A stamping part surface detection method and welding slag cleaning device

Abstract

The present invention relates to the technical field of stamping part surface detection, and in particular to a stamping part surface detection method and a welding slag cleaning device. The detection method proposed in the present invention is based on the processing of three-dimensional point cloud data, is not affected by changes in ambient light, can more accurately separate defective areas from the background, and can effectively identify welding slag even in complex geometric shapes or deep recesses, and can achieve comprehensive detection of stamping parts with complex shapes, significantly improving the accuracy and efficiency of detection; by presetting a basic threshold and dynamically adjusting it according to the reflection characteristics, it can be ensured that the stamping parts can obtain stable detection and cleaning results under detection conditions in different light environments.

Description

Technical Field

[0001] The invention relates to the technical field of stamping part surface detection, and in particular to a stamping part surface detection method and a welding slag cleaning device. Background Art

[0002] The cylinder block guard is a component installed on an engine or mechanical equipment. It is mainly used to protect the internal mechanical components from the external environment, such as preventing dust, dirt and other impurities from entering. For automobiles, the cylinder block exhaust pipe guard can also help prevent burns during maintenance and avoid fires caused by flammable objects directly contacting the high-temperature exhaust pipe.

[0003] At present, the detection methods for cylinder guards include:

[0004] 1. Non-contact high-precision detection: Traditional methods rely on manual or contact measuring tools, which is not only inefficient, but also difficult to achieve comprehensive and accurate detection of cylinder guard stamping parts with complex shapes. Moreover, welding slag cleaning often requires a lot of manual intervention, which increases labor costs and may lead to inconsistent operations.

[0005] 2. Inspection through machine vision inspection system: In addition to manual inspection, the current existing technology also uses cameras and image processing software to automatically identify welding slag on the surface of the cylinder cover. However, this method is sensitive to light conditions. Changes in ambient light may affect the detection results, and it is difficult to detect welding slag hidden in complex geometric shapes or deep recesses.

[0006] 3. Detection by ultrasonic testing equipment: Use ultrasonic probe to emit high-frequency sound waves, and analyze the reflected signals to detect defects inside and under the surface of the cylinder shield weld, such as pores and cracks, which indirectly reflects the presence of welding slag. This method is mainly used for internal defect detection of cylinder shields, and has limited effect on direct detection of surface welding slag.

[0007] Therefore, a method for detecting the surface of a stamping part and a welding slag cleaning device that can solve the above-mentioned problems are needed. Summary of the invention

[0008] The present invention provides a method for detecting the surface of a stamping part and a device for cleaning welding slag. The detection method proposed in the present invention is based on the processing of three-dimensional point cloud data, is not affected by changes in ambient light, can more accurately separate defective areas from the background, and can effectively identify welding slag even in complex geometric shapes or deep recesses, thereby enabling comprehensive detection of stamping parts with complex shapes, significantly improving the accuracy and efficiency of detection. By presetting a basic threshold and dynamically adjusting it according to the reflection characteristics, it can be ensured that the stamping parts can obtain stable detection and cleaning results under detection conditions in different light environments.

[0009] The technical solution adopted by the present invention to solve the above technical problems is:

[0010] A method for detecting the surface of a stamping part comprises the following steps:

[0011] Step a. Before the stamping part is conveyed to the stamping part conveying station, the stamping part is three-dimensionally scanned using laser scanners arranged at different angles, wherein each laser scanner emits a laser beam and receives a reflected laser signal;

[0012] Step b. The computer constructs three-dimensional point cloud data of the stamping part according to the received laser signal; performs preliminary filtering on the three-dimensional point cloud data to remove noise points; and forms a point cloud map by calculating the coordinates and reflection intensity of each point;

[0013] Step c. extracting key geometric features of the stamping part from the three-dimensional point cloud data, wherein the key geometric features include edge features and curvature mutation points of the stamping part;

[0014] Step d. Separating the defect area from the background in the point cloud image to form a binary defect area image; performing morphological processing on the separated defect area;

[0015] Step e. Performing a connected domain analysis on the defective area after morphological processing to obtain defect location information of the stamping part;

[0016] Step f. According to the defect position information of the stamped parts, the stamped parts are transferred from the stamped parts conveying platform to the working area of ​​the ultrasonic vibration mechanism by the loading robot arm, and the welding slag on the stamped parts is shaken off by the ultrasonic vibration mechanism.

[0017] Step g. After the stamping parts have been processed by the ultrasonic vibration mechanism, the loading and unloading robots place the stamping parts in a sandblasting machine for automatic sandblasting.

[0018] Furthermore, in step c, the key geometric features of the stamping part are extracted from the three-dimensional point cloud data, and the key geometric features include the edge features and curvature mutation points of the stamping part, including:

[0019] Step c1: For point P in the 3D point cloud data i , calculate the mean curvature H and Gaussian curvature K in its local neighborhood;

[0020] Step c2: For each point P i , calculate its point P with the neighborhood j The curvature difference between , the calculation formula is , where C represents the curvature and mean is the average value of the curvature in the neighborhood;

[0021] Step c3: setting a threshold T to distinguish between normal curvature changes and sudden changes in curvature;

[0022] Step c4: For all >Point P of T i , marking it as the curvature mutation point;

[0023] Step c5: Use the sliding window method or the K-nearest neighbor method to determine the i Neighborhood of;

[0024] Step c6: Apply the Canny edge detection algorithm to identify edge points based on the curvature change;

[0025] Step c7: Use an edge tracking algorithm to connect isolated edge points into continuous edge lines to form edge features of the stamping part.

[0026] Furthermore, in step c1, for point P in the three-dimensional point cloud data i , calculating the average curvature H and Gaussian curvature K in its local neighborhood includes:

[0027] Step c1-1: For point P in the 3D point cloud data i , select its k nearest neighbors as the local neighborhood, and use the octree spatial index structure to accelerate the nearest neighbor search;

[0028] Step c1-2: Use the least squares fitting method to fit P i The local neighborhood point P j Fit a plane or quadratic surface and extract the principal curvatures k1 and k2;

[0029] Step c1-3: By calculating each point P in the local neighborhood j The average curvature H is obtained by taking the average of the principal curvatures, and the calculation formula is: , where k1 is the maximum principal curvature, indicating that at this point P i The maximum curvature of the surface along a certain direction at point P; k2 is the minimum principal curvature, indicating that at this point P i The minimum curvature of the surface in another direction;

[0030] Step c1-4: By calculating each point P in the local neighborhood j The Gaussian curvature K is obtained by multiplying the principal curvatures of .

[0031] Furthermore, the performing of morphological processing on the separated defective region in step d includes: performing a morphological opening operation on the separated defective region to remove isolated small noise points, and then performing a closing operation to fill small holes in the defective region to enhance connectivity of the defective region;

[0032] Furthermore, before separating the defective area in the point cloud image from the background in step d, a basic threshold is preset according to the reflection characteristics of the stamping material, and the basic threshold is adjusted according to changes in the ambient light intensity and the distribution of the point cloud reflection intensity in the three-dimensional point cloud data to ensure accurate separation of subsequent defective areas.

