Steel labeling and code spraying method based on visual system

Abstract

The invention discloses a steel labeling and code spraying method based on a visual system, belongs to the technical field of steel labeling and code spraying, and provides a steel channel and H-shaped steel labeling and code spraying area positioning algorithm by combining a plane fitting function and an OBB orientation bounding box algorithm according to the characteristic that a steel channel and H-shaped steel patch code spraying area is flat. An accurate area suitable for labeling and code spraying operation on the steel is effectively identified and positioned; according to the characteristic that the cross sections of the angle steel and the U-shaped steel are bent, an RANSAC algorithm is combined with a PCA algorithm, a labeling and code-spraying area positioning algorithm for the angle steel and the U-shaped steel is provided, and an accurate area suitable for labeling and code-spraying operation on the steel is effectively recognized and positioned.

Description

Technical Field

[0001] The invention relates to the technical field of steel labeling and coding, and in particular to a steel labeling and coding method based on a vision system. Background Art

[0002] With the advancement of science and technology and industrial development, the demand for automation technology in the steel production field is increasing. Most steel mills produce mainly channel steel, H-shaped steel, angle steel, and U-shaped steel. Labeling and coding are key links in the management and tracking of steel products. The accuracy and efficiency of their execution have a significant impact on the production quality and economic benefits of enterprises. In recent years, most steel companies still use manual operation to label and spray code. This traditional practice has a series of problems under certain conditions, including low production efficiency. Manual operation requires operators to identify and mark steel one by one, which makes the labeling and spraying process lack automation and cannot achieve large-scale and efficient production; limited operation consistency. Manual operation is affected by the operator's skill level and working environment, and it is difficult to ensure that the labeling and spraying position and accuracy of each steel are consistent; dangerous working environment. At the steel production site, operators need to be in harsh working environments such as high temperature, high noise and high dust for a long time, which not only increases the danger of work, but also may affect the operator's health and increase the risk of work-related accidents; high labor costs, manual operation requires a lot of manpower input, manual operation requires training and skill improvement, which increases the management cost and time cost of the enterprise; problems with manual intervention restrict the production efficiency and quality level of steel enterprises. Therefore, realizing the automation, precision and intelligence of the labeling and spraying process of steel production is of great significance to improving the efficiency, quality and safety of steel production.

[0003] In recent years, with the rapid development of robot technology and image processing technology, automated production has become a trend in the manufacturing industry. Steel mills have also begun to try to use intelligent visual robots to replace manual production. By using image processing technology to analyze and identify data in real time, they can accurately determine the safe area suitable for labeling and coding, thereby controlling the robot to achieve automatic identification and labeling and coding automation. This combination of technologies has brought more efficient and accurate labeling and coding solutions to steel production. Summary of the invention

[0004] The technical problem to be solved by the present invention is: how to solve the problems existing in the prior art such as low production efficiency, limited operation consistency, dangerous working environment, high labor cost, etc., and provide a steel labeling and coding method based on a visual system.

[0005] The present invention solves the above technical problems through the following technical solutions, and the present invention comprises the following steps:

[0006] S1: Data Collection

[0007] Use the line structured light mode 3D camera to collect data on steel samples and obtain point cloud data;

[0008] S2: Data preprocessing

[0009] Decode and preprocess the collected point cloud data to obtain effective steel point cloud images;

[0010] S3: Labeling and coding safety area positioning

[0011] Based on the effective steel point cloud image, the corresponding steel coding area positioning algorithm is used according to the type of steel to obtain the labeling coding safety area, and then the coordinate information of the labeling coding safety area is obtained;

[0012] S4: Labeling and coding operations

[0013] According to the identified coordinate information of the labeling and coding safety area, the robot end effector is moved to the specified position and maintains the correct coding posture to complete the labeling and coding tasks.

[0014] Furthermore, in the step S2, the image captured by the line structured light mode 3D camera is encoded inside the camera to form data in dat and xml formats, and the data in dat and xml formats are decoded and converted into a point cloud structure image and its intensity information, so as to obtain a steel point cloud image.

