity of the same pattern in the set after the maintenance are calculated. The thresholds are set and judged: the first preset threshold (such as 30%) and the second preset threshold (such as 25%) are set. The frequency drop ratio is calculated: (the first occurrence frequency - the second occurrence frequency) / the first occurrence frequency, and the severity drop ratio is calculated: (the first average severity - the second average severity) / the first average severity. If the frequency drop ratio > 30%, or the severity drop ratio > 25%, or both ratios exceed the respective thresholds, the system automatically determines that the burr distribution pattern is effectively improved after the maintenance.

[0035] In a preferred embodiment, the obtaining of the preset angle images of the stamping part to be detected under multiple different preset viewing angles comprises the following steps: obtaining the type of the stamping part to be detected, as a target stamping part type; calling and traversing a collection angle database to obtain a preset viewing angle set corresponding to the target stamping part type, the preset viewing angle set comprising multiple preset viewing angles; the collection angle database comprising multiple stamping part types and the preset viewing angle set corresponding to each stamping part type; The control servo motor drives the rotating disc and the stamping part fixed thereon to rotate to each preset viewing angle, and after stopping at each angle, triggers the CCD detection camera to capture an image, which is taken as the preset angle image at the preset viewing angle.

[0036] Specifically, the stamping part type of the stamping part to be detected refers to the category or model of the stamping part, for example, "left front door inner plate", "engine cover outer plate", "B column reinforcing plate", etc. Different types of workpieces have completely different geometric shapes, edge directions and positions prone to burrs. In the detection station, the workpiece can be preliminarily classified by scanning the two-dimensional code / RFID tag attached to the workpiece or using a visual recognition algorithm, so as to automatically obtain the target stamping part type. Collection angle database: this is a core database, which stores the mapping relationship between various stamping part types and their corresponding preset viewing angle sets. Implementation: this database can be a simple table structure, for example, a SQL database table, containing two columns: stamping part type and optimal angle set, such as [0, 45, 90, 135, 180, 225, 270, 315]. Preset viewing angle set: for a specific type of stamping part, a set of camera observation angles (usually in degrees) that can most effectively capture all potential burr features of the stamping part after optimization calculation. Exemplarily, for "left front door inner plate", its preset viewing angle set can be [0°, 15°, 30°, 90°, 180°, 270°], which ensures that the burrs of the complex flanging and bending parts can be clearly visible at least at one angle. Servo motor and rotating disc: constitute a high-precision rotary positioning system. Implementation: the servo motor receives angle instructions from the upper computer and drives the rotating disc connected on its axis to rotate. The stamping part to be detected is firmly installed in the center of the rotating disc through a customized fixture, ensuring that its rotation axis coincides with the rotating disc axis to avoid large deviation during shooting. The CCD detection camera is a high-resolution and high-sensitivity industrial area array camera, which is fixedly installed directly above the rotating disc on one side, and its field of view covers the workpiece on the rotating disc. The lighting system (such as ring-shaped LED lamp, coaxial light, etc.) needs to ensure that the illumination of the workpiece edge is uniform and has no obvious reflection at any angle.

[0037] The present embodiment avoids the massive redundant images and data processing pressure brought by 360° indiscriminate full-angle collection, and only collects a few most valuable angles, greatly improving the detection efficiency. The best viewing angle "tailored" for different workpieces ensures that even on the most complex geometric structure, the key edge can be covered by at least one high-quality angle, laying a solid foundation for subsequent high-precision detection. The whole process does not need manual intervention to adjust the angle, and the system automatically adapts to different types of workpieces, which is very suitable for mixed production, multi-variety and small-batch modern flexible manufacturing scenes.

[0038] Further, the method further comprises constructing the acquisition angle database; The constructing the acquisition angle database comprises the following steps: Selecting a plurality of test samples corresponding to the stamping part type, ensuring that the test samples cover the common burr states of the stamping part type; Fixing a test sample and controlling it to rotate uniformly around an axis, while continuously shooting by the CCD detection camera to obtain a continuous video stream of the test sample for one rotation; Frame processing the continuous video stream to obtain a plurality of high-definition images corresponding to uniformly distributed angles, constituting a full-view image set of the test sample; Obtaining a pixel-level segmentation mask of the burr area in each high-definition image in the full-view image set; Based on the segmentation mask, calculating the average contrast of the burr area in each high-definition image; Selecting the high-definition images with an average contrast higher than a third preset threshold value, and taking the rotation angle corresponding to the high-definition images as the effective candidate view angle of the test sample; Based on the effective candidate view angle of each test sample corresponding to the stamping part type, obtaining the preset view angle set of the stamping part type; Cycling the above steps to obtain the preset view angle set corresponding to all stamping part types respectively, to obtain the acquisition angle database.