[0033] Furthermore, the connected domain analysis of the defective area after morphological processing in step e includes: determining the specific position and size of each defect on the stamping part; calculating the area and perimeter parameters of each connected domain, and recording its position information, and visually marking the identified defects on the display interface, using red contour lines to mark the defective area; generating a test report containing defect location and size information.

[0034] Furthermore, the number of the laser scanners is more than three, and the laser scanners are evenly distributed around the stamping part.

[0035] Furthermore, in step f, the ultrasonic vibration mechanism emits high-frequency vibrations through the ultrasonic transducer to loosen and fall off the welding slag attached to the surface of the stamping part; the state change of the defective area is monitored by the ultrasonic vibration displacement sensor to confirm whether the welding slag is effectively removed; if there is still uncleaned welding slag on the stamping part after ultrasonic treatment, repeat step f until the welding slag is completely removed, and the stamping part after slag removal is transported to the next workstation by the unloading robot arm.

[0036] A welding slag cleaning device based on a stamping part surface detection method comprises a stamping part conveying platform and a workbench, wherein a loading robot arm, an ultrasonic vibration mechanism and a unloading robot arm are installed on the workbench, and the stamping part conveying platform is used to convey the stamping parts to the bottom of the loading robot arm.

[0037] Furthermore, the ultrasonic vibration mechanism includes an ultrasonic transducer and an ultrasonic vibration displacement sensor.

[0038] The advantages of the present invention are:

[0039] 1. The detection method proposed in the present invention is based on the processing of three-dimensional point cloud data, is not affected by changes in ambient light, can more accurately separate defective areas from the background, and can effectively identify welding slag even in complex geometric shapes or deep recesses. It can achieve comprehensive detection of stamping parts with complex shapes, significantly improving the accuracy and efficiency of detection. By presetting the basic threshold and dynamically adjusting it according to the reflection characteristics, it can ensure that the stamping parts can obtain stable detection and cleaning results under the detection conditions of different light environments.

[0040] 2. The present invention realizes the automation of the entire process from automatic scanning of the stamping part surface, defect identification to automatic removal of welding slag. The stamping part is transferred to the working area of ​​the ultrasonic vibration mechanism by the loading robot arm, and the ultrasonic transducer is used to emit high-frequency vibration to loosen and remove the welding slag, and finally transported to the next workstation by the unloading robot arm. This highly automated process significantly reduces the manpower demand, ensures the consistency and reliability of the operation, and thus improves the production efficiency and product quality of stamping parts.

[0041] 3. The present invention uses laser scanners arranged at multiple angles for three-dimensional scanning, and combines computer processing to generate detailed three-dimensional point cloud data. This method not only achieves non-contact high-precision detection, but also can adapt to various complex geometric shapes, ensuring the comprehensiveness and accuracy of the detection results. At the same time, by performing preliminary filtering on the three-dimensional point cloud data to remove noise points, the quality of subsequent analysis is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 The figure is a schematic flow chart of a method for detecting the surface of a stamping part according to the present invention.

[0044] Figure 2 A schematic structural diagram of a welding slag cleaning device based on a stamping part surface detection method provided by the present invention.

[0045] Figure 3 A schematic diagram of the layout of a sandblasting machine based on a stamping part surface detection method provided by the present invention.

[0046] in:

[0047] 1. Stamping parts conveyor; 2. Stamping parts; 3. Loading robot arm;

[0048] 4. Ultrasonic vibration mechanism; 5. Unloading robot arm; 6. Workbench;

[0049] 7. Feeding line; 8. Laser marking mechanism; 9. Loading and unloading robot;

[0050] 10. Automatic positioning machine; 11. Sandblasting machine; 12. Visual inspection mechanism;

[0051] 13. Packing robot. DETAILED DESCRIPTION

[0052] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Embodiment 1: At present, in the field of cylinder shield detection and welding slag cleaning, although the non-contact high-precision detection method can avoid the limitations of contact measurement, it still has difficulties in processing cylinder shield stamping parts with complex shapes. This method is difficult to achieve comprehensive and accurate detection, especially for the identification of welding slag in hidden areas or parts with complex geometric shapes. In addition, the welding slag cleaning process relies heavily on manual intervention, which not only increases labor costs, but also may lead to inconsistent operations and affect the quality of the final product. Although the application of machine vision inspection systems has improved the degree of automation, its performance is seriously dependent on ambient light conditions. Changes in illumination may lead to unstable detection results, especially in industrial environments where light conditions are difficult to maintain constant. More importantly, this method is difficult to effectively detect welding slag located in complex geometries or deep recesses, which are often high-incidence areas of welding slag residue. Although ultrasonic detection equipment can detect internal defects, it has limited direct detection effects on surface welding slag and is difficult to meet the needs of surface quality control of cylinder shields. Specifically, these problems are particularly prominent on a production line for automobile engine cylinder shields with a high degree of automation. Assume that the production line needs to process 100 cylinder shields per hour, and each shield needs to go through welding, inspection, cleaning and other processes. Due to the complex shape of the cylinder shield, which contains multiple curved surfaces and grooves, traditional machine vision inspection systems often miss and misdetect during inspection. For example, in a groove with a depth of more than 5mm, the detection accuracy may drop below 70%. Although manual inspection can make up for this deficiency, it is slow and the inspection time for each shield may take 3-5 minutes, which seriously affects production efficiency. Even more difficult is the welding slag cleaning link. Existing automated cleaning equipment is difficult to adapt to welding slag of different positions and sizes, and the cleaning effect is unstable. Sometimes, some tiny welding slag (less than 0.5mm in diameter) may be missed, and these seemingly insignificant welding slag may fall off during the long-term operation of the engine, causing serious mechanical failures. Although manual cleaning can improve the cleaning quality, it is inefficient and prone to omissions due to operator fatigue.

[0054] In view of the above problems, the present invention has explored a variety of possible technical solutions by analyzing these technical difficulties. In view of the challenge that non-contact detection methods are difficult to handle with complex shapes, the present invention's solutions include: single camera 3D reconstruction, binocular vision reconstruction and laser scanning reconstruction. After comparative analysis, laser scanning technology stands out due to its high precision and strong anti-interference ability.

[0055] Specifically, the present invention uses a laser scanner arranged at multiple angles to obtain complete three-dimensional information of stamped parts (cylinder shields) from different angles. By emitting and receiving laser signals, the computer can construct dense point cloud data, which provides an accurate geometric basis for subsequent defect identification. In order to solve the problem that traditional methods are difficult to identify hidden defects, the present invention innovatively proposes a feature extraction method based on curvature analysis. By calculating the local curvature changes of each point in the point cloud data, edge features and curvature mutation points can be accurately located. This method can not only identify subtle defects that are easily overlooked in conventional visual inspections, but also effectively distinguish normal geometric changes from abnormal defect areas.

