Motor vehicle tail gas detection intelligent intubation control method and system based on image recognition

Abstract

The invention relates to the technical field of tail gas pipe insertion control, in particular to a motor vehicle tail gas detection intelligent pipe insertion control method and system based on image recognition. In order to solve the technical problems of low intellectualization and poor accuracy of motor vehicle tail gas intubation detection in the prior art, two-dimensional coordinates of an automobile tail image are converted into three-dimensional coordinates of a mechanical arm, the position of an exhaust pipe in a working space of the mechanical arm is determined, and a moving path is planned in combination with angle information of a robot mechanical arm joint; then, according to the moving path of the mechanical arm of the robot, a tail gas collecting pipe at the tail end of the mechanical arm is moved to directly face an exhaust pipe, a segmented speed control method is adopted in the tail gas collecting pipe inserting process, the inserting pressure of the tail gas collecting pipe is obtained in real time, and when the pressure is larger than a pressure threshold value, the measures of pulling out and then inserting are taken; the intubation control method is high in adaptive capacity and high in intelligence, high automation of the tail gas detection intubation process can be achieved, and the tail gas detection accuracy and efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of exhaust pipe insertion control, and in particular to an intelligent intubation control method and system for motor vehicle exhaust detection based on image recognition. Background Art

[0002] During the vehicle exhaust detection process, it is usually necessary to insert the exhaust collection tube into the vehicle exhaust pipe. At present, this operation is mainly completed by manual control, which not only consumes human resources, but also has difficulty in ensuring the insertion depth, resulting in large errors, affecting the accuracy of the exhaust detection results. In addition, the operator frequently approaches the exhaust outlet and inhales exhaust pollutants, which also has a significant adverse effect on personnel health. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent intubation control method and system for motor vehicle exhaust gas detection based on image recognition.

[0004] The technical solution of the present invention is as follows: An intelligent intubation control method for motor vehicle exhaust detection based on image recognition includes the following operations: S1, obtaining a rear image of a car, and identifying an exhaust pipe from the rear image of the car; S2. After converting the two-dimensional coordinates of the image of the rear of the car into the three-dimensional coordinates of the robot arm, path planning is performed based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain the moving path of the robot arm; the method for obtaining the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm is specifically as follows: based on the intrinsic parameters and extrinsic parameters of the camera, a camera imaging model and a conversion relationship model between the camera coordinate system and the robot arm coordinate system are constructed; after substituting the two-dimensional coordinates of the exhaust pipe into the camera imaging model, the conversion relationship model is processed by the least squares method to obtain the world vertical coordinate of the exhaust pipe; the world vertical coordinate of the exhaust pipe and the two-dimensional coordinate of the exhaust pipe are substituted into the camera imaging model for solution to obtain the world horizontal coordinate and the world vertical coordinate of the exhaust pipe; the world horizontal coordinate, the world vertical coordinate and the world vertical coordinate of the exhaust pipe form the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm; S3. According to the moving path of the robot arm, move the exhaust gas collection pipe at the end of the robot arm to face the exhaust pipe; insert the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold using a segmented speed control method in a preset insertion direction; during the insertion of the exhaust gas collection pipe, obtain the insertion pressure of the exhaust gas collection pipe in real time; if the insertion pressure of the exhaust gas collection pipe is greater than the pressure threshold, pull out the exhaust gas collection pipe in a direction opposite to the insertion direction to the pull-out depth threshold, and then execute the exhaust gas collection pipe insertion action.

[0005] The operation of obtaining the moving path of the robot arm in S2 is specifically as follows: the initial angles of the joints of the robot arm are used as the root nodes of the random tree to obtain the initial random tree; the initial random tree is iteratively processed by node expansion until the distance between the generated new node and the target angle point is less than the distance threshold, and the iterative processing ends; the new node whose distance to the target angle point is less than the distance threshold is used as the target node; in the process of tracing back from the target node to the root node, the edges between the nodes in the random tree are traversed to obtain several motion paths; based on the path cost function, several motion paths are screened, and the motion path with the smallest path cost value is used as the moving path of the robot arm; the target angle point is the point at the end of the robot arm corresponding to the standard exhaust pipe position.

