Industrial mechanical arm spraying track automatic generation method applied to automobile spraying

Abstract

The invention relates to the technical field of automobile spraying, in particular to an industrial mechanical arm spraying track automatic generation method applied to automobile spraying, which comprises the following steps: firstly, adopting a Kinect depth camera to scan and generate automobile surface contour point cloud data in real time, then preprocessing the related point cloud data, extracting key feature points in an automobile surface contour, and finally, generating a spraying track of an industrial mechanical arm; and calculating contour points where the slices intersect with the automobile surface point cloud data based on a point cloud slicing method, generating a spraying path after the contour points are sequenced, finally calculating normal vectors of track points on the spraying path, determining the pose of the tail end of a spraying gun in the space, and circularly transmitting the pose obtained in real time to a control center. And the mechanical arm is controlled to move according to the generated spraying track. The method does not need to depend on an automobile surface model file during spraying, the automobile surface spraying track is autonomously planned and generated through the industrial mechanical arm in the automobile surface spraying process, dynamic adjustment can be conducted according to the actual situation of the automobile surface, and the accuracy of the whole spraying process is ensured.

Description

Technical Field

[0001] The invention relates to the technical field of automobile spraying, and in particular to a method for automatically generating spraying trajectories of an industrial robot arm applied to automobile spraying. Background Art

[0002] Spraying is an important part of the product manufacturing process and is widely used in the fields of automobiles, furniture, ships, aerospace, and military equipment. With the rapid development of advanced intelligent technologies such as robots and visual perception, since the 1950s, automation technologies represented by industrial robotic arms have gradually replaced manual spraying in the automotive spraying industry at home and abroad. Its advantages in automotive spraying operations are mainly reflected in: First, it can replace manual operation and can spray continuously, at high speed, and stably, greatly improving production efficiency and production speed; second, it can accurately control the thickness and uniformity of the spraying and improve the quality of the spraying; third, it can work in dangerous environments such as toxic, harmful, high temperature, and high pressure, reducing the risk of injury to operators and improving work safety.

[0003] Although my country's industrial robotic arms started late, they have now achieved development from small to medium-sized scales, and the product variety and technical level are also constantly improving. According to online data statistics: In 2022, the market size of domestic brand industrial robotic arms in my country reached 7.12 billion yuan, and foreign brands were close to 10.71 billion yuan. There is still a certain gap compared with foreign countries.

[0004] Insufficient and deficient existing products: There are two ways for industrial robotic arms to plan the trajectory of automobile spraying: manual teaching and traditional offline programming methods. The manual teaching method requires workers to manually guide the robotic arm to complete specific spraying paths and actions. It has disadvantages such as low programming efficiency, reliance on worker experience, and low paint utilization. It is only suitable for some simple spraying occasions. The other method based on traditional offline programming is currently the mainstream method. It is necessary to import the precise model of the automobile surface into the programming software, and then select appropriate spraying process parameters to generate the trajectory. However, this method has limitations. It relies on the precise three-dimensional model of the automobile surface to be sprayed, resulting in the inability to spray the automobile without ready-made automobile surface contour data. For example, due to the large number of models of major automobile brands and their differentiation, it is difficult for automobile repairs to have all model models, making it possible for some household automobile surface spraying after-sales repairs to be done only through manual teaching. Summary of the invention

[0005] The purpose of the present invention is to provide a method for automatically generating spray trajectories of an industrial robot arm used in automobile spraying, aiming to overcome the shortcomings of traditional off-line programming technology in automobile spraying that relies on workpiece model files, reconstruct the automobile surface model in real time, and detect the area to be sprayed by image recognition, so as to improve the efficiency and intelligence of the spraying process.

[0006] To achieve the above object, the present invention provides a method for automatically generating a spraying trajectory of an industrial robot arm applied to automobile spraying, comprising the following steps:

[0007] Step 1: Start the Kinect depth camera to scan the car surface profile data;

[0008] Step 2: Install a masking cloth on the non-spraying area of ​​the car, extract the color features of different areas on the car surface based on the color image collected by the Kinect depth camera, divide the spraying area, and determine the spraying area;

[0009] Step 3: Slice the point cloud data of the car surface in the spraying area to obtain the complete spraying path;

[0010] Step 4: Obtain the spray gun posture based on the contour points of the spray path, and realize automatic spraying through the EtherCAT communication protocol driver.