[0056] In terms of defect area processing, the present invention designs a systematic detection scheme including background separation, morphological processing and connected domain analysis. By dynamically adjusting the threshold and performing morphological opening and closing operations, noise points can be effectively removed and the connectivity of the defect area can be enhanced. Finally, through precise position marking and report generation, visualization and quantitative analysis of defects are achieved.

[0057] In order to solve the key problem of welding slag cleaning, the present invention introduces an ultrasonic vibration-dropping mechanism. The mechanism loosens and drops the welding slag through high-frequency vibration, and monitors the cleaning effect in real time through a displacement sensor to ensure that the surface welding slag is completely removed. If it is not completely removed, the detection system will automatically repeat the process to ensure the integrity of the cleaning. The core innovation of the present invention is to propose a stamping part surface detection method that combines laser three-dimensional scanning and ultrasonic vibration. This method obtains accurate three-dimensional point cloud data of stamping parts through multi-angle laser scanning, extracts key geometric features using curvature analysis, and accurately identifies defective areas through morphological processing and connected domain analysis. Finally, an ultrasonic vibration-dropping mechanism is used to clean the welding slag from the detected defective area, realizing the full process automation from detection to cleaning.

[0058] Figure 1 It is a flow chart of the method steps of the present invention, such as Figure 1As shown, a method for detecting the surface of a stamping part provided by the present invention is specifically divided into the following key steps: First, before the stamping part is conveyed to the conveyor table, a plurality of laser scanners arranged at different angles are used to perform a three-dimensional scan on the stamping part. Each laser scanner emits a laser beam and receives the reflected laser signal. This multi-angle scanning method can effectively avoid data loss caused by occlusion and ensure that complete information on the surface of the stamping part is obtained. The number of laser scanners is more than three, and the laser scanners are evenly distributed around the stamping part. The scanning angle of each laser scanner can be dynamically adjusted according to the specific shape and size of the stamping part to ensure that the entire surface area is covered. In the specific implementation, the number and arrangement of laser scanners are crucial. By evenly distributing more than three laser scanners around the stamping part, the surface details of the stamping part can be captured from different angles. This multi-angle scanning method can overcome the blind spot problem of single-angle scanning and ensure comprehensive detection of the surface of the stamping part. In practical applications, the arrangement of laser scanners can present a variety of geometric configurations. For example, a triangular distribution method can be adopted to ensure that the angle between each scanner is between 30 and 60 degrees. Each scanner emits a laser beam and receives the reflected signal, and the computer performs 3D point cloud data reconstruction and defect analysis. In specific implementation, three or four laser scanners can be selected, and their installation positions are on fixed brackets. The distance between the scanners can be adjusted according to the size and shape of the stamped parts (usually a spacing of 15 to 25 cm is maintained to ensure good overlapping coverage). For complex stamped parts such as cylinder shields, this multi-angle scanning method can effectively overcome the limitations of traditional detection methods. Compared with single-angle or manual visual inspection, multi-angle laser scanning can capture more surface details and improve the accuracy and reliability of defect detection. Practice has shown that this multi-angle laser scanning technology has significant advantages in stamping surface inspection. Through evenly distributed laser scanners, the detection system's ability to identify surface defects of stamping parts with complex geometries can be significantly improved, providing more accurate technical means for industrial quality control.

[0059] After scanning the stamped parts, the computer constructs the 3D point cloud data of the stamped parts according to the received laser signal. The point cloud map is formed by calculating the coordinates and reflection intensity of each point. The 3D point cloud data is initially filtered to remove noise points and improve the accuracy of subsequent processing. Then, the key geometric features of the stamped parts, including edge features and curvature mutation points, are extracted from the 3D point cloud data. This step is crucial for identifying defects, because defects such as welding slag usually cause abnormal changes in local geometric features. After that, the defect area in the point cloud map is separated from the background to form a binary defect area image. Morphological processing, such as opening and closing operations, is performed on the separated defective areas to remove isolated small noise points and enhance the connectivity of the defective areas. Subsequently, the connected domain analysis is performed on the defective areas after morphological processing to obtain the defect location information of the stamped parts. This step can accurately locate the defects and provide accurate location guidance for subsequent welding slag cleaning. Finally, according to the defect location information, the stamped parts are moved from the conveyor table to the working area of ​​the ultrasonic vibration mechanism by the loading robot arm. The ultrasonic vibration mechanism loosens and removes the welding slag through high-frequency vibration, thereby achieving automatic cleaning of defects.

[0060] The design of the above detection method fully considers the complexity of stamping surface detection and the difficulty of welding slag cleaning. By combining high-precision three-dimensional scanning technology and efficient ultrasonic cleaning technology, the present invention can realize comprehensive detection and cleaning of complex-shaped stamping parts, significantly improving the efficiency and accuracy of detection and cleaning.

[0061] As a preferred embodiment, the present invention can be applied to the production line of automobile engine cylinder shields. Specifically, after the cylinder shield is stamped, three laser scanners are first used to perform a three-dimensional scan on it. The three laser scanners are arranged above the stamped part, 45 degrees to the left front, and 45 degrees to the right front, respectively. The scanning frequency is 100Hz, and the scanning accuracy can reach 0.1mm. After the scanning is completed, the computer uses a point cloud processing algorithm to construct a three-dimensional point cloud model of the cylinder shield. For a typical cylinder shield, the amount of point cloud data is about 1 million points. Subsequently, a spherical neighborhood with a radius of 2mm is used to calculate the local curvature, and the edge features and curvature mutation points are extracted. The curvature difference threshold T is set to 0.05, and the points exceeding the threshold are marked as potential defect points. Next, the defect area is segmented based on the reflection intensity value. The basic threshold is set to 150 (range 0-255) and dynamically adjusted according to the ambient light intensity. The segmented defective area is opened by a 3x3 pixel structure element, followed by a 5x5 pixel structure element closed. Then, a connected domain analysis is performed, and connected areas with an area greater than 0.5mm² are marked as valid defects. For each defective area, its center coordinates, area, and perimeter information are recorded. Finally, the loading robot moves the cylinder cover to the working area of ​​the ultrasonic vibration mechanism according to the defect location information. The ultrasonic vibration mechanism operates at a frequency of 20kHz and an amplitude of 50μm, and each defective area is processed for 5 seconds. The state changes of the defective area are monitored in real time by a displacement sensor until the welding slag is completely removed.

[0062] Through this implementation mode, the present invention can complete the comprehensive inspection and welding slag cleaning of a cylinder body shield within 15 seconds, thereby greatly improving the production efficiency and product quality.

[0063] In step c, the present invention extracts key geometric features of the stamping part from the three-dimensional point cloud data. The key geometric features include edge features and curvature mutation points of the stamping part.

[0064] Step c1: For point P in the 3D point cloud data i , calculate the mean curvature H and Gaussian curvature K in its local neighborhood.

[0065] Step c2: For each point P i , calculate its point P with the neighborhood j The curvature difference between , the calculation formula is , where C represents the curvature and mean is the average value of the curvature in the neighborhood.