[0006] The path cost function is obtained by the following formula: , Q For the path value, d(p i ,p i+1 ) For Node p i With Node p i+1 The distance between i i+1 For Node p i+1 Angle, i i For Node p i Angle, q e According to the energy consumption of the robot arm when moving along the motion path, I is the total number of nodes in the motion path.

[0007] During the iterative processing of node expansion, the specific operation of the current node expansion processing is as follows: in the random tree generated in the previous round, a leaf node is randomly generated within the range of the minimum moving angle and the maximum moving angle of each root node to obtain several current transition leaf nodes; the nodes in the random tree generated in the previous round that are closest to the corresponding current transition leaf nodes are used as the current nearest leaf nodes corresponding to each root node; the current nearest leaf nodes are moved along the direction of the corresponding current transition leaf node after the step size threshold is reached, and are used as the current new nodes of the corresponding root nodes; the current new nodes are connected to the corresponding current nearest leaf nodes to implement the update iteration of the random tree, and the random tree generated in the current round is obtained, which is used to perform the operations of the next round of node expansion processing.

[0008] In S1, the operation of identifying the exhaust pipe from the car rear image is specifically as follows: marking the edge position points in the car rear image to obtain a marked car rear image; covering the marked car rear image with detection frames, and taking the detection frame corresponding to the minimum distance between the detection center point and the center point of the real frame in the same area as the classification frame of the corresponding area; classifying all classification frames based on the shape, pixel distribution and area of ​​the exhaust pipe, and taking the classification result of the exhaust pipe and the classification frame whose corresponding classification probability value is greater than the probability threshold as the exhaust pipe identification frame; the exhaust pipe identification frame is processed by Hough transform to obtain the center position of the exhaust pipe; the coordinates of the center position of the exhaust pipe in the car rear image are taken as the two-dimensional coordinates of the exhaust pipe.

[0009] The method for obtaining edge position points is specifically as follows: grayscale processing is performed on the image of the rear of the car to obtain a grayscale image of the car; a weighted average of the pixels of each position point and other position points within a first neighborhood range in the grayscale image of the car is used as a new pixel value of the corresponding position point to obtain a denoised image of the car; it is determined whether the pixel value of each position point in the denoised image of the car is greater than an edge pixel threshold; if the pixel value of the current position point is not less than the first edge pixel threshold, the current position point is an edge position point; if the pixel value of the current position point is between the first edge pixel threshold and the second edge pixel threshold, and the current position point is adjacent to a determined edge position point, the current position point is an edge position point; the first edge pixel threshold is greater than the second edge pixel threshold.

[0010] An intelligent intubation control system for motor vehicle exhaust detection based on image recognition, used to implement the above-mentioned intelligent intubation control method for motor vehicle exhaust detection based on image recognition, comprising: A robot arm, with an exhaust gas collection pipe installed at the end, for executing the action of moving the exhaust gas collection pipe according to the instructions transmitted by the control unit; A camera, mounted on the robot arm, is used to obtain images of the rear of the car; An image processing unit, used for processing the image of the rear of the car acquired by the camera, and identifying the exhaust pipe from the image of the rear of the car; A computing unit, configured to convert the two-dimensional coordinates of the rear image of the vehicle into three-dimensional coordinates of the robot arm, and perform path planning based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain a moving path of the robot arm; The control unit is used to move the exhaust gas collection pipe at the end of the robot arm to face the exhaust pipe according to the movement path of the robot arm in the calculation unit; insert the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold in a preset insertion direction and using a segmented speed control method; and when the exhaust gas collection pipe insertion pressure is greater than the pressure threshold during the exhaust gas collection pipe insertion process, pull out the exhaust gas collection pipe in a direction opposite to the insertion direction to the pull-out depth threshold, and then perform the exhaust gas collection pipe insertion action; A pressure sensor for obtaining the tail gas collection pipe insertion pressure; A depth sensor, used to obtain the depth of the exhaust gas collecting pipe inserted into the exhaust pipe; The exhaust gas collection pipe fixing clamp is installed at the end of the robot arm and is used to fix the exhaust gas collection pipe.