[0039] Specifically, test sample: refers to the physical sample of the stamping part used for the visual angle optimization experiment. A batch of samples that can represent all common burr states of this type of workpiece need to be carefully selected, for example: both good products without burrs and products with small, medium and severe burrs, as well as samples with burrs in different positions. The more the number of samples and the more comprehensive the coverage, the more reliable the final angle set. Continuous video stream and frame processing: fix a test sample on a turntable, set the turntable to rotate continuously at a constant speed (such as 1° / s). At the same time, set the CCD camera to continuously shoot at a high frame rate (such as 30 fps) to obtain a video of the workpiece rotating one revolution. Then, use a video frame extraction algorithm (such as extracting frames at fixed rotation angle intervals) to decompose the video into a series of high-definition images, for example, extract a frame every 2° of rotation, and finally obtain 180 images, which constitute the full-view image set of the sample. Pixel-level segmentation mask: a binary image in which white pixels represent pixels belonging to the burr area and black pixels represent the background. Obtaining the mask is a prerequisite for calculating the contrast. This can be achieved in several ways: manual annotation: manually outline the burr area in each image using image annotation tools by a professional. It has the highest accuracy but is time-consuming. Semi-automatic algorithm: use traditional image segmentation algorithms (such as threshold segmentation, edge detection) or pre-trained machine learning models (such as U-Net) for preliminary segmentation, and then manually check and correct. This is a practical solution that balances efficiency and accuracy. Average contrast: an indicator used to quantify the difference in brightness or color between the burr area and the surrounding background area at a specific angle. The greater the difference, the higher the contrast, and the clearer the burr is visible. The calculation formula is usually: Average contrast = (average gray value of burr area - average gray value of background area) / (average gray value of burr area + average gray value of background area). Using the obtained segmentation mask, the burr and background areas can be easily separated and their average gray values can be calculated. Effective candidate view angle: for a single test sample, all rotation angles corresponding to images with an average contrast higher than a third preset threshold (for example, 0.4) are considered "good angles" for detecting the burr of the sample. Preset view angle set: after statistical screening, a set of robust angles for detecting all workpieces of this type is finally determined based on the effective candidate view angles of all test samples of this type.

[0040] In a preferred embodiment, the preset view angle set of the stamping part type is obtained based on the effective candidate view angles of each test sample corresponding to the stamping part type, comprising the following steps: aggregate the effective candidate view angles corresponding to all test samples of the stamping part type to form an initial set containing a plurality of effective candidate view angles; statistically analyze the frequency of occurrence of each effective candidate view angle in the initial set; Screening the effective candidate view angles with the occurrence frequency higher than the preset frequency threshold as the preset view angles corresponding to the stamping part type, to constitute the preset view angle set of the stamping part type.

[0041] Specifically, the initial set: when constructing the preset view angle set of a certain stamping part type (such as "Type_A"), it is first necessary to summarize the respective effective candidate view angles of all test samples of the type. Exemplarily, it is assumed that 10 test samples are selected for "Type_A", and each sample obtains 10-20 effective candidate view angles. All the effective candidate view angles of the 10 samples are put into an empty set to obtain the initial set. It may contain a large number of repeated angle values (for example, 0°, 5°, 10°, 0°, 355°, 0°, 90°...). Occurrence frequency: refers to the proportion of the number of occurrences of a certain candidate view angle (such as 0°) in the entire initial set to the total number of times. Normalize all angles in the initial set (such as all mapped to the range of 0-360°). Count statistically. For example, the initial set has 500 view angle data, of which 0° (or the interval of 359.5°-0.5°) occurs 150 times. Then the occurrence frequency of the candidate view angle 0° = 150 / 500 = 30%. The preset frequency threshold is used to screen out those view angles that occur frequently enough and are therefore considered to be generally effective. This threshold needs to be set according to the number of samples and requirements. For example, the preset frequency threshold can be set to 15%. This means that an angle must be considered effective in at least 15% of the test sample data for it to be ultimately selected. Screening and composition: traverse each unique candidate view angle in the initial set, and if its calculated occurrence frequency is higher than 15%, select the view angle as a preset view angle of the stamping part type. The set composed of all the selected preset view angles is the final preset view angle set.