[0066] Step c3: Set a threshold T to distinguish between normal curvature changes and sudden changes in curvature.

[0067] Step c4: For all >Point P of T i , marking it as a curvature mutation point.

[0068] Step c5: Use the sliding window method or the K-nearest neighbor method to determine the i 's neighborhood.

[0069] Step c6: Apply the Canny edge detection algorithm to identify edge points based on the curvature change; specifically, the edge points refer to the locations on the surface of the stamping part where the geometric features change significantly, and these locations are defective areas of the stamping part.

[0070] Step c7: Use an edge tracking algorithm to connect isolated edge points into continuous edge lines to form edge features of the stamping part.

[0071] Specifically, in the surface detection method of stamping parts, when extracting key geometric features from three-dimensional point cloud data, it is necessary to accurately identify edge features and curvature mutation points to facilitate subsequent defect detection and welding slag cleaning. In response to this technical problem, the present invention proposes a specific step for extracting key geometric features of stamping parts from three-dimensional point cloud data. The step first calculates the average curvature and Gaussian curvature in the local neighborhood of each point in the point cloud data, then calculates the curvature difference between the point and the neighborhood, distinguishes normal curvature changes from curvature mutations by setting a threshold, and uses an edge detection algorithm to identify edge points. In specific implementation, firstly, the average curvature H and Gaussian curvature K in the local neighborhood of each point in the three-dimensional point cloud data are calculated. These curvature values ​​reflect the curvature degree of the surface at this point. Then, the curvature difference between each point and its neighborhood points is calculated, and the curvature change is judged by the preset threshold T. When the curvature difference exceeds the threshold, the point is marked as a curvature mutation point. In order to improve the accuracy of feature extraction, a sliding window or K nearest neighbor method is used to determine the neighborhood range of the point. On this basis, the Canny edge detection algorithm is applied to identify edge points, and the edge tracking algorithm is used to connect discrete edge points into continuous edge lines.

[0072] Specifically, the local neighborhood can be set to contain 15 nearest neighbor points, and the curvature difference threshold T is set to 0.05. i , calculate the Euclidean distance between it and the neighboring points, and select the 15 points with the closest distance as its local neighborhood. The calculated average curvature H and Gaussian curvature K are used to characterize the local shape characteristics of the surface. i When the curvature difference with its neighboring points is greater than 0.05, it is marked as a curvature mutation point. The Canny algorithm is used with the low threshold set to 0.1 and the high threshold set to 0.3 to detect edge points.

[0073] In this way, the edge features and curvature mutation points of the stamping surface can be accurately identified, providing reliable geometric feature information for subsequent defect detection. Compared with the traditional edge detection method based on two-dimensional images, this solution makes full use of the spatial information of three-dimensional point cloud data to improve the accuracy and robustness of feature extraction. This feature extraction method can not only effectively identify the edge and curvature mutation of stamping parts, but also provide an important reference for subsequent welding slag detection and cleaning, and improve the reliability of detection.

[0074] In the above step c1, for a point Pi in the three-dimensional point cloud data, calculating the average curvature H and Gaussian curvature K in its local neighborhood includes:

[0075] Step c1-1: For a point Pi in the 3D point cloud data, select its k nearest neighbor points as the local neighborhood, and use the octree spatial index structure to accelerate the nearest neighbor search;

[0076] Step c1-2: Use the least squares fitting method to fit P i The local neighborhood point P j Perform plane or quadratic surface fitting to extract the principal curvatures k1 and k2; (plane fitting can provide normal vector information; surface fitting can provide richer geometric features)

[0077] Step c1-3: By calculating each point P in the local neighborhood j The average curvature H is obtained by taking the average of the principal curvatures, and the calculation formula is: , where k1 is the maximum principal curvature, indicating that at this point P i The maximum curvature of the surface along a certain direction at point P; k2 is the minimum principal curvature, indicating that at this point P i The minimum curvature of the surface along another direction (perpendicular to the k1 direction);

[0078] Step c1-4: By calculating each point P in the local neighborhood j The Gaussian curvature K is obtained by multiplying the principal curvatures of .

[0079] The present invention improves the efficiency and accuracy of curvature calculation by using the octree spatial index structure and least squares fitting through the above method. The octree spatial index structure can quickly locate the nearest neighbor point and reduce the search time. The least squares fitting can more accurately describe the geometric features of the local surface, thereby improving the calculation accuracy of the principal curvature. Specifically, the octree spatial index structure is a three-dimensional space division technology that can recursively divide the three-dimensional space into eight subspaces. When constructing the octree, the point cloud data is organized into a hierarchical structure, so that a large number of irrelevant points can be quickly excluded when searching for the nearest neighbor point, thereby greatly improving the search efficiency. Least squares fitting is a commonly used mathematical optimization technique for finding the best function match for a set of data. In the present invention, it is used to fit the point cloud data of the local neighborhood to obtain a more accurate surface description. By minimizing the sum of squares of the fitting error, the best fitting plane or quadratic surface can be obtained, thereby more accurately calculating the principal curvature.

[0080] Among them, the mean curvature H and Gaussian curvature K are important parameters for describing the local geometric features of the surface. The mean curvature H reflects the average curvature of the surface at that point, while the Gaussian curvature K reflects the intrinsic curvature of the surface at that point. The accurate calculation of these two parameters is crucial for subsequent curvature analysis and feature extraction. Through the above method, the present invention can calculate the curvature information of the surface of the stamping more efficiently and accurately. This not only improves the efficiency of the entire detection process, but also enhances the ability to detect tiny defects on the surface of the stamping. For example, when detecting welding slag or tiny dents, accurate curvature information can help the detection system better distinguish between normal surface changes and abnormal defects. As a preferred embodiment, k=20 can be selected as the number of points in the local neighborhood. This value can provide sufficient local information without introducing too much computational burden (in practical applications, this value can be appropriately adjusted according to the density of the point cloud data and the complexity of the stamping). During the construction of the octree, a minimum voxel size, such as 0.1 mm, can be set as a stopping condition for space division (this value should be determined based on the accuracy of the point cloud data and the size of the stamping part. Too small voxels may increase unnecessary computational complexity, while too large voxels may reduce search accuracy).

[0081] For the least squares fitting, first try to use plane fitting. If the fitting error exceeds the preset threshold (for example, 0.05mm), then switch to quadratic surface fitting. This adaptive fitting strategy can minimize the computational complexity while ensuring accuracy. When calculating the principal curvatures k1 and k2, the eigenvalue decomposition method can also be used. Specifically, for the surface obtained by fitting at point P i The shape operator is calculated at , and then the eigenvalues ​​of the shape operator are solved. These eigenvalues ​​correspond to the principal curvatures.