[0011] Among them, there are two cameras, one camera is installed on the main pile fixed by the robot mechanical arm to obtain the car image, and the other camera is installed at the end of the robot mechanical arm to obtain the car tail image when the car tail exists in the car image.

[0012] The beneficial effects of the present invention are as follows: the present invention provides an intelligent intubation control method for exhaust gas detection of motor vehicles based on image recognition, which first identifies the exhaust pipe through image recognition to determine the position and shape of the exhaust pipe at the rear of the car; then, by converting the two-dimensional coordinates of the image of the rear of the car into the three-dimensional coordinates of the robotic arm, the position of the exhaust pipe in the working space of the robotic arm is determined, and the moving path is planned in combination with the joint angle information of the robot robotic arm, providing clear motion instructions for the robotic arm; finally, according to the moving path of the robot robotic arm, the exhaust gas collecting pipe at the end of the mobile robotic arm is directly opposite to the exhaust pipe, and a segmented speed control method is adopted during the insertion of the exhaust gas collecting pipe, which can better control the insertion accuracy during the insertion process, and at the same time, the insertion pressure of the exhaust gas collecting pipe is obtained in real time, and the measures of pulling out and reinserting are taken when the pressure is greater than the pressure threshold, which can effectively prevent the exhaust gas collecting pipe or the exhaust pipe from being damaged due to excessive insertion force; the intubation control method has strong adaptability and high intelligence, can realize the high automation of the exhaust gas detection intubation process, reduces the tediousness and uncertainty of manual operation, and can improve the accuracy and efficiency of exhaust gas detection. DETAILED DESCRIPTION

[0013] This embodiment provides an intelligent intubation control system for motor vehicle exhaust detection based on image recognition, which is used to implement an intelligent intubation control method for motor vehicle exhaust detection based on image recognition, including: A robot arm, with an exhaust gas collection pipe installed at the end, for executing the action of moving the exhaust gas collection pipe according to the instructions transmitted by the control unit; A camera is installed on the robot arm to obtain an image of the rear of the car. There are two cameras, one of which is installed on the main fixed pile of the robot arm to obtain an image of the car, and the other is installed at the end of the robot arm to obtain an image of the rear of the car when the image processing unit outputs an image of the car with the rear of the car in it, so as to improve recognition accuracy. An image processing unit, used for processing the image of the rear of the car acquired by the camera, and identifying the exhaust pipe from the image of the rear of the car; A computing unit, configured to convert the two-dimensional coordinates of the rear image of the vehicle into three-dimensional coordinates of the robot arm, and perform path planning based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain a moving path of the robot arm; A control unit is used to move the exhaust gas collection pipe at the end of the robot arm to face the exhaust pipe according to the movement path of the robot arm in the calculation unit; insert the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold in a preset insertion direction and using a segmented speed control method; and when the exhaust gas collection pipe insertion pressure is greater than the pressure threshold during the exhaust gas collection pipe insertion process, the exhaust gas collection pipe is pulled out in a direction opposite to the insertion direction until the pull-out depth threshold, or the exhaust gas collection pipe is pulled out until the pull-out depth reaches the corresponding pull-out depth threshold, and then the exhaust gas collection pipe insertion action is performed; A pressure sensor for obtaining the tail gas collection pipe insertion pressure; A depth sensor, used to obtain the depth of the exhaust gas collecting pipe inserted into the exhaust pipe; The tail gas collection pipe fixing clamp is installed at the end of the robot arm and is used to fix the tail gas collection pipe; The exhaust gas collection pipe retracting device is installed at the end of the robot arm and is connected to the exhaust gas collection pipe. It is used for automatic retraction of the exhaust gas collection pipe during the movement of the robot arm.