[0015] Furthermore, in step S2, the preprocessing operation includes the following process:

[0016] S21: Obtaining the decoded steel point cloud image;

[0017] S22: Apply a bilateral filter to filter out noise points in the steel point cloud image;

[0018] S23: using a statistical filter combined with a statistical outlier removal algorithm to remove multiple outliers in the steel point cloud image;

[0019] S24: downsampling the steel point cloud image using a voxel filter;

[0020] S25: Use the RANSAC algorithm to fit the plane equation, segment the foreground and background of the steel point cloud image, and extract the point cloud image of the upper steel structure, that is, obtain the effective steel point cloud image.

[0021] Furthermore, in step S3, the types of steel include four categories, namely H-shaped steel, angle steel, U-shaped steel and channel steel; among them, channel steel and H-shaped steel obtain the labeling and coding safety area through the first steel coding area positioning algorithm, and angle steel and U-shaped steel obtain the labeling and coding safety area through the second steel coding area positioning algorithm.

[0022] Furthermore, the specific processing process of the first steel material coding area positioning algorithm is as follows:

[0023] S311: Randomly select points

[0024] In each iteration, n points are randomly selected from the point cloud of the valid steel point cloud image, and a plane model is fitted based on these points. The plane model is defined by the parameters [a, b, c, d] in the plane equation ax+by+cz+d=0;

[0025] S312: Calculate the distance from a point to a plane

[0026] For each point in the point cloud, calculate its vertical distance to the fitting plane, where the fitting plane is the plane model mentioned above;

[0027] S313: Determine interior points

[0028] According to the set distance threshold, points with a distance less than the distance threshold are classified as inliers, that is, they are considered to be located on the current fitting plane, and the area composed of inliers is the labeling safety area;

[0029] S314: Evaluating Plane Models

[0030] The quality of the plane models was evaluated based on the number of inliers in each plane model.

[0031] Furthermore, the specific processing process of the second steel material coding area positioning algorithm is as follows:

[0032] S321: Use the RANSAC algorithm to preliminarily process the point cloud data of the effective steel point cloud image, iteratively optimize the plane model parameters to improve the fitting degree, identify and fit the main fitting plane in the angle steel and U-shaped steel point cloud, remove the noise and outliers of the plane model, and retain the point cloud data associated with the main fitting plane, wherein the main fitting plane is an area not less than 400mm*80mm;

[0033] S322: Apply the PCA algorithm to the point cloud data preprocessed by the RANSAC algorithm, introduce the KD-Tree structure to organize the point cloud data, and divide the point cloud space into different areas. Use the PCA algorithm to calculate the point cloud normal vector, and further refine and extract the specific geometric features of the outer curved surface of the steel according to the normal vector;

[0034] S323: After determining the outer curved surface of the steel material for labeling and coding, continue to use the OBB directional bounding box algorithm, combined with the average slope information of the outer surface of the steel material, to obtain a minimum circumscribed rectangular bounding box to cover the extracted slope. Within this bounding box, further locate and obtain the labeling and coding safety area and its coordinate information.

[0035] Furthermore, in step S323, the specific processing process is as follows:

[0036] S3231: After determining the outer curved surface of the steel material to be labeled and coded, use the OBB directional bounding box algorithm; by analyzing the direction of the extracted outer curved surface of the steel material, use the OBB directional bounding box algorithm to generate a minimum circumscribed rectangular bounding box aligned with the direction of the surface;

[0037] S3232: Calculate the average slope information of the outer surface of the steel material, and use the information to adjust the positioning of the OBB bounding box; finally, obtain a minimum rectangular bounding box that can tightly surround the outer curved surface of the steel material;

[0038] S3233: In the smallest rectangular bounding box obtained in step S3232, a 400mm*80mm labeling and coding safety area is located inward from 30mm along the positive direction of the x-axis of the bounding box as the starting point, and its specific coordinates are extracted.