[0082] Through this specific implementation, the method of the present invention can achieve good results in practical applications. For example, for a typical automobile engine cylinder shield stamping part, its surface area is about 0.5 square meters, and the point cloud data contains about 1 million points. Using the method of the present invention, the curvature calculation of all points can be completed within 5 seconds, while the traditional method requires more than 30 seconds. At the same time, the average error of the calculated curvature value can be controlled within 0.01mm^-1, which is accurate enough for detecting tiny surface defects (such as welding slag with a diameter of 0.5mm). Moreover, the method of the present invention has significant improvements in efficiency and accuracy. The traditional curvature calculation method usually uses a simple differential approximation. Although the calculation speed is fast, the accuracy is low, especially when processing complex surfaces, it is prone to large errors. Some high-precision methods, such as using complex differential geometry algorithms, have high accuracy but high computational complexity and are difficult to apply to real-time detection of welding slag. The method of the present invention combines octree spatial indexing and least squares fitting to greatly improve the calculation efficiency while ensuring accuracy, and realizes real-time and high-precision detection of welding slag on the surface of stamping parts.

[0083] The present invention performs morphological processing on the separated defective area in step d, including: performing a morphological opening operation on the separated defective area to remove isolated small noise points, and then performing a closing operation to fill the small holes in the defective area to enhance the connectivity of the defective area; in image processing and point cloud data processing, "small holes" refer to small holes or discontinuous parts existing in the detected defective area. These small holes are caused by noise and scanning error factors. In morphological operations, filling these small holes can enhance the connectivity of the defective area and make it more complete, thereby improving the accuracy of subsequent analysis. Specifically, in the process of stamping surface inspection, small holes may appear in the following situations: Small holes in point cloud data: In three-dimensional point cloud data, due to the resolution limitation of the laser scanner or the loss of reflected signals, small holes may appear in certain areas. These small holes will affect the overall identification and measurement of the defective area. Small holes in the defective area after binarization: After the point cloud data is converted into a binary image (i.e., the defective area is separated from the background), small holes may appear inside the defective area due to threshold segmentation or noise. These small holes will make the defective area discontinuous, affecting the subsequent connected domain analysis and defect classification.

[0084] By filling the small holes in the defect area, the clarity and integrity of the defect area can be further improved. The adaptive morphological kernel size adjustment technology is applied to dynamically adjust the kernel size of the morphological operation according to the size and shape of the defect area to obtain more accurate results.

[0085] Among them, morphological processing is an image processing method based on mathematical morphology theory, which is mainly used to change the shape and structure of the target in the image. In the present invention, morphological processing is applied to the optimization of defect areas to improve the accuracy and reliability of defect detection. Specifically, morphological opening and closing operations are two basic morphological operations. The opening operation consists of an erosion operation followed by an expansion operation, while the closing operation consists of an expansion operation followed by an erosion operation. The combination of these two operations can effectively remove noise and enhance the connectivity of the target area. In practical applications, the specific parameters of the morphological processing need to be adjusted according to the material, surface characteristics and typical size of the defects of the stamping parts. For example, for stamping parts with rougher surfaces, it is necessary to use larger structural elements for opening operations to effectively remove the noise caused by surface roughness. For the case where the expected defects are smaller, it is necessary to use smaller structural elements for closing operations to avoid defect omissions caused by overfilling. As a preferred embodiment, a circular or square structural element can be used for morphological operations. The size of the structural element can be set to between 1 / 4 and 1 / 2 of the expected minimum defect size. For example, if the smallest defect is expected to be 2 mm in diameter, a structural element size between 0.5 mm and 1 mm can be selected. In addition, morphological processing can be combined with other image processing techniques to further improve the effect of defect detection. For example, before morphological processing, Gaussian filtering can be applied to smooth the image and reduce small noise. After morphological processing, edge detection algorithms can be used to accurately locate the boundaries of defects.

[0086] By implementing this morphological processing step, the present invention can effectively optimize the shape and structure of the defect area and improve the accuracy and reliability of defect detection. Compared with the traditional simple threshold segmentation method, the method of the present invention can better handle defects on complex surfaces and reduce the probability of false detection and missed detection. At the same time, due to the high computational efficiency of morphological operations, this method can also meet the needs of real-time detection and is suitable for online detection on high-speed production lines.

[0087] Before separating the defective area in the point cloud image from the background in step d, a basic threshold is preset according to the reflective characteristics of the stamping material, and the basic threshold is adjusted according to the changes in the ambient light intensity and the distribution of the point cloud reflection intensity in the three-dimensional point cloud data to ensure the accurate separation of the subsequent defective area. This method improves the accuracy and robustness of defective area separation by dynamically adjusting the threshold to adapt to different environmental conditions and material properties. Specifically, the preset basic threshold provides an initial reference point for defective area separation, and the dynamic adjustment based on the ambient light intensity and the point cloud reflection intensity distribution ensures that the threshold can adapt to changes in actual detection conditions.

[0088] In practical applications, this dynamic threshold adjustment method can be implemented through the following steps:

[0089] First, based on the reflective properties of the stamping material, an initial basic threshold is determined through experiments or empirical data. This basic threshold is usually set to a value that can effectively distinguish defective areas from normal surfaces under standard lighting conditions. Before each inspection, the inspection system in the computer evaluates the current ambient light intensity. This can be achieved by installing light sensors in the inspection area. If a significant change in the ambient light intensity is detected, the inspection system will adjust the basic threshold accordingly. For example, when the light is strong, the threshold needs to be increased to avoid misjudgment; when the light is weak, the threshold needs to be lowered to ensure that defects are not missed. The inspection system analyzes the distribution of the reflection intensity of the point cloud in the 3D point cloud data. This can be achieved by calculating the statistical characteristics of the reflection intensity of the point cloud (such as mean, standard deviation, etc.). If the reflection intensity distribution is found to be significantly different from the expected, the inspection system will further fine-tune the threshold. For example, if the overall reflection intensity is found to be high, the threshold needs to be increased accordingly; otherwise, the threshold needs to be lowered. Finally, the inspection system applies the adjusted threshold to the separation process of the defective area. Through this dynamic adjustment mechanism, the inspection system can maintain a high defect detection accuracy even when the environmental conditions change or the surface state of the stamping part is different.

[0090] As a preferred implementation, the following specific steps may be used to implement dynamic threshold adjustment:

[0091] Preset base threshold: For cylinder cover stamping parts, the base threshold under standard lighting conditions (e.g., 500 lux) is set to 100 (assuming a reflection intensity range of 0-255). Ambient light intensity assessment: Use a light sensor to monitor ambient light intensity in real time. When a change in light intensity of more than 20% is detected, the threshold adjustment is triggered. For example, if the light intensity increases to 600 lux, increase the threshold by 5-10%. Point cloud reflection intensity analysis: Calculate the mean and standard deviation of the reflection intensity in the current point cloud data. If the mean deviates from the expected value (e.g., the expected value is 128) by more than 20%, or the standard deviation is outside the expected range (e.g., the expected range is 20-40), further adjustments are made. For example, if the mean rises to 160, increase the threshold by another 5-10%. Threshold application: Apply the final adjusted threshold to the defect area separation algorithm. For example, if the initial threshold is 100 and is adjusted to 115 after ambient light and reflection intensity analysis, use 115 as the final separation threshold. Feedback mechanism: After completing a test, the detection system will evaluate the quality of the separation results (for example, through manual sampling or automatic evaluation algorithms). If the separation effect is found to be poor, the detection system will record the relevant parameters for future threshold adjustment optimization.