[0014] This embodiment also provides an intelligent intubation control method for motor vehicle exhaust detection based on image recognition, and the specific process is as follows.

[0015] S1. Acquire a rear image of a car, and identify an exhaust pipe from the rear image of the car.

[0016] By using image recognition to identify the exhaust pipe, the position and shape of the exhaust pipe at the rear of the car can be accurately determined, thereby improving the degree of automation and helping to improve the accuracy of subsequent intubation.

[0017] The specific operation of identifying the exhaust pipe from the car rear image is as follows: marking the edge position points in the car rear image to obtain a marked car rear image that is easy to accurately and quickly identify; covering the marked car rear image with detection frames, and using the detection frame corresponding to the minimum distance between the detection center point and the center point of the real frame in the same area as the classification frame of the corresponding area; based on the shape, pixel distribution and area of ​​the exhaust pipe, all classification frames are classified, and the classification frame with the classification result of exhaust pipe and the corresponding classification probability value greater than the probability threshold is used as the exhaust pipe identification frame; the exhaust pipe identification frame is processed by Hough transform to obtain the center position of the exhaust pipe; the coordinates of the center position of the exhaust pipe in the car rear image are used as the two-dimensional coordinates of the exhaust pipe.

[0018] The method for obtaining edge position points is specifically as follows: grayscale the image of the rear of the car to obtain a grayscale image of the car; the weighted average of the pixels of each position point and other position points within the first neighborhood range in the grayscale image of the car is used as the new pixel value of the corresponding position point to obtain a denoised image of the car; determine whether the pixel value of each position point in the denoised image of the car is greater than the edge pixel threshold; if the pixel value of the current position point is not less than the first edge pixel threshold, the current position point is an edge position point; if the pixel value of the current position point is between the first edge pixel threshold and the second edge pixel threshold, and the current position point is adjacent to the determined edge position point, the current position point is an edge position point. The first edge pixel threshold is greater than the second edge pixel threshold.

[0019] The method for obtaining edge position points can also be specifically as follows: grayscale the image of the rear of the car to obtain a grayscale image of the car; use the weighted average of the pixels of each position point and other position points within the first neighborhood range in the grayscale image of the car as the new pixel value of the corresponding position point to obtain a denoised image of the car; obtain the pixel gradient amplitude and pixel gradient direction of each position point in the denoised image of the car; compare the pixel gradient amplitude of each position point with the gradient amplitudes of all adjacent pixel points within the second neighborhood range and along the corresponding pixel gradient direction; if the pixel gradient amplitude of the current position point is the maximum value, the current position point is an edge position point.

[0020] S2. After converting the two-dimensional coordinates of the rear image of the car into the three-dimensional coordinates of the robot arm, path planning is performed based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain the movement path of the robot arm.

[0021] By converting the two-dimensional coordinates of the car's rear image into the three-dimensional coordinates of the robotic arm, the position of the exhaust pipe in the robotic arm's workspace can be accurately determined, and the movement path can be planned in combination with the robot arm's joint angle information, providing the robotic arm with clear motion instructions, achieving a high degree of automation in the exhaust gas detection intubation process, reducing the tediousness and uncertainty of manual operation, and improving the degree of automation of the entire detection process.

[0022] First, the two-dimensional coordinates of the car's rear image are converted into the three-dimensional coordinates of the robot's working arm, and the two-dimensional coordinates of the exhaust pipe are converted into the three-dimensional coordinates of the robot's mechanical arm. The position information of the exhaust pipe in the three-dimensional coordinates of the robot's mechanical arm is obtained, which is convenient for the subsequent planning of the robot's mechanical arm's moving path.