[0039] Furthermore, the robot end effector includes a vacuum suction cup and a spray terminal, the vacuum suction cup is used to capture and place the labeling material, and the spray terminal is used to perform coding operations after labeling is completed.

[0040] Compared with the prior art, the present invention has the following advantages: the steel labeling and coding method based on the vision system, according to the flatness of the patch coding area of ​​channel steel and H-shaped steel, by combining the plane fitting function and the OBB directional bounding box algorithm, proposes a positioning algorithm for the labeling and coding area of ​​channel steel and H-shaped steel, which effectively identifies and locates the accurate area on the steel that is suitable for labeling and coding operations; according to the curved cross-section characteristics of angle steel and U-shaped steel, by combining the RANSAC algorithm with the PCA algorithm, a positioning algorithm for the labeling and coding area of ​​angle steel and U-shaped steel is proposed, which effectively identifies and locates the accurate area on the steel that is suitable for labeling and coding operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flow chart of the steel labeling and coding method based on the visual system of the present invention;

[0042] Figure 2a is a finished product image of a channel steel in an embodiment of the present invention;

[0043] Figure 2b is an image of a finished H-beam in an embodiment of the present invention;

[0044] Figure 2c is a finished product image of an angle steel in an embodiment of the present invention;

[0045] Figure 2d is a finished product image of a U-shaped steel in an embodiment of the present invention;

[0046] Figure 3a is a valid steel point cloud image of the channel steel in the embodiment of the present invention;

[0047] Figure 3b is a valid steel point cloud image of the H-shaped steel in the embodiment of the present invention;

[0048] Figure 3c is a valid steel point cloud image of the angle steel in the embodiment of the present invention;

[0049] Figure 3d It is a valid steel point cloud image of the U-shaped steel in the embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0051] Embodiment 1

[0052] like Figure 1 As shown, this embodiment provides a technical solution: a steel labeling and coding method based on a visual system, comprising the following steps:

[0053] The Sick Ruler E600 model line structured light mode 3D camera was used to collect data on the steel samples;

[0054] Decode the collected point cloud data, and then use pre-processing methods such as point cloud filtering, denoising and background removal to improve the quality of the point cloud data, thereby improving the accuracy of the inkjet positioning algorithm;

[0055] According to the type of steel, the corresponding steel coding area positioning algorithm is used to accurately locate a 400mm*80mm labeling and coding safety area, providing clear spatial positioning guidance for subsequent robot coding and labeling operations;

[0056] According to the coordinate information of the labeling and coding safety area, the robot end effector is moved to the specified position and maintains the correct coding posture to complete the labeling and coding tasks.

[0057] In this embodiment, the image captured by the Sick Ruler E600 model line structured light mode 3D camera is encoded inside the camera to form data in a specific format (dat and xml formats). This encoded data cannot be directly applied to point cloud image processing, so it must first be decoded and converted into a point cloud structure image and its intensity information, that is, a steel point cloud image is obtained.

[0058] In this embodiment, the process of preprocessing the point cloud data is as follows:

[0059] The preprocessing operation mainly includes point cloud filtering, denoising, background removal and downsampling methods; the specific process is as follows: obtain the decoded steel point cloud image; apply the bilateral filter to filter out the noise points in the steel point cloud image, while maintaining the important detail features in the steel point cloud image; use the statistical filter combined with the statistical outlier removal (SOR) algorithm to specifically target and remove a large number of outliers in the image, thereby improving the overall accuracy and reliability of the data; use the voxel filter to downsample the steel point cloud image, while reducing the number of point clouds, it also retains the key detail features of the steel point cloud image, and improves the efficiency of subsequent steel point cloud processing; use the RANSAC algorithm to fit the plane equation, segment the foreground and background of the steel point cloud image, and realize the accurate extraction of the upper steel structure point cloud image, that is, to obtain the effective steel point cloud image; this process reduces the total parameter amount of the point cloud data by about 84%, while ensuring the data quality, significantly reducing the computational burden of subsequent processing and analysis.