[0092] Through this dynamic threshold adjustment method, the detection method provided by the present invention can effectively cope with different environmental conditions and significantly improve the accuracy and reliability of defect area separation. This not only improves the detection efficiency, but also reduces the probability of misjudgment and missed judgment, thereby improving the reliability of the entire stamping surface detection. Specifically, the dynamic threshold adjustment method of the present invention has the following advantages: Strong adaptability: Compared with the fixed threshold method, the present invention can automatically adapt to changes in ambient light and differences in material properties to maintain the stability of detection. Higher accuracy: By considering multiple factors (ambient light, point cloud reflection intensity distribution) to adjust the threshold, the present invention can more accurately distinguish between defective areas and normal surfaces. Good robustness: Even in complex or changing detection environments, the present invention can maintain a high detection accuracy and reduce the need for human intervention. Easy to implement: The method of the present invention can be easily integrated into the existing three-dimensional point cloud data processing process without significantly changing the hardware settings or detection process.

[0093] The present invention performs connected domain analysis on the defective area after morphological processing in step e, including: determining the specific position and size of each defect on the stamping part; calculating the area and perimeter parameters of each connected domain, and recording its position information, for the identified defects, visually marking them on the display interface, and marking the defective area with a red outline; generating a test report containing defect location and size information.

[0094] The present invention can more accurately locate and quantify the defects on the surface of stamping parts through the above method. By calculating the area and perimeter of the connected domain, the specific size information of the defect can be obtained, thereby providing a more accurate basis for subsequent defect evaluation and processing. At the same time, the defect area is marked with a red outline on the display interface, which can intuitively present the test results and facilitate the operator to quickly identify the problem area. In addition, the automatically generated test report contains detailed defect location and size information, which provides strong support for quality control and production management. Specifically, when performing the connected domain analysis, the detection method of the present invention first marks the binary image after morphological processing. Each adjacent non-zero pixel forms a connected domain, representing a potential defect area. For each connected domain, its area (i.e., the number of pixels) and perimeter (i.e., the number of boundary pixels) are calculated. These parameters can be used to evaluate the size and shape of the defect. Furthermore, the present invention adopts coordinate mapping technology to convert the connected domain position in the image into the position in the actual coordinate system of the stamping part. This requires the pre-establishment of the conversion relationship between the image coordinate system and the stamping part coordinate system. Thus, the actual position of each defect on the stamping part can be accurately located. In terms of visual marking, the present invention uses computer graphics technology to draw red contour lines on the original image. This marking method not only highlights the defect area, but also retains the background information of the original image, so that the operator can better understand the context of the defect. For the generation of the inspection report, the present invention designs a structured data format that contains the following information: defect number, defect type (such as welding slag, dent), defect location (x, y, z coordinates), defect area, defect perimeter, and severity assessment. This formatted report facilitates subsequent data analysis and statistics.

[0095] As a preferred embodiment, the present invention can also introduce a machine learning algorithm to assist in defect classification. By training a neural network model, the detection system can automatically classify the detected defects into different types, such as welding slag, scratches, dents, etc. This intelligent classification can further improve the accuracy and efficiency of detection.

[0096] For example, in practical applications, the method of the present invention can be implemented by the following specific steps:

[0097] Preprocessing: Noise reduction and edge enhancement are performed on the binary image after morphological processing.

[0098] Connected domain labeling: Use the 8-connected algorithm to label the image and assign a unique identifier to each connected region.

[0099] Feature extraction: The following features are calculated for each labeled connected domain:

[0100] Area: Count the number of pixels in a connected domain.

[0101] Perimeter: Calculates the number of pixels along the border of a connected domain.

[0102] Centroid: Calculates the coordinates of the geometric center of a connected domain.

[0103] Shape Factor: Calculates shape characteristics such as circularity.

[0104] Coordinate mapping: Use a pre-calibrated coordinate transformation matrix to transform the image coordinates into the actual coordinates of the stamping part.

[0105] Visualization Marker: Draw red contour lines on the original image using image processing libraries such as OpenCV, and set the line width to 2-3 pixels to ensure visibility.

[0106] Report Generation: Creates a structured report in JSON or XML format with detailed information for each defect.

[0107] Machine learning classification: Use a pre-trained convolutional neural network model to classify defects and output the probability distribution of defect types.

[0108] Through this implementation, the present invention can achieve high-precision and high-efficiency surface defect detection of stamping parts. Compared with traditional manual detection, this method greatly improves the accuracy and consistency of detection. At the same time, through automated report generation and visual display, the readability and practicality of the detection results are significantly improved. In addition, the method of the present invention has significant advantages over simple image processing technology. Traditional image processing methods can usually only detect simple surface defects, but it is difficult to process stamping parts with complex shapes. By combining three-dimensional point cloud data and advanced image analysis technology, the present invention can more comprehensively capture the geometric features of the surface of stamping parts, thereby improving the detection ability of various types of defects. Compared with the method that relies solely on ultrasonic detection, the method of the present invention has more advantages in surface defect detection. Ultrasonic detection is mainly used for the identification of internal defects, while the method of the present invention focuses on the precise positioning and quantification of surface defects, and can provide more intuitive and detailed surface quality information. Through precise connected domain analysis, intuitive visual marking and detailed report generation, it provides strong technical support for improving the level of product quality control.

[0109] In step f of the present invention, the ultrasonic vibration mechanism emits high-frequency vibration through the ultrasonic transducer to loosen and remove the welding slag attached to the surface of the stamping part; the state change of the defective area is monitored by the ultrasonic vibration displacement sensor to confirm whether the welding slag is effectively removed; if there is still uncleaned welding slag on the stamping part after ultrasonic treatment, step f is repeated until the welding slag is completely removed, and the stamping part after slag removal is transported to the next station through the unloading robot arm. This improved welding slag cleaning method mainly solves the problems of low efficiency and incomplete cleaning of traditional cleaning methods. By introducing the ultrasonic vibration mechanism, the welding slag attached to the surface of the stamping part can be loosened and removed more effectively. The use of ultrasonic vibration displacement sensors ensures real-time monitoring and feedback of the cleaning process, and improves the accuracy and completeness of the cleaning. The repeated processing mechanism further ensures the thorough removal of the welding slag.