[0023] Among them, the method for obtaining the position information of the exhaust pipe in the three-dimensional coordinates of the robot mechanical arm is specifically as follows: based on the intrinsic parameters and extrinsic parameters of the camera, a camera imaging model and a conversion relationship model between the camera coordinate system and the robot mechanical arm coordinate system (world coordinate system) are constructed; after substituting the two-dimensional coordinates of the exhaust pipe into the camera imaging model, the conversion relationship model is processed by the least squares method to obtain the world vertical coordinate of the exhaust pipe; the world vertical coordinate of the exhaust pipe and the two-dimensional coordinate of the exhaust pipe are substituted into the camera imaging model for solution to obtain the world horizontal coordinate of the exhaust pipe and the world vertical coordinate of the exhaust pipe; the world horizontal coordinate of the exhaust pipe, the world vertical coordinate of the exhaust pipe and the world vertical coordinate of the exhaust pipe form the position information of the exhaust pipe in the three-dimensional coordinates of the working robot mechanical arm.

[0024] Then, based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm, the movement path of the robot arm is obtained.

[0025] The operation of obtaining the moving path of the robot manipulator is specifically as follows: the initial angles of each joint of the robot manipulator are used as the root nodes of the random tree to obtain the initial random tree; the initial random tree is iteratively processed by node expansion until the distance between the generated new node and the target angle point is less than the distance threshold, and the iterative processing ends; the new node whose distance to the target angle point is less than the distance threshold is used as the target node; in the process of tracing back from the target node to the root node, the edges between the nodes in the random tree are traversed to construct a motion path from the starting state to the target state, and several motion paths are obtained; based on the path cost function, several motion paths are screened, and the motion path with the smallest path cost value is used as the moving path of the robot manipulator.

[0026] The above-mentioned target angle point is the point at the end of the robotic arm corresponding to the position of the standard exhaust pipe, that is, the position point where the exhaust gas collection pipe at the end of the robot robotic arm is opposite to the exhaust pipe. It is also the initial position point of the end of the robot robotic arm when performing the insertion process. The target angle point can be calculated based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot robotic arm.

[0027] In the above-mentioned node expansion iterative processing, the operation of the current node expansion processing is specifically as follows: in the random tree generated in the previous round, a leaf node is randomly generated within the range of the minimum moving angle and the maximum moving angle of each root node to obtain several current transition leaf nodes; each current transition leaf node corresponds to its own root node; the node closest to the corresponding current transition leaf node in the random tree generated in the previous round is used as the current nearest leaf node corresponding to each root node; the current nearest leaf node is moved along the direction of the corresponding current transition leaf node by the step threshold, and is used as the current new node of the corresponding root node; the current new node is connected to the corresponding current nearest leaf node (the current new node belonging to the same root node is connected to the current nearest leaf node), and the random tree generated in the current round is updated and iterated to obtain the random tree for performing the next round of node expansion processing operations. The step threshold is obtained based on the length of the robot arm and the working angle range.

[0028] At the same time, the path cost function is obtained by the following formula: , Q For the path value, d(p i ,p i+1 ) For Node p i With Node p i+1 The distance between i i+1 For Node p i+1 Angle, i i For Node p i Angle, q e According to the energy consumption of the robot arm when moving along the motion path, I is the total number of nodes in the motion path.

[0029] The operation for obtaining the moving path of the robot mechanical arm can also be: determine whether the position point of the exhaust pipe in the three-dimensional coordinates of the robot mechanical arm is within the moving range of the end of the robot mechanical arm; if so, the line connecting the position point of the exhaust pipe in the three-dimensional coordinates of the robot mechanical arm and the position point at the end of the mechanical arm corresponding to the standard exhaust pipe position is used as the moving path of the robot mechanical arm; if not, projecting the initial position point of the end of the robot mechanical arm onto the point on the horizontal plane where the position point of the exhaust pipe in the three-dimensional coordinates of the robot mechanical arm is located as the middle point, and the 90° broken line formed by the initial position point of the end of the robot mechanical arm, the middle point and the position point at the end of the mechanical arm corresponding to the standard exhaust pipe position is used as the moving path of the robot mechanical arm.