[0060] In this embodiment, the types of steel and the differences between the types are mainly as follows:

[0061] There are four main types of steel, including H-shaped steel, angle steel, U-shaped steel and channel steel. Among them, the cross section of the channel steel looks like a "C" shape, with the inner side remaining straight, the outer side slightly vertical inward, and the side and bottom connected to form a groove; the cross section of the H-shaped steel is shaped like the English letter "H", with two wider parallel flanges, closely connected to the middle column (web); the cross section of the angle steel is "L" shaped, consisting of two vertically arranged planes, shaped like an angle; the cross section of the U-shaped steel is "U" shaped, similar to the channel steel, but the flanges on both sides are usually smaller and the outer surface is curved. Therefore, according to the shape of the steel cross section and the location of the labeling and coding, different algorithms are used to obtain the labeling and coding safety area. These four types of steel can be divided into two categories: the labeling and coding locations of the channel steel and H-shaped steel are the bottom surface of the groove and the surface of the middle column, both of which are flat; the labeling and coding locations of the angle steel and U-shaped steel are the outer surfaces, both of which are curved. Figure 2a-2d These are the finished product images of four types of steel products. Figure 3a-3dThe valid steel point cloud images collected are of four types of steel. The red parts are the labeling and coding locations of various steel products.

[0062] In this embodiment, the 400mm*80mm labeling and coding safety area needs to meet two major prerequisites: first, ensure that the labeling and coding safety area has sufficient visibility to facilitate subsequent identification and reading; second, avoid conflicts between the area and any existing obstacles or marks on the steel. Only by meeting the above two requirements can the coding operation be successfully completed.

[0063] In this embodiment, the corresponding coding area positioning algorithm includes:

[0064] According to the shape of the steel cross-section and the location of the labeling and coding, the steel coding area positioning algorithm is also divided into two categories: channel steel and H-shaped steel labeling and coding area positioning algorithm, angle steel and U-shaped steel labeling and coding area positioning algorithm.

[0065] In this embodiment, the main process of the channel steel and H-beam labeling and coding area positioning algorithm includes:

[0066] Given that the labeling and coding positions of channel steel and H-beam are both in the horizontal plane - the channel steel is located at the bottom of its outer groove, while the H-beam chooses the flat part of the middle column. Specifically, the purpose of this algorithm is to determine a 400mm*80mm rectangular area on the horizontal plane for labeling and coding tasks. First, the visual system obtains the processed point cloud data of the channel steel or H-beam, fits a plane equation by applying the Segment plane plane fitting function, and returns the equation of the final fitted plane and the index of the points that constitute the plane. The specific steps are as follows:

[0067] S1: Randomly select points

[0068] In each iteration, n points are randomly selected from the point cloud of the valid steel point cloud image, and a plane model is fitted based on these points. The plane model is defined by the parameters [a, b, c, d] in the plane equation ax+by+cz+d=0;

[0069] S2: Calculate the distance from the point to the plane

[0070] For each point in the point cloud, calculate its perpendicular distance to the fitting plane;

[0071] S3: Determine interior points

[0072] According to the set distance threshold, points with a distance less than the distance threshold are classified as inliers, that is, they are considered to be located on the current fitting plane, and the area formed by these inliers is the labeling safety area;

[0073] S4: Evaluating the Plane Model

[0074] The quality of the plane model is evaluated based on the number of inliers of each plane model. The quality of the plane model is determined by the number of its inliers. The more inliers, the more reliable the plane model is.