[0011] Optionally, the execution process of step 1 includes the following steps:

[0012] Step 1.1: Perform hand-eye calibration between the robotic arm and the Kinect depth camera to determine the coordinate transformation relationship between the robotic arm and the camera;

[0013] Step 1.2: Determine the scanning distance of the Kinect depth camera from the car surface.

[0014] Optionally, the process of determining the spraying area in step 2 includes the following steps:

[0015] Step 2.1: Extract car surface color features based on HSV color space;

[0016] Step 2.2: Set color threshold to segment the image;

[0017] Step 2.3: Use edge detection to determine the spraying area.

[0018] Optionally, the process of processing the automobile surface point cloud data in step 3 includes the following steps:

[0019] Step 3.1: Point cloud preprocessing, feature direction recognition and coordinate transformation of point cloud data in the spraying area;

[0020] Step 3.2: Point cloud slicing;

[0021] Step 3.3: Use the intersection method to calculate the slice contour points;

[0022] Step 3.4: Sort the slice contour points and fit the spraying path.

[0023] Optionally, the execution process of step 4 is specifically to calculate the normal vector based on the contour point obtained in step 3, and then offset the normal vector by a certain distance to obtain the end posture of the spray gun; finally, the solved end posture is cyclically sent from the host computer to the servo driver that controls the movement of the robotic arm through the EtherCAT communication protocol, so that the robotic arm can realize automatic spraying of the car surface according to the spraying trajectory planning.

[0024] The present invention provides an automatic generation method of spraying trajectory of an industrial robot arm applied to automobile spraying. First, a Kinect depth camera is used to scan and generate point cloud data of the automobile surface contour in real time, and then the relevant point cloud data is preprocessed to extract key feature points in the automobile surface contour. In view of the complex shapes of the automobile surface contour such as curved surface and concave-convex parts, the contour points where the slices intersect with the automobile surface point cloud data are calculated based on the point cloud slicing method, and the spraying path is generated after the contour points are sorted. Finally, the normal vector of the trajectory point on the spraying path is calculated, and the normal vector is offset by a certain distance to determine the position and posture of the spray gun end in space. The real-time acquired position and posture are cyclically transmitted to the control center, so that the industrial spraying robot arm is controlled according to the generated spraying trajectory. The present invention does not need to rely on the automobile surface model file during spraying. The automobile surface spraying trajectory is generated by the industrial robot arm in the process of automobile surface spraying, and can be dynamically adjusted according to the actual situation of the automobile surface, thereby ensuring the accuracy of the entire spraying process. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 The present invention is a schematic flow chart of the steps of a method for automatically generating a spraying trajectory of an industrial robot arm used for automobile spraying.

[0027] Figure 2 It is a schematic diagram of "eye outside the hand" selected for the hand-eye calibration of the present invention.

[0028] Figure 3 It is a schematic diagram of the outline of a part of the automobile surface point cloud slice of the present invention.

[0029] Figure 4 It is a comparative diagram of the projection method and the intersection method.

[0030] Figure 5 It is a schematic diagram of point cloud slicing of the surface model in the present invention.

[0031] Figure 6It is a schematic diagram of the arrangement of the slice intersection trajectories of the present invention.

[0032] Figure 7 It is a schematic diagram of spray gun posture selection of the present invention.

[0033] Figure 8 This is a schematic diagram of the EtherCAT master station communication process.

[0034] Fig. 9 It is a schematic diagram of the overall environment of a specific embodiment of the present invention.

[0035] Fig.10 It is a schematic diagram of automobile surface point cloud data according to a specific embodiment of the present invention.

[0036] Fig.11 It is a schematic diagram of the door spraying area of ​​a specific embodiment of the present invention.

[0037] Fig.12 It is a schematic diagram of a point cloud file of a door spraying area according to a specific embodiment of the present invention.

[0038] Fig.13 It is a schematic diagram of point cloud coordinate transformation according to a specific embodiment of the present invention.

[0039] Fig.14 It is a schematic diagram of obtaining the normal vector at the spray gun posture of a specific embodiment of the present invention.