[0030] S3. According to the moving path of the robot arm, move the exhaust gas collection pipe at the end of the robot arm to face the exhaust pipe; insert the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold using a segmented speed control method in a preset insertion direction; during the insertion of the exhaust gas collection pipe, obtain the insertion pressure of the exhaust gas collection pipe in real time; if the insertion pressure of the exhaust gas collection pipe is greater than the pressure threshold, pull out the exhaust gas collection pipe in a direction opposite to the insertion direction to the pull-out depth threshold, and then execute the exhaust gas collection pipe insertion action.

[0031] According to the moving path of the robot manipulator, the exhaust gas collection pipe at the end of the mobile manipulator is facing the exhaust pipe, and a segmented speed control method is used during the insertion of the exhaust gas collection pipe, which can better control the insertion accuracy during the insertion process. At the same time, the insertion pressure of the exhaust gas collection pipe can be obtained in real time, and the measures of pulling out and reinserting can be taken when the pressure is greater than the pressure threshold. This can effectively prevent the exhaust gas collection pipe or the exhaust pipe from being damaged due to excessive insertion force, improve the adaptability of the insertion method, enhance the ability to cope with complex situations, and improve the intelligence of the intubation method.

[0032] First, according to the moving path of the robot arm, the exhaust gas collection pipe at the end of the robot arm is moved to face the exhaust pipe, that is, the end of the robot arm is moved to the target angle point so that the exhaust gas collection pipe faces the exhaust pipe.

[0033] Then, in a preset insertion direction, the exhaust gas collection tube is inserted into the exhaust pipe to a preset depth threshold (preferably 350mm) using a segmented speed control method. In the segmented speed control method, the principle of decreasing speed is adopted, and the speed when approaching the exhaust pipe (such as 1 - 2cm / s) is greater than the speed during the insertion process (such as 0.5 - 1cm / s). In the stage of approaching the exhaust pipe, a relatively fast speed can make the exhaust gas collection tube quickly approach the exhaust pipe and reduce the overall intubation time. When approaching the exhaust pipe, the speed is reduced to 0.5 - 1cm / s for insertion. This slower speed allows the robotic arm to more accurately control the position of the exhaust gas collection tube, ensuring that the exhaust gas collection tube is accurately docked with the exhaust pipe opening, which helps to avoid insertion position deviation caused by excessive speed, thereby improving the accuracy of intubation.

[0034] At the same time, during the insertion process of the exhaust gas collection tube, the insertion pressure of the exhaust gas collection tube is obtained in real time; if the insertion pressure of the exhaust gas collection tube is greater than the pressure threshold, it means that the exhaust gas collection tube may be stuck. At this time, the exhaust gas collection tube is pulled out in the direction opposite to the insertion direction to the pull-out depth threshold, or the exhaust gas collection tube is pulled out to the pull-out depth reaching the corresponding pull-out depth threshold, and then the exhaust gas collection tube insertion action is executed according to the corresponding insertion direction and insertion speed, and the exhaust gas collection tube continues to be inserted into the exhaust pipe.

[0035] The present embodiment provides an intelligent intubation control method for exhaust gas detection of motor vehicles based on image recognition. First, the position and shape of the exhaust pipe at the rear of the car are determined by identifying the exhaust pipe through image; then, the position of the exhaust pipe in the working space of the robotic arm is determined by converting the two-dimensional coordinates of the image of the rear of the car into the three-dimensional coordinates of the robotic arm, and the moving path is planned in combination with the joint angle information of the robot robotic arm, providing clear motion instructions for the robotic arm; finally, according to the moving path of the robot robotic arm, the exhaust gas collecting pipe at the end of the mobile robotic arm is directly opposite to the exhaust pipe, and a segmented speed control method is adopted during the insertion of the exhaust gas collecting pipe, which can better control the insertion accuracy during the insertion process, and at the same time, the insertion pressure of the exhaust gas collecting pipe is obtained in real time, and the measures of pulling out and reinserting are taken when the pressure is greater than the pressure threshold, which can effectively prevent the exhaust gas collecting pipe or the exhaust pipe from being damaged due to excessive insertion force; the intubation control method has strong adaptability and high intelligence, can realize a high degree of automation of the exhaust gas detection intubation process, reduces the tediousness and uncertainty of manual operation, and can improve the accuracy and efficiency of exhaust gas detection.