[0075] In this embodiment, the main process of the angle steel and U-shaped steel labeling and coding area positioning algorithm includes:

[0076] Considering that the labeling and coding positions of angle steels and U-shaped steels are all on the outer curved surfaces of the curves, due to the curvature characteristics of these positions and the angle and accuracy of the 3D camera shooting, when obtaining point cloud images of curved steels such as angle steels and U-shaped steels, there may be missing or discontinuous information at the bends. If the plane fitting function is directly used to fit the plane, it is often difficult to accurately capture and locate the outer side of such steels, and the sensitivity to data discontinuities is high. Therefore, for objects with complex geometric shapes such as angle steels and U-shaped steels, it is necessary to explore more accurate fitting techniques to identify and locate the precise position of the outer curved surface, so as to ensure the accuracy and efficiency of the labeling and coding operations. The present invention adopts the RANSAC algorithm combined with the principal component analysis (PCA) algorithm to specifically extract features from the outer curved surfaces of angle steels and U-shaped steels. The specific steps are as follows:

[0077] S1: Use the RANSAC algorithm to preliminarily process the point cloud data of the effective steel point cloud image, iteratively optimize the plane model parameters to improve the fitting degree, identify and fit the main fitting planes in the angle steel and U-shaped steel point clouds, remove the noise and outliers of the plane model, and retain the point cloud data associated with the main fitting plane, where the main fitting plane is an area not less than 400mm*80mm, which is convenient for subsequent labeling and coding operations;

[0078] S2: Apply the PCA algorithm to the point cloud data preprocessed by the RANSAC algorithm, introduce the KD-Tree structure to organize the point cloud data, and divide the point cloud space into different areas. Use the PCA algorithm to calculate the point cloud normal vector, and further refine and extract the specific geometric features of the outer curved surface of the steel according to the normal vector; the specific steps are as follows:

[0079] S21: For the point cloud data preprocessed by the RANSAC algorithm, a KD-tree structure is constructed, which can accelerate the query operation of the point cloud, thereby achieving more efficient processing of the point cloud data;

[0080] S22: Based on the KD-Tree structure, the point cloud data is divided into different spatial regions to achieve more detailed analysis of the point cloud data;

[0081] S23: In each divided spatial region, the PCA algorithm is used to calculate the normal vector of the point cloud in the region, and the geometric features of the outer curved surface of the steel are specifically extracted.

[0082] S3: After determining the outer curved surface of the steel that is suitable for labeling and coding, continue to use the OBB algorithm, combined with the average slope information of the outer surface of the steel, to obtain a minimum circumscribed rectangular bounding box to cover the extracted slope. Within this bounding box, a 400mm*80mm labeling and coding safety area is further located. Among them, the OBB (Oriented Bounding Box) oriented bounding box algorithm is a technology used to calculate the minimum circumscribed rectangular bounding box of an object. By performing principal component analysis on the point cloud data of the object, the bounding box is aligned with the average slope information of the outer surface of the steel to generate a rectangular bounding box that matches the shape and direction of the object, thereby more tightly surrounding the object. The specific steps are as follows:

[0083] S31: After determining the outer curved surface of the steel material to be labeled and coded, the OBB oriented bounding box algorithm is used. By analyzing the direction of the extracted curved surface, the OBB algorithm is used to generate a minimum circumscribed rectangular bounding box aligned with the surface direction to tightly cover the outer curved surface of the steel material.

[0084] S32: Calculate the average slope information of the outer surface of the steel material, where the outer surface of the angle steel refers to the plane area outside the two right-angle sides of the angle steel, and the outer surface of the U-shaped steel refers to the outer curved area of ​​the convex surface of the steel material. This information is used to adjust the positioning of the OBB bounding box so that it more accurately surrounds the curved surface. Finally, a minimum rectangular bounding box is obtained, which can tightly surround the curved surface of the outer side of the steel material.

[0085] S33: In the smallest bounding box obtained above, 30 mm from the positive direction of the x-axis of the bounding box is used as the starting point, and a 400 mm*80 mm labeling and coding safety area is accurately positioned inward, and the labeling and coding safety area is further positioned and its specific coordinates are extracted to provide an accurate spatial reference for the coding operation.

[0086] In this embodiment, the robot end effector includes: a dedicated vacuum suction cup and a spray terminal. In the robot system for realizing automatic labeling and coding operations, the vacuum suction cup is responsible for accurately capturing and placing the labeling material, while the spray terminal is responsible for performing the coding operation after the labeling is completed.