[0040] Fig.15 It is a schematic diagram of spray path fitting of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0041] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0042] See also Figure 1 The present invention provides a method for automatically generating a spraying trajectory of an industrial robot arm used for automobile spraying, comprising the following steps:

[0043] S1: Start the Kinect depth camera to scan the car surface profile data;

[0044] S2: A masking cloth is installed on the non-spraying area of ​​the car, and the color features of different areas on the car surface are extracted based on the color image collected by the Kinect depth camera to divide the spraying area and determine the spraying area;

[0045] S3: Process the point cloud data of the car surface in the spraying area to obtain the complete spraying path;

[0046] S4: The spray gun posture is obtained based on the contour points of the spray path, and automatic spraying is achieved through the EtherCAT communication protocol driver.

[0047] The following is a further explanation based on the execution steps:

[0048] The step S1 of starting the depth camera to scan the surface profile data of the car includes:

[0049] (1) Hand-eye calibration of the robotic arm and Kinect depth camera

[0050] Since the end of the robot arm carries a spray gun, the Kinect depth camera is installed near the base of the robot arm. That is, the hand-eye calibration selects the "eye outside the hand" method, such as Figure 2 shown.

[0051] The purpose of hand-eye calibration is to determine the coordinate transformation relationship between the robot arm and the camera, that is, to determine the transformation matrix base H cam First, obtain the transformation matrix between the calibration plate coordinate system and the camera coordinate system through camera calibration cam H cal ; Transformation matrix between tool coordinate system and robot base coordinate system base H tool Obtained from the kinematic analysis of the robot arm; since the calibration plate coordinate system and the tool coordinate system keep the same posture when the robot arm moves, the transformation matrix between the two coordinate systems is tool H cal remains unchanged, so the robot arm holds the calibration plate and moves to two positions at will. tool H cal Keeping it unchanged, we can solve base H cam , as shown in formula (1).

[0052] tool H base1 · base1 H cam1 · cam1 H cal = tool H base2 · base2 H cam2 · cam2 H cal (1)

[0053] By shifting the terms in formula (1) left and right, we can get formula (2):

[0054]

[0055] Since the relative position of the camera coordinate system and the robot base coordinate system remains unchanged, the coordinate transformation matrix between the two remains stable and can be regarded as a fixed parameter of the system, that is, base H cam1 and base H cam2 are equal and will not change, so formula (2) can be simplified to:

[0056] AX=XB (3)

[0057] Among them, X is the transformation matrix between the camera coordinate system and the robot base coordinate system, that is, base H cam Transformation matrix.

[0058] (2) Determine the scanning distance between the camera and the car surface

[0059] The scanning distance is determined based on the depth camera obtaining the current depth image D(x,y), where D(x,y) represents the depth value of the depth image at the position (x,y). By setting a target depth threshold, only images within the threshold are required, so that the limit of the scanning distance can be determined. After the scanning distance is determined, the Kinect camera is used to collect color images, depth images, and point cloud data of the car surface.

[0060] The spraying area division of step S2 includes:

[0061] (1) Car surface color feature extraction

[0062] Since the HSV color space is fuller and richer than the RGB color space, in order to facilitate the subsequent extraction of automobile surface color features, the collected RGB colors need to be converted into HSV colors. The conversion process is as follows:

[0063] Normalize the RGB values ​​and divide the R, G, and B component values ​​by 255 to get values ​​between 0 and 1:

[0064]

[0065] Calculate the maximum and minimum values:

[0066]

[0067] Calculate hue (H):

[0068]

[0069] Calculate saturation (S):

[0070]

[0071] Calculate the brightness (V):

[0072] V=C max (8)

[0073] After the color space conversion, color features of different regions are extracted based on color aggregation vectors. That is, the image is divided into several connected regions according to the connectivity between pixels, and then the number of pixels in each independent connected domain C is counted. By setting several color threshold intervals, it is determined whether the pixels in C are aggregated or non-aggregated. The aggregated and non-aggregated regions are the color features of different regions.

[0074] Set color threshold to segment the image

[0075] After the color features of the car surface are extracted, the color features after the above color aggregation can be segmented by setting an appropriate color threshold to achieve image segmentation. If the spray area to be segmented is gray, the HSV threshold limit is set as follows, that is, only the area within this color threshold is required to achieve segmentation.