Claims

1. An intelligent intubation control method for motor vehicle exhaust detection based on image recognition, characterized in that: The following operations are included: S1, obtaining a rear image of a car, and identifying an exhaust pipe from the rear image of the car; S2, after converting the two-dimensional coordinates of the rear image of the car into the three-dimensional coordinates of the robot arm, path planning is performed based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain the moving path of the robot arm; The method for obtaining the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm is specifically as follows: based on the intrinsic parameters and extrinsic parameters of the camera, a camera imaging model and a conversion relationship model between the camera coordinate system and the robot arm coordinate system are constructed; after substituting the two-dimensional coordinates of the exhaust pipe into the camera imaging model, the conversion relationship model is processed by the least square method to obtain the exhaust pipe world vertical coordinate; the exhaust pipe world vertical coordinate and the exhaust pipe two-dimensional coordinate are substituted into the camera imaging model for solution to obtain the exhaust pipe world horizontal coordinate and the exhaust pipe world vertical coordinate; the exhaust pipe world horizontal coordinate, the exhaust pipe world vertical coordinate and the exhaust pipe world vertical coordinate form the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm; S3, according to the moving path of the robot arm, the exhaust gas collection pipe at the end of the robot arm is moved to face the exhaust pipe; Inserting the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold in a preset insertion direction and using a segmented speed control method; During the exhaust gas collection pipe insertion process, the exhaust gas collection pipe insertion pressure is obtained in real time; if the exhaust gas collection pipe insertion pressure is greater than the pressure threshold, the exhaust gas collection pipe is pulled out in the direction opposite to the insertion direction to the pulling-out depth threshold, and then the exhaust gas collection pipe insertion action is performed.

2. The intelligent intubation control method for motor vehicle exhaust detection based on image recognition according to claim 1 is characterized in that: In S2, the operation of obtaining the moving path of the robot arm is specifically: The initial angles of each joint of the robot arm are used as the root nodes of the random tree to obtain the initial random tree; The initial random tree is iterated through node expansion until the distance between the generated new node and the target angle point is less than the distance threshold, and then the iterative process ends; The new node whose distance to the target angle point is less than the distance threshold is used as the target node; In the process of tracing back from the target node to the root node, the edges between the nodes in the random tree are traversed to obtain several motion paths; Based on the path cost function, several motion paths are screened, and the motion path with the smallest path cost value is used as the moving path of the robot manipulator; The target angle point is the point at the end of the robot arm corresponding to the standard exhaust pipe position.

3. The intelligent intubation control method for motor vehicle exhaust detection based on image recognition according to claim 2 is characterized in that: The path cost function is obtained by the following formula: ， Q For the path value, d(p i ,p i+1 ) For Node p i With Node p i+1 The distance between θ i+1 For Node p i+1 Angle, θ i For Node p i Angle, q e According to the energy consumption of the robot arm when moving along the motion path, I is the total number of nodes in the motion path.

4. The intelligent intubation control method for motor vehicle exhaust detection based on image recognition according to claim 2 is characterized in that: During the iterative processing of node expansion, the specific operations of the current node expansion processing are: In the random tree generated in the previous round, a leaf node is randomly generated within the range of the minimum moving angle and the maximum moving angle of each root node to obtain several current transition leaf nodes; The nodes in the random tree generated in the previous round that are closest to the corresponding current transition leaf nodes are used as the current closest leaf nodes corresponding to each root node; The current nearest leaf node is moved along the direction of the corresponding current transition leaf node by the step size threshold, and is used as the current new node of the corresponding root node; Connect the current new node with the corresponding current nearest leaf node to implement the update iteration of the random tree, and obtain the random tree generated in the current round, which is used to perform the operation of the next round of node expansion processing.