[0087] Embodiment 2

[0088] This embodiment further illustrates the steel labeling and coding method in the first embodiment, as follows:

[0089] Step 1: Data preparation

[0090] First, the image of the steel is acquired and decoded using the Sick Ruler E600 model 3D camera in line structured light mode, from the data in a special format into point cloud data type and intensity information.

[0091] Step 2: Data processing

[0092] Denoising: Apply bilateral filters to filter out noise points in steel point cloud images while maintaining important detail features in steel point cloud images; Remove discrete points: Use statistical filters combined with statistical outlier removal (SOR) algorithms to specifically target and remove a large number of outliers in the image, thereby improving the overall accuracy and reliability of the data; Downsampling: Use voxel filters to downsample steel point cloud images, which reduces the number of point clouds while retaining the key detail features of steel point cloud images, thereby improving the efficiency of subsequent steel point cloud processing; Fitting plane operation: Use the RANSAC algorithm to fit the plane equation, segment the foreground and background of the steel point cloud image, and achieve accurate extraction of the upper steel structure point cloud image.

[0093] Step 3: Algorithm processing

[0094] According to the type of steel, the corresponding steel coding area positioning algorithm is adopted.

[0095] For channel steel and H-beam, a plane is obtained by using the Segment plane plane fitting function; based on the fitted plane, the limits of the minimum circumscribed rectangular oriented bounding box are calculated; in order to avoid placing the label on the edge of the steel, the calculation starts from the position 30mm from the positive direction of the x-axis of the bounding box and is positioned inward to a 400mm*80mm label safety area. This method ensures the accuracy of the label position while avoiding any possible edge interference.

[0096] For angle steel and U-shaped steel, the optimized RANSAC algorithm is used to successfully fit the required inclined surface; on this basis, combined with the average slope of the inclined surface and applying the OBB algorithm, the covariance matrix of the point cloud is eigen-decomposed by using the oriented bounding box (OBB) method, and the main directional eigenvectors of the point cloud distribution are obtained to obtain three orthogonal principal axis directions, which correspond to the length, width and height directions of the OBB respectively; then, the algorithm determines the minimum and maximum coordinate values ​​of the point cloud along these principal axis directions to determine the specific size of the OBB, and calculates the minimum circumscribed rectangular oriented bounding box covering the inclined surface; consistent with the strategy for processing channel steel and H-shaped steel, in order to prevent the labeling and coding from being placed on the edge of the steel, the starting point is 30mm along the positive direction of the x-axis of the bounding box, and the labeling and coding safety area of ​​400mm*80mm is accurately located inward. This safety area is consistent with the average slope of the fitted inclined surface to ensure the accuracy and safety of the coding operation.

[0097] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A steel labeling and coding method based on a visual system, characterized in that: The following steps are involved: S1: Data Collection Use the line structured light mode 3D camera to collect data on steel samples and obtain point cloud data; S2: Data preprocessing Decode and preprocess the collected point cloud data to obtain effective steel point cloud images; S3: Labeling and coding safety area positioning Based on the effective steel point cloud image, the corresponding steel coding area positioning algorithm is used according to the type of steel to obtain the labeling coding safety area, and then the coordinate information of the labeling coding safety area is obtained; S4: Labeling and coding operations According to the identified coordinate information of the labeling and coding safety area, the robot end effector is moved to the specified position and maintains the correct coding posture to complete the labeling and coding tasks.

2. According to the method of steel labeling and coding based on a visual system according to claim 1, it is characterized in that: In step S2, the image captured by the line structured light mode 3D camera is encoded inside the camera to form data in dat and xml formats. The data in dat and xml formats are decoded and converted into a point cloud structure image and its intensity information, thereby obtaining a steel point cloud image.