[0076]

[0077] (3) Determination of spraying area

[0078] Edge detection is the last step in determining the spraying area. Currently, the commonly used edge detection algorithms include Canny operator, Laplace operator and Sobel operator. The Canny operator uses two different thresholds to detect strong edges and weak edges, and has strong robustness, so it is selected for the edge detection task of the spraying area. The Canny algorithm is mainly divided into the following four steps for edge detection:

[0079] ① Using Gaussian filtering to remove noise can effectively reduce the interference of pseudo edges.

[0080] ② Calculate the image gradient in the x and y directions to obtain a collection of potential image edges.

[0081] ③ Non-maximum suppression: retain the directions with large pixel value changes, eliminate the directions with small pixel value changes, and refine the edges.

[0082] ④Threshold screening: Edge screening is performed by setting two thresholds. Values ​​greater than the high threshold are directly judged as strong edges, and values ​​less than the low threshold are directly excluded. Edges between the two are also considered edges if they are strong edges, otherwise they are non-edges.

[0083] The processing of the automobile surface point cloud data in step S3 includes:

[0084] (1) Point cloud preprocessing

[0085] First, the feature direction and coordinate transformation of the point cloud data in the spraying area are performed to facilitate the subsequent point cloud slicing operation. The essence of the point cloud slicing algorithm is: given a point cloud set P = {P1, P2, ... Pn}, a plane is made along a given direction and with it as the normal vector, so as to calculate the contour line of the point cloud set P located on any plane E. In view of this, the center of gravity, eigenvalues ​​and eigenvectors of the point cloud can be obtained by calculating the covariance matrix of the point cloud data points, and then the matrix transformation is performed to adjust the posture of the point cloud so that the direction of the main eigenvector is parallel to a certain coordinate axis, and finally the center of mass coordinates are moved to coincide with the origin. The specific algorithm flow is as follows.

[0086] ① Calculate the centroid of the point cloud set P

[0087]

[0088] ②Decentralize the data and obtain a new data set p':

[0089]

[0090] ③ Calculate the covariance matrix C of the decentralized data:

[0091]

[0092] ④ Decompose the covariance matrix to obtain a 3×1 matrix V consisting of eigenvalues a and a 3×3 matrix V consisting of the corresponding eigenvectors e , the directions pointed by these three vectors are the three directions where the point cloud features are most obvious.

[0093] C.V. e =V a ·V e (13)

[0094] ⑤ From the characteristics of the covariance matrix, we can know that the covariance matrix is ​​a semi-positive definite matrix whose eigenvectors are perpendicular to each other, so the eigenvectors of the point cloud can be used to rotate the point cloud so that the three eigenvectors are parallel to the coordinate axis and the center of mass of the point cloud is translated to the origin of the coordinate system. The transformation matrix is ​​shown in formula (14), and the center of mass translation is shown in formula (15).

[0095]

[0096] In the formula is the transposed matrix of the eigenvector matrix, p c is the centroid coordinate. After the above transformation matrix processing, the point cloud can be transformed to a position that is conducive to slicing processing.

[0097] (2) Point cloud slicing

[0098] Select m planes E with equal spacing and parallel to each other from the preprocessed point cloud data. i (i=1,2,···,m), passing through plane E according to the normal vector direction of the point cloud data i To intercept the point cloud model of the car surface, m-1 data bands can be obtained, and the strip data can be used to calculate the point cloud section profile on the slice. The point cloud slice profile of some car surfaces is as follows: Figure 3 shown.

[0099] (3) Calculation of contour points in slices

[0100] There are two methods to calculate the data points on the point cloud slice: projection method and intersection method, such as Figure 4 As shown. The projection method is to project the point clouds on the left and right sides of the slice onto the slice, and the projection points are the contour points of the slice. The intersection method is to match the nearest points on the left and right sides of the slice, and the intersection of the line connecting the two points and the slice is used as the contour point. Since the projection method will have a large deviation when projecting at a point far away from the slice, the intersection method is used to calculate the slice contour points.

[0101] (4) Slice contour point sorting

[0102] The spatial arrangement of the point cloud data is disordered, and the calculated slice contour points need to be sorted to fit a spraying path. Assume that the point cloud slice direction is along the y-axis, such as Figure 5 As shown in the figure, since the y values ​​of the point cloud data in the same slice are equal, the sorting of the point cloud data can be converted into the sorting of the data on the plane, that is, the sorting of the point cloud slice data points in ascending order of the x-axis coordinates, as shown in the figure. Figure 6 shown.