[0110] Specifically, the welding slag cleaning method of the present invention includes the following key features: First, the ultrasonic vibration mechanism is the core component of the method. The ultrasonic transducer uses an ultrasonic device with a power of 6000W and 4200W, which can emit high-frequency vibrations, which can effectively loosen the welding slag attached to the surface of the stamping part. The frequency of the high-frequency vibration can be adjusted according to the material of the stamping part and the characteristics of the welding slag (usually between 15kHz and 20kHz. The amplitude of the vibration can also be adjusted as needed, generally between 10μm and 100μm), and the ultrasonic power parameter is between 1500w-3800w. Secondly, the ultrasonic vibration displacement sensor is used to monitor the state changes of the defective area. This sensor can accurately detect slight changes in the surface of the stamping part to determine whether the welding slag has been removed. Ensure that the surface changes after the welding slag falls off can be captured. Furthermore, the present invention introduces a repeated processing mechanism. If there is still welding slag remaining after the first ultrasonic treatment, the detection system will automatically repeat the processing. This process can be repeated as many times as needed until it is confirmed that all welding slag has been removed. The time for each repeated processing can be adjusted according to the actual situation, usually between 1 second and 30 seconds. The unloading robot can accurately transport the cleaned stamping parts to the next station, reducing the need for manual operation.

[0111] In practical applications, the welding slag cleaning method of the present invention can effectively solve the problems existing in the traditional cleaning method. For example, for a typical automobile engine cylinder shield stamping part, the traditional manual cleaning method may take 5-10 minutes to complete the welding slag cleaning of a workpiece, and the cleaning effect is often not ideal. However, using the method of the present invention, the entire cleaning process can be completed within 1-2 minutes, and the cleaning effect is more thorough.

[0112] When implementing it, you can follow the steps below:

[0113] First, the stamping parts are transported to the bottom of the loading robot arm through the conveyor table, and the loading robot arm places the stamping parts accurately in the working area of ​​the ultrasonic vibration mechanism. The working area is usually designed as a shockproof platform to ensure that the ultrasonic vibration can be effectively transmitted to the stamping parts. The ultrasonic transducer is activated to generate high-frequency vibration. The vibration frequency is set to 15kHz and the amplitude is 100μm. These parameters can be fine-tuned according to the specific material of the stamping parts and the characteristics of the welding slag. The ultrasonic vibration displacement sensor starts working and monitors the state changes of the surface of the stamping parts in real time. The sampling frequency of the sensor is set to 1000Hz to ensure that small surface changes can be captured. The detection system sets the initial cleaning time to 20 seconds. Within these 20 seconds, the ultrasonic action time is set according to the actual effect. After 20 seconds, the detection system analyzes the data of the displacement sensor. If it is detected that the welding slag has completely fallen off, it will proceed to the next step; if there is still welding slag remaining, the detection system will automatically perform a second cleaning, and the time is set to 15 seconds. Repeat the above steps until it is confirmed that all welding slag has been removed. The detection system is set to repeat up to 3 times to prevent excessive processing from damaging the surface of the stamping parts. After the cleaning is completed, the unloading robot arm moves the stamping parts from the working area of ​​the ultrasonic vibration mechanism to the next workstation. The moving speed of the robot arm is set to 0.5m / s to ensure smooth and accurate operation. Through this method, efficient and thorough welding slag cleaning can be achieved, while ensuring the surface quality of the stamping parts. Compared with traditional manual cleaning or simple mechanical cleaning, it has the following advantages: the cleaning efficiency is significantly improved, the cleaning effect is more thorough, and the ultrasonic vibration can penetrate into the small gaps and depressions of the stamping parts to remove the welding slag that is difficult to reach by traditional methods. At the same time, it also reduces manual operation, labor intensity and labor costs. In general, the welding slag cleaning method proposed in the present invention effectively solves the efficiency and quality problems in the cleaning of welding slag of stamping parts, and provides better guarantees for the subsequent processing and use of stamping parts.

[0114] Figure 2 A schematic diagram of the structure of a welding slag cleaning device based on a stamping part surface detection method provided by the present invention is shown in FIG. Figure 2As shown, the present invention also provides a welding slag cleaning device based on the stamping part surface detection method, including a stamping part conveying platform and a workbench, on which a loading robot arm, an ultrasonic vibration and dropping mechanism and a unloading robot arm are installed, and the stamping part conveying platform is used to convey the stamping part to the bottom of the loading robot arm. The welding slag cleaning device realizes the automatic cleaning process of welding slag on the surface of the stamping part by integrating the stamping part conveying platform, the workbench, the loading robot arm, the ultrasonic vibration and dropping mechanism and the unloading robot arm. The stamping part conveying platform conveys the stamping part to be processed to the bottom of the loading robot arm, and the loading robot arm transfers the stamping part to the working area of ​​the ultrasonic vibration and dropping mechanism for welding slag cleaning. After the cleaning is completed, the unloading robot arm conveys the stamping part to the next station. This design effectively improves the efficiency of welding slag cleaning and reduces labor costs. Specifically, the stamping part conveying platform can adopt a conveyor belt or a roller conveying mechanism for continuous and stable conveying of stamping parts. As the basic platform of the entire device, the workbench provides a stable support and can be made of metal material to withstand the weight of each mechanism and the vibration generated during the movement. The loading robot arm can adopt a multi-axis robot arm, such as a six-axis robot arm, to achieve flexible grasping and placing actions. The end of the robot arm can be equipped with a clamp or suction cup suitable for the shape of the stamping part to ensure stable grasping. The control and detection system of the loading robot arm can be synchronized with the stamping part conveyor table to achieve accurate material picking time. The ultrasonic vibration mechanism includes an ultrasonic transducer and an ultrasonic vibration displacement sensor, wherein the ultrasonic transducer can be made of piezoelectric ceramic material and can convert electrical energy into high-frequency mechanical vibration. The vibration frequency is set in the range of 20-40kHz. The design of the unloading robot arm is the same as that of the loading robot arm, and a multi-axis structure is also adopted to achieve flexible operation. Therefore, the welding slag cleaning device proposed in the present invention can effectively solve the problems of low efficiency and high labor cost of welding slag cleaning on the surface of stamping parts. Through the automated design, the cleaning efficiency is greatly improved and the need for manual intervention is reduced. At the same time, the application of ultrasonic vibration technology ensures the thoroughness and uniformity of cleaning and improves the surface quality of stamping parts. The stamping parts conveyor adopts a belt conveyor with a width of 800mm, and the conveying speed is adjustable in the range of 0.1-0.5m / s. The workbench is welded with 40mm thick steel plates, and the dimensions are 2000mm×1500mm×800mm (length×width×height). The ends of the loading and unloading robotic arms are equipped with pneumatic clamps, and the clamping force can be adjusted between 50-200N. The ultrasonic transducer has a power of 1000W, an operating frequency of 28kHz, and an adjustable amplitude range of 20-100μm. In practical applications, the device can handle stamping parts of various shapes within a size of 500mm×400mm×200mm. During the cleaning process, the ultrasonic vibration time can be adjusted according to the welding slag situation, and is generally set in the range of 10-30 seconds. Through visual inspection and intelligent control, the device can automatically judge the cleaning effect and perform secondary cleaning if necessary.The entire cleaning process takes no more than 1 minute on average, greatly improving efficiency.

[0115] Compared with the prior art, the welding slag cleaning device proposed in the present invention has the following advantages: First, by integrating the functions of stamping part transmission, loading and unloading, and ultrasonic cleaning, it realizes full automation and significantly improves production efficiency. Second, the application of ultrasonic vibration technology can more effectively remove welding slag than traditional mechanical brushing or manual cleaning methods, especially for welding slag cleaning on complex-shaped stamping parts.