5. The intelligent intubation control method for motor vehicle exhaust detection based on image recognition according to claim 1 is characterized in that: In S1, the operation of identifying the exhaust pipe from the rear image of the vehicle is specifically as follows: Mark the edge position points in the car tail image to obtain a marked car tail image; cover the marked car tail image with detection frames, and use the detection frame corresponding to the minimum distance between the detection center point and the center point of the real frame in the same area as the classification frame of the corresponding area; classify all classification frames based on the shape, pixel distribution and area of ​​the exhaust pipe, and use the classification result of the exhaust pipe and the classification frame with a corresponding classification probability value greater than the probability threshold as the exhaust pipe recognition frame; The exhaust pipe identification frame is processed by Hough transform to obtain the center position of the exhaust pipe; the coordinates of the center position of the exhaust pipe in the rear image of the car are used as the two-dimensional coordinates of the exhaust pipe.

6. The intelligent intubation control method for motor vehicle exhaust detection based on image recognition according to claim 5 is characterized in that: The specific method for obtaining edge position points is: The image of the rear of the car is grayed to obtain a gray image of the car; the weighted average of the pixels of each position point and other position points within the first neighborhood in the gray image of the car is used as a new pixel value of the corresponding position point to obtain a denoised image of the car; Determine whether the pixel value of each position point in the car denoised image is greater than the edge pixel threshold; If the pixel value of the current position point is not less than the first edge pixel threshold, the current position point is an edge position point; If the pixel value of the current position point is between the first edge pixel threshold and the second edge pixel threshold, and the current position point is adjacent to the determined edge position point, then the current position point is an edge position point; The first edge pixel threshold is greater than the second edge pixel threshold.

7. An intelligent intubation control system for motor vehicle exhaust detection based on image recognition, used to implement the intelligent intubation control method for motor vehicle exhaust detection based on image recognition as claimed in claim 1, characterized in that: include: A robot arm, with an exhaust gas collection pipe installed at the end, for executing the action of moving the exhaust gas collection pipe according to the instructions transmitted by the control unit; A camera, mounted on the robot arm, is used to obtain images of the rear of the car; An image processing unit, used for processing the image of the rear of the car acquired by the camera, and identifying the exhaust pipe from the image of the rear of the car; A computing unit, configured to convert the two-dimensional coordinates of the rear image of the vehicle into three-dimensional coordinates of the robot arm, and perform path planning based on the position information of the exhaust pipe in the three-dimensional coordinates of the robot arm and the joint angle information of the robot arm to obtain a moving path of the robot arm; A control unit, used for moving the exhaust gas collecting pipe at the end of the robot arm to face the exhaust pipe according to the moving path of the robot arm in the calculation unit; Inserting the exhaust gas collection pipe into the exhaust pipe to a preset depth threshold in a preset insertion direction and using a segmented speed control method; and when the exhaust gas collecting pipe insertion pressure is greater than a pressure threshold during the exhaust gas collecting pipe insertion process, the exhaust gas collecting pipe is pulled out in a direction opposite to the insertion direction to a pulling out depth threshold, and then the exhaust gas collecting pipe insertion action is performed; Pressure sensor for obtaining the exhaust gas collection pipe insertion pressure.

8. The intelligent intubation control system for motor vehicle exhaust detection based on image recognition according to claim 7 is characterized in that: There are two cameras, one camera is installed on the main fixed pole of the robot arm to obtain the car image, and the other camera is installed at the end of the robot arm to obtain the car tail image when the car tail exists in the car image.

9. The intelligent intubation control system for motor vehicle exhaust detection based on image recognition according to claim 7 is characterized in that: Also includes: A depth sensor, used to obtain the depth of the exhaust gas collecting pipe inserted into the exhaust pipe; The exhaust gas collection pipe fixing clamp is installed at the end of the robot arm and is used to fix the exhaust gas collection pipe.

Images