[0103] Step S4 includes the process of obtaining the spray gun posture and posture transmission, which specifically includes:

[0104] (1) Obtaining the spray gun posture

[0105] It can be seen from the spraying process that when spraying the workpiece surface, the spray gun direction is generally perpendicular to the workpiece surface. Therefore, the normal vector of the contour point in the point cloud slice can be used as the direction of the spray gun axis, and then the contour point can be offset along the normal line by a certain height to obtain the end position of the spray gun, such as Figure 7 As shown, where n p is the normal vector of point P, and D is the offset distance.

[0106] When solving the normal vector of the projection point in the point cloud slice, the normal vector can be obtained by selecting the nearest neighboring points of the point to fit a local small plane. Suppose the plane is a point cloud dataset containing N points, and each point p i There are three coordinates (x i,y i ,z i ), where i = 1, 2, ..., N. First, the coordinates of each point are centered to obtain the centralized coordinates, as shown in formula (16).

[0107]

[0108] In the formula is the mean coordinate value in the point cloud dataset.

[0109] Next, the covariance matrix C of the centered coordinates is calculated using the formula shown in formula (17), and then the eigenvalue decomposition is performed to solve the corresponding eigenvector, and finally the eigenvector with the smallest eigenvalue is selected as the normal vector of the local microplane.

[0110]

[0111] The normal vector obtained only determines the line where the normal vector is located, and the normal vector needs to be oriented. p , any point p i , whose normal vector is n i , satisfying formula (18) can make the normal vector of the point point toward the viewpoint, thereby determining the direction of the normal vector.

[0112] n i ·(V p -p i )>0 (18)

[0113] (2) Spray gun position transmission

[0114] Because the bottom of the robot moves with the car, the obtained posture needs to be solved based on the world coordinates using the kinematic and dynamic equations of the robot to calculate the position and posture of the spray gun. In the host computer, the solved spray gun posture is transmitted to the robot motion controller built in the host computer through the geometry_msgs / Pose message type in ROS. The controller then packages the posture data into EtherCAT frames and transmits it to the servo driver through the EtherCAT communication protocol. After the servo driver parses the posture data, it sends the position control command to the servo motor, so that the robot moves according to the planned spray trajectory. The robot posture message format is shown in Table 1, and the EtherCAT communication process is shown in Table 1. Figure 8 shown.

[0115] Table 1 Gun posture message format

[0116]

[0117] Further, the following is a supplementary description in conjunction with specific embodiments:

[0118] The overall environment of this embodiment is shown as follows Fig. 9 shown.

[0119] In step 1 of this embodiment, after the robot arm and the camera complete the hand-eye calibration, the Kinect camera is started to collect the surface data of the car. The mobile car at the bottom of the robot arm moves around the car in a "mouth" shape to scan the surface of the car to obtain complete point cloud data. The surface point cloud data of the car is as follows: Fig.10 However, in actual painting tasks, it is generally not necessary to scan the car all the way around. For example, in after-sales maintenance of car painting, it is only necessary to control the bottom moving trolley to move to the paint touch-up area on the car surface, and then the camera scans, which greatly reduces the painting task time.

[0120] The spraying area division in step 2 also includes the following three steps:

[0121] (1) Extracting color features of the vehicle surface based on the color image scanned by the depth camera;

[0122] (2) Convert the RGB color space into the HSV color space, and set the color threshold based on the converted HSV color space to segment the image;

[0123] (3) The spraying area is determined, and the Canny algorithm is selected for edge detection. The selected door spraying area is shown in Figure 11. The effect of the point cloud file is shown in Figure 11. Fig.12 shown.

[0124] Step 3 is to process the point cloud data of the automobile surface. In the point cloud preprocessing process, after the transformation matrix processing, the point cloud can be transformed to a position that is conducive to slicing processing, such as Fig.13 shown.

[0125] Then, the point cloud is sliced ​​and the contour points in the slices are calculated to complete the sorting of the slice contour points.