[0042] The following describes the implementation of the steps: Data aggregation: the system reads the list of valid candidate viewing angles for all test samples under the model "Type_A" in the database, and combines them to form a large list. Divide the 360-degree range into several small intervals, for example, one interval per degree. Traverse the large list and classify each angle value into the corresponding small interval. Count the number of angles in each interval. Calculate the frequency: for each angle interval, such as 0°-1°, calculate its frequency: frequency = (number of angles in the interval) / (total length of the large list). Threshold screening: compare the frequency of each interval with the preset frequency threshold (such as 15%). Keep the central angles represented by all intervals with a frequency greater than 15% (such as 0.5°, 90.5°, 180.5°, 270.5°). Set output: round the remaining angle values to the nearest whole degree and remove duplicates, and finally get the preset viewing angle set of "Type_A", for example {0, 90, 180, 270}.

[0043] In a preferred embodiment, the respective calculation of the local image feature saliency of the physical location point in each of the target region images comprises the following steps: Respectively extracting multi-dimensional image features corresponding to the region where the physical location point is located in each of the target region images; Fusion calculation of the multi-dimensional image features to generate a quantitative score value representing the visual saliency of the physical location point in each of the target region images; According to the quantitative score value of the visual saliency of the physical location point in each of the target region images, the local image feature saliency in each of the target region images is obtained.

[0044] Specifically, multi-dimensional image features refer to numerical vectors extracted from images that can describe different attributes (such as shape, texture, etc.) of local regions of images. Common dimensions include: gradient features: describe edge intensity and direction. Texture features: describe the roughness, regularity, etc. of the region. The higher the value of the quantitative score value, the more clear and prominent the visual features of the local region where the current physical location point is located in the specific target region image, and the more conducive to burr identification and judgment.

[0045] The following describes the implementation of the steps: Input: for a physical location point P_i on the edge segment to be tested, and an image Img_j in the target region image sequence. Local image block extraction: in the image Img_j, take the pixel coordinates of the point P_i as the center, and extract a fixed size (such as 32 32pixel) square image patch Patch_ij. Multi-dimensional feature extraction: compute multi-dimensional image features for the small image patch Patch_ij, and input the multi-dimensional image feature vectors into a pre-defined fusion model (e.g. a weighted summation formula), and the model outputs a final quantized score value Score_ij. This value is the local image feature saliency of point P_i in image Img_j. Repeat the steps for each image Img_j in the sequence, and obtain a set of scores [Score_i1, Score_i2,..., Score_iN].

[0046] Further, the respective extraction of the multi-dimensional image features corresponding to the region where the physical location point is located in each of the target region images comprises the following steps: For each of the target region images, a local image patch of a predetermined size is extracted centered at the physical location point; A gradient magnitude map and a gradient direction histogram are calculated as gradient features of the local image patch; A local binary pattern feature is extracted by an LBP operator as a texture feature of the local image patch; The gradient features and the texture features are taken as the multi-dimensional image features.

[0047] Specifically, the local image patch is a small region image centered at the physical location point and is a basic unit for feature calculation. Implementation: The size needs to be balanced. Too small may not contain enough information, and too large may introduce too much irrelevant background interference. Usually, 16 16, 32 32 or 64 64 pixels. The gradient magnitude map describes the edge intensity at each pixel point in the image. Implementation: The Sobel operator is used to perform convolution in the X and Y directions respectively to obtain gradient components Gx and Gy. The gradient magnitude of each pixel point is calculated according to the formula: Magnitude = sqrt(Gx 2 + Gy 2 ), and the calculated magnitude map is the gradient feature. The burr usually appears as a prominent edge, so it will produce a higher gradient magnitude at its location. The gradient direction histogram describes the statistical characteristics of the edge direction distribution within the image patch. Implementation: Based on the calculation of Gx and Gy, the direction of each pixel point is calculated: Theta = arctan2(Gy, Gx). Then the 0-360° direction range is divided into several intervals (bins), and the number of gradient directions of all pixels in the image patch falling into each interval is counted to form a histogram. This variant of HOG (Histogram of Oriented Gradients) is also a gradient feature and has a certain effect on describing the morphology of the burr.