3. According to the method of steel labeling and coding based on a visual system according to claim 1, it is characterized in that: In step S2, the preprocessing operation includes the following process: S21: Obtaining the decoded steel point cloud image; S22: Apply a bilateral filter to filter out noise points in the steel point cloud image; S23: using a statistical filter combined with a statistical outlier removal algorithm to remove multiple outliers in the steel point cloud image; S24: downsampling the steel point cloud image using a voxel filter; S25: Use the RANSAC algorithm to fit the plane equation, segment the foreground and background of the steel point cloud image, and extract the point cloud image of the upper steel structure, that is, obtain the effective steel point cloud image.

4. According to the method of steel labeling and coding based on a visual system according to claim 1, it is characterized in that: In step S3, the types of steel include four categories, namely H-shaped steel, angle steel, U-shaped steel and channel steel; among them, channel steel and H-shaped steel obtain the labeling and coding safety area through the first steel coding area positioning algorithm, and angle steel and U-shaped steel obtain the labeling and coding safety area through the second steel coding area positioning algorithm.

5. According to the method of steel labeling and coding based on a visual system according to claim 4, it is characterized in that: The specific processing process of the first steel material coding area positioning algorithm is as follows: S311: Randomly select points In each iteration, n points are randomly selected from the point cloud of the valid steel point cloud image, and a plane model is fitted based on these points. The plane model is defined by the parameters [a, b, c, d] in the plane equation ax+by+cz+d=0; S312: Calculate the distance from a point to a plane For each point in the point cloud, calculate its vertical distance to the fitting plane, where the fitting plane is the plane model mentioned above; S313: Determine interior points According to the set distance threshold, points with a distance less than the distance threshold are classified as inliers, that is, they are considered to be located on the current fitting plane, and the area composed of inliers is the labeling safety area; S314: Evaluating Plane Models The quality of the plane models was evaluated based on the number of inliers in each plane model.

6. According to the method of steel labeling and coding based on a visual system according to claim 5, it is characterized in that: The specific processing process of the second steel material coding area positioning algorithm is as follows: S321: Use the RANSAC algorithm to preliminarily process the point cloud data of the effective steel point cloud image, iteratively optimize the plane model parameters to improve the fitting degree, identify and fit the main fitting plane in the angle steel and U-shaped steel point cloud, remove the noise and outliers of the plane model, and retain the point cloud data associated with the main fitting plane, wherein the main fitting plane is an area not less than 400mm*80mm; S322: Apply the PCA algorithm to the point cloud data preprocessed by the RANSAC algorithm, introduce the KD-Tree structure to organize the point cloud data, and divide the point cloud space into different areas. Use the PCA algorithm to calculate the point cloud normal vector, and further refine and extract the specific geometric features of the outer curved surface of the steel according to the normal vector; S323: After determining the outer curved surface of the steel material for labeling and coding, continue to use the OBB directional bounding box algorithm, combined with the average slope information of the outer surface of the steel material, to obtain a minimum circumscribed rectangular bounding box to cover the extracted slope. Within this bounding box, further locate and obtain the labeling and coding safety area and its coordinate information.

7. The steel labeling and coding method based on a visual system according to claim 6 is characterized in that: In step S323, the specific processing process is as follows: S3231: After determining the outer curved surface of the steel material to be labeled and coded, use the OBB directional bounding box algorithm; by analyzing the direction of the extracted outer curved surface of the steel material, use the OBB directional bounding box algorithm to generate a minimum circumscribed rectangular bounding box aligned with the direction of the surface; S3232: Calculate the average slope information of the outer surface of the steel material, and use the information to adjust the positioning of the OBB bounding box; finally, obtain a minimum rectangular bounding box that can tightly surround the outer curved surface of the steel material; S3233: In the smallest rectangular bounding box obtained in step S3232, a 400mm*80mm labeling and coding safety area is located inward from 30mm along the positive direction of the x-axis of the bounding box as the starting point, and its specific coordinates are extracted.

8. A steel labeling and coding method based on a visual system according to claim 1 or 7, characterized in that: The robot end effector comprises a vacuum suction cup and a spraying terminal. The vacuum suction cup is used to capture and place the labeling material, and the spraying terminal is used to perform the coding operation after the labeling is completed.

Images