[0126] From the spraying process in step 4, it can be known that when spraying the surface of the workpiece, the direction of the spray gun perpendicular to the surface of the workpiece is generally used. Therefore, when spraying the car door, the direction of the spray gun axis can be determined according to the normal vector of the contour point in the point cloud slice, and then the contour point is offset along the normal to a certain height to obtain the end posture of the spray gun. After the end spray gun posture of each contour point is connected, spraying can be carried out according to the spray trajectory generated after the above point cloud slicing. In addition, a path point of about 3 cm is added to both ends of the contour line obtained by slicing to ensure the integrity of the spraying along the trajectory. The spray gun posture is obtained as follows Fig.14 As shown, the spraying trajectory planning is as follows Fig.15The specific spraying process is as follows: when the spray gun moves along the first row of planned trajectory to the spraying edge, it continues to spray according to the added path points to ensure that the spraying above the door is complete; after the spraying of the first row of planned trajectory is completed, the spray gun position moves down according to the point cloud slice interval d to spray the second row of planned trajectory, and the spraying direction is opposite to the first row of planned trajectory, and it is circulated in sequence until the spraying of the car door is completed.

[0127] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0128] (1) The present invention enables an industrial robot arm to autonomously plan and generate a spraying trajectory for the surface of a car during the spraying process. By combining image recognition technology to accurately detect and divide the area to be sprayed, and performing downsampling, eigenvalue solving, coordinate transformation and other processing on the point cloud data in the spraying area, the key feature points in the spraying area are automatically extracted for point cloud slicing to generate a complete spraying path.

[0129] (2) The point cloud slicing technology used in the present invention to generate the spray trajectory can accurately plan the spray path on automobile surfaces of different shapes and sizes, such as planes, curved surfaces, and complex geometric shapes. This flexibility enables it to perform well in various spraying tasks and meet the needs of different models and components. In conjunction with EtherCAT communication technology, the trajectory planning results are sent to the robot arm controller in real time. Even if the position or posture of the workpiece changes, it can be dynamically adjusted according to the actual situation of the automobile surface, ensuring the accuracy of the entire spraying process.

[0130] What is disclosed above is only a preferred embodiment of the present invention, and it certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for automatically generating spraying trajectories of industrial robot arms used for automobile spraying, characterized in that: The following steps are involved: Step 1: Start the Kinect depth camera to scan the car surface profile data; Step 2: Install a masking cloth on the non-spraying area of ​​the car, extract the color features of different areas on the car surface based on the color image collected by the Kinect depth camera, divide the spraying area, and determine the spraying area; Step 3: Slice the point cloud data of the car surface in the spraying area to obtain the complete spraying path; Step 4: Obtain the spray gun posture based on the contour points of the spray path, and realize automatic spraying through the EtherCAT communication protocol driver.

2. The method for automatically generating spraying trajectories of an industrial robot arm for automobile spraying as claimed in claim 1, characterized in that: The execution process of step 1 includes the following steps: Step 1.1: Perform hand-eye calibration between the robotic arm and the Kinect depth camera to determine the coordinate transformation relationship between the robotic arm and the camera; Step 1.2: Determine the scanning distance of the Kinect depth camera from the car surface.

3. The method for automatically generating spraying trajectories of industrial robot arms for automobile spraying as claimed in claim 2, characterized in that: The process of determining the spraying area in step 2 includes the following steps: Step 2.1: Extract car surface color features based on HSV color space; Step 2.2: Set color threshold to segment the image; Step 2.3: Use edge detection to determine the spraying area.

4. The method for automatically generating spraying trajectories of industrial robot arms for automobile spraying as claimed in claim 3, characterized in that: The process of processing the automobile surface point cloud data in step 3 includes the following steps: Step 3.1: Point cloud preprocessing, feature direction recognition and coordinate transformation of point cloud data in the spraying area; Step 3.2: Point cloud slicing; Step 3.3: Use the intersection method to calculate the slice contour points; Step 3.4: Sort the slice contour points and fit the spraying path.

5. The method for automatically generating spraying trajectories of industrial robot arms for automobile spraying as claimed in claim 4, characterized in that: The execution process of step 4 is to calculate the normal vector based on the contour point obtained in step 3, and then offset the normal vector by a certain distance to obtain the end position of the spray gun; finally, the solved end position is sent from the host computer to the servo driver that controls the movement of the robotic arm through the EtherCAT communication protocol, so that the robotic arm can realize automatic spraying of the car surface according to the spraying trajectory planning.

Images