[0048] Specifically, LBP operator: Local Binary Pattern, is a powerful texture descriptor. For each pixel in the image block, it compares it with its 3 8 pixels in the 3 neighborhood. If the neighborhood pixel value is greater than the center pixel value, mark the position as 1, otherwise 0. In this way, an 8-bit binary number (usually read in clockwise order) is obtained, which is converted to decimal number, that is, the LBP code of the center pixel. Finally, calculate the histogram of all pixel LBP codes in the entire image block, which is the texture feature of the image block. The surface texture of burr area is usually different from the smooth base material, and the LBP feature can effectively capture this difference.

[0049] The features of the embodiment are complementary: the gradient feature is sensitive to edges and shapes, and can effectively capture the "convex" shape of the burr; the texture feature is sensitive to surface material changes, and can effectively distinguish the rough surface of the burr from the smooth surface of the workpiece. Combined with both, almost all the core visual properties of the burr are covered. High computational efficiency: Sobel and LBP are both classic operators with light computation and extremely high efficiency, which are very suitable for large-scale use in industrial detection scenes that require real-time performance, ensuring the high-speed operation of the system.

[0050] The principles and implementations of the present application are described in this paper by applying specific examples, and the above examples are only used to help understand the method and its core idea. The above description is only the preferred embodiment of the present application. It should be pointed out that due to the limited nature of the language expression, there are objectively infinite specific structures, and for ordinary technical personnel in this technical field, without departing from the principles of the present application, a number of improvements, refinements or changes can be made, or the above technical features can be combined in an appropriate way; these improvements, refinements, changes or combinations, or the application of the inventive concept and technical solution to other occasions without improvement, shall be regarded as the protection scope of the present application.

Claims

1. A method for burr data acquisition and image processing applicable to automotive stamping parts, characterized in that, Includes the following steps: Acquire images of the stamping part to be inspected at preset angles from multiple preset viewpoints; The target region image of the edge segment to be tested of the stamping part to be tested is extracted from each of the preset angle images, and a target region image sequence is generated; wherein, the edge segment to be tested is formed by connecting a number of physical location points; For any physical location point on the edge segment to be tested, traverse all target region images in the target region image sequence and calculate the local image feature saliency of the physical location point in each target region image; The target region image with the highest local image feature saliency is selected as the source image for spur identification of the physical location point; Based on the analysis source image assigned to each physical location point on the edge segment to be tested, the burr region on the edge segment to be tested is identified; The geometric dimensions of the burr region are quantized, and the burr data results are output.

2. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 1, characterized in that: After identifying the burr region on the edge segment under test based on the analysis source image assigned to each physical location point on the edge segment under test, the method further includes the following steps: Based on the size and location information of all identified burr areas, a burr distribution map corresponding to the stamping part to be inspected is generated; The historical defect pattern library is retrieved and traversed to obtain the defect cause corresponding to the burr distribution pattern of the stamping part to be inspected; wherein, the historical defect pattern library includes a variety of burr distribution patterns and their corresponding defect causes, and the defect causes include mold wear type or mold failure type. Based on the defect cause corresponding to the burr distribution pattern of the stamping part to be inspected, a corresponding mold maintenance prompt is generated and sent to the production management terminal.

3. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 2, characterized in that: This method also includes building a library of historical defect patterns; The construction of the historical defect pattern library includes the following steps: The data on burrs in the stamped part is correlated with the mold identity data of the mold used in the production of the stamped part; the burr data includes the burr distribution map. The mold identification data associated with the mold used is linked to its maintenance record data; Based on the temporal correlation between the burr data and the maintenance record data, the correspondence between the burr distribution map and the mold wear type or the mold failure type is obtained, so as to obtain the historical defect pattern library.

4. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 3, characterized in that: The method for obtaining the correspondence between the burr distribution map and the mold wear type or the mold failure type based on the time-series correlation between the burr data and the maintenance record data includes the following steps: Based on the maintenance time in the maintenance record data, determine the set of stamped parts produced before and after the maintenance; Comparative analysis of the burr distribution patterns of each stamped part in the stamped parts set produced before and after maintenance; Determine whether the frequency or severity of any burr distribution pattern corresponding to the burr distribution pattern in the stamped parts produced after repair has decreased in the set. If so, then establish the correspondence between the burr distribution pattern corresponding to the burr distribution map and the mold wear type or mold fault type resolved in this maintenance operation.

5. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 4, characterized in that: The determination of whether the frequency or severity of any burr distribution pattern corresponding to the burr distribution pattern in the stamped parts produced after repair has decreased includes the following steps: Calculate the first occurrence frequency and first average severity of the first burr distribution pattern in the stamped parts set before repair; Calculate the second occurrence frequency and second average severity of the first burr distribution pattern in the set of stamped parts after repair; If the rate of decrease of the second occurrence frequency relative to the first occurrence frequency is greater than the first preset threshold, and / or the rate of decrease of the second average severity relative to the first average severity is greater than the second preset threshold, then it is determined that the occurrence frequency or severity of any of the burr distribution patterns corresponding to the burr distribution patterns has decreased.

6. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 1, characterized in that: The process of acquiring images of the stamping part to be inspected at preset angles from multiple preset viewpoints includes the following steps: Obtain the stamping type of the stamping part to be tested, and use it as the target stamping type; The acquisition angle database is retrieved and traversed to obtain a set of preset viewing angles corresponding to the target stamping part type. The set of preset viewing angles includes multiple preset viewing angles. The acquisition angle database includes multiple stamping part types and the set of preset viewing angles corresponding to each stamping part type. The servo motor drives the rotating disk and the stamping part to be inspected fixed on it to rotate to each preset viewing angle. After pausing at each angle, the CCD inspection camera is triggered to acquire an image, which is used as the preset angle image under the preset viewing angle.

7. The method for burr data acquisition and image processing applicable to automotive stamping parts according to claim 6, characterized in that: The method also includes constructing the acquisition angle database; The construction of the acquisition angle database includes the following steps: Select multiple test samples corresponding to the stamping part type to ensure that the test samples cover the common burr conditions of the stamping part type; A test sample is fixed and controlled to rotate at a constant speed around an axis, while the CCD detection camera continuously captures images to obtain a continuous video stream of one revolution of the test sample. The continuous video stream is processed by frame segmentation to obtain multiple high-definition images corresponding to uniformly distributed angles, which constitute the full-view image set of the test sample; Obtain the pixel-level segmentation mask of the spiky region in each of the high-definition images in the full-view image set; Based on the segmentation mask, the average contrast of the spiky regions in each of the high-definition images is calculated; The high-definition images with an average contrast higher than a third preset threshold are selected, and their corresponding rotation angles are used as the effective candidate viewpoints of the test samples. Based on the effective candidate viewpoints of each test sample corresponding to the stamping part type, the preset viewpoint set for the stamping part type is obtained; Repeat the above steps to obtain the preset viewing angle set corresponding to each of the stamping part types, so as to obtain the acquisition angle database.

8. The method for burr data acquisition and image processing of automotive stamping parts according to claim 7, characterized in that: The process of obtaining the preset view set for each test sample corresponding to the stamping part type based on the effective candidate viewpoints includes the following steps: Summarize the effective candidate viewpoints corresponding to all test samples under the aforementioned stamping part type to form an initial set containing multiple effective candidate viewpoints; The frequency of occurrence of each of the effective candidate viewpoints in the initial set is counted; The valid candidate viewpoints that appear more frequently than a preset frequency threshold are selected as the preset viewpoints corresponding to the stamping part type, so as to form the preset viewpoint set for the stamping part type.

9. The method for burr data acquisition and image processing of automotive stamping parts according to claim 1, characterized in that: The step of calculating the saliency of local image features of the physical location point in each of the target region images includes the following steps: Extract multi-dimensional image features from the image of each target region corresponding to the region where the physical location point is located; The multi-dimensional image features are fused and calculated to generate a quantitative score value that represents the visual salience of the physical location point in each of the target region images; The saliency of local image features in each target region image is obtained based on the quantitative score of the visual saliency of the physical location point in each target region image.

10. The method for burr data acquisition and image processing of automotive stamping parts according to claim 9, characterized in that: The step of extracting multi-dimensional image features from each target region image corresponding to the region where the physical location point is located includes the following steps: For each target region image, a local image patch of a predetermined size is extracted with the physical location point as the center; Calculate the gradient magnitude map and gradient direction histogram as the gradient features of the local image patch; Local binary pattern features are extracted using the LBP operator and used as texture features for the local image patch; The gradient features and the texture features are used as the multi-dimensional image features.

Images