[0116] Figure 3 The present invention provides a schematic diagram of the layout of a sandblasting machine based on a stamping part surface detection method. The present invention also includes step g. The stamping part treated by the ultrasonic vibration mechanism is moved to the laser marking mechanism through the feeding line for laser marking, and the number of times the stamping part is repaired is marked. Then, the stamping part is placed in the sandblasting machine for automatic sandblasting by the loading and unloading robot. Specifically, the sandblasting machine mainly uses compressed air as power, and the sandblasting air pressure is between 0.3-0.7MPA. The 0.8mm bearing steel particle abrasive is sprayed at high speed to the workpiece surface through the nozzle, and hits the shield surface, thereby achieving the purpose of cleaning, roughening and improving the surface quality of the workpiece. The compressed air not only provides power, but also accelerates the abrasive through the nozzle, so that it hits the surface of the stamping part, thereby achieving the expected processing effect. The above method realizes the reuse of the cylinder shield product, solving the problem that it could not be repaired before. After the sandblasting is completed, the quality of the stamping part is inspected by the camera in the visual inspection mechanism, and the stamping parts that pass the inspection are automatically boxed and packaged by the boxing robot.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting the surface of a stamping part, characterized in that: The following steps are involved: Step a. Before the stamping part is conveyed to the stamping part conveying station, the stamping part is three-dimensionally scanned using laser scanners arranged at different angles, wherein each laser scanner emits a laser beam and receives a reflected laser signal; Step b. The computer constructs three-dimensional point cloud data of the stamping part according to the received laser signal; performs preliminary filtering on the three-dimensional point cloud data to remove noise points; By calculating the coordinates and reflection intensity of each point, a point cloud image is formed; Step c. extracting key geometric features of the stamping part from the three-dimensional point cloud data, wherein the key geometric features include edge features and curvature mutation points of the stamping part; Step d. Separating the defect area from the background in the point cloud image to form a binary defect area image; performing morphological processing on the separated defect area; Step e. Performing a connected domain analysis on the defective area after morphological processing to obtain defect location information of the stamping part; Step f. According to the defect position information of the stamped part, the stamped part is transferred from the stamped part conveying platform to the working area of ​​the ultrasonic vibration mechanism by the loading robot arm, and the welding slag on the stamped part is shaken off by the ultrasonic vibration mechanism; Step g. After the stamping parts have been processed by the ultrasonic vibration mechanism, the loading and unloading robots place the stamping parts in a sandblasting machine for automatic sandblasting.

2. A method for detecting the surface of a stamping part according to claim 1, characterized in that: In the step c, key geometric features of the stamping part are extracted from the three-dimensional point cloud data. The key geometric features include edge features and curvature mutation points of the stamping part. Step c1: For point P in the 3D point cloud data i , calculate the mean curvature H and Gaussian curvature K in its local neighborhood; Step c2: For each point P i , calculate its point P with the neighborhood j The curvature difference between , the calculation formula is , where C represents the curvature and mean is the average value of the curvature in the neighborhood; Step c3: setting a threshold T to distinguish between normal curvature changes and sudden changes in curvature; Step c4: For all >Point P of T i , marking it as the curvature mutation point; Step c5: Use the sliding window method or the K-nearest neighbor method to determine the i Neighborhood of; Step c6: Apply the Canny edge detection algorithm to identify edge points based on the curvature change; Step c7: Use an edge tracking algorithm to connect isolated edge points into continuous edge lines to form edge features of the stamping part.

3. A method for detecting the surface of a stamping part according to claim 2, characterized in that: In step c1, for point P in the three-dimensional point cloud data i , calculating the average curvature H and Gaussian curvature K in its local neighborhood includes: Step c1-1: For point P in the 3D point cloud data i , select its k nearest neighbors as the local neighborhood, and use the octree spatial index structure to accelerate the nearest neighbor search; Step c1-2: Use the least squares fitting method to fit P i The local neighborhood point P j Fit a plane or quadratic surface and extract the principal curvatures k1 and k2; Step c1-3: By calculating each point P in the local neighborhood j The average curvature H is obtained by taking the average of the principal curvatures, and the calculation formula is: , where k1 is the maximum principal curvature, indicating that at this point P i The maximum curvature of the surface along a certain direction at point P; k2 is the minimum principal curvature, indicating that at this point P i The minimum curvature of the surface in another direction; Step c1-4: By calculating each point P in the local neighborhood j The Gaussian curvature K is obtained by multiplying the principal curvatures of .

4. A method for detecting the surface of a stamping part according to claim 1, characterized in that: The morphological processing performed on the separated defective area in step d includes: performing a morphological opening operation on the separated defective area to remove isolated small noise points, and then performing a closing operation to fill small holes in the defective area to enhance the connectivity of the defective area.

5. A method for detecting the surface of a stamping part according to claim 1, characterized in that: Before separating the defective area in the point cloud image from the background in step d, a basic threshold is preset according to the reflective characteristics of the stamping material, and the basic threshold is adjusted according to the change in ambient light intensity and the distribution of point cloud reflection intensity in the three-dimensional point cloud data to ensure accurate separation of subsequent defective areas.

6. A method for detecting the surface of a stamping part according to claim 1, characterized in that: The connected domain analysis of the defective area after morphological processing in step e includes: determining the specific position and size of each defect on the stamping part; calculating the area and perimeter parameters of each connected domain, and recording its position information, and visually marking the identified defects on the display interface, using red contour lines to mark the defective area; generating a test report containing defect location and size information.

7. A method for detecting the surface of a stamping part according to claim 1, characterized in that: The number of the laser scanners is more than three, and the laser scanners are evenly distributed around the stamping part.

8. A method for detecting the surface of a stamping part according to claim 1, characterized in that: In step f, the ultrasonic vibration-removing mechanism emits high-frequency vibrations through the ultrasonic transducer to loosen and fall off the welding slag attached to the surface of the stamping part; the state change of the defective area is monitored by the ultrasonic vibration-removing displacement sensor to confirm whether the welding slag is effectively removed; if there is still uncleaned welding slag on the stamping part after ultrasonic treatment, the ultrasonic vibration-removing mechanism is used to continuously vibrate the welding slag on the stamping part until the welding slag is completely removed, and the stamping part after slag removal is transported to the next workstation by the unloading robot arm.

9. A welding slag cleaning device based on the stamping part surface detection method according to claim 1, comprising a stamping part conveying table and a workbench, characterized in that: A loading robot arm, an ultrasonic vibration mechanism and a unloading robot arm are installed on the workbench, and the stamping part conveying platform is used to convey the stamping parts to the bottom of the loading robot arm.

10. The welding slag cleaning device according to claim 9, characterized in that: The ultrasonic vibration mechanism includes an ultrasonic transducer and an ultrasonic vibration displacement sensor.

Images