Semi-autonomous cutting and removal method for water stop needles used in underwater crack repair of tunnels

Abstract

The present invention discloses a semi-autonomous cutting and removal method for water stop needles in underwater crack repair of tunnel. Based on the underwater small target detection method of improved YOLOv5, the water stop needles are identified in the preprocessed crack environment image. According to the recognition result and the binocular camera disparity map, the three-dimensional position coordinates of the water stop needles in the base coordinate system of the robotic arm are calculated. The water stop needle with the closest distance is selected as the operation task target. Through the inverse kinematics solution of the robotic arm, the target pose of the end of the underwater robot robotic arm is obtained. The improved RRT algorithm is used to plan the path of the underwater robot robotic arm. Finally, the operator uses the incremental position control method to operate the robotic arm to execute the cutting action through the master-slave force feedback human-machine interface. The method of the present invention improves the automation level and safety degree of the operation process.

Description

Technical Field

[0001] The invention belongs to the technical field of underwater robot detection and operation, and particularly relates to a semi-autonomous cutting and removal method for water-stop needles used for underwater crack repair of tunnel Background Technique

[0002] Hydraulic tunnels can be used for irrigation, power generation, water supply, water discharge, water conveyance, construction diversion and navigation. They play an important role in China's water conservancy projects and are important infrastructure related to national development and people's livelihood security. However, long-term water flow scouring and natural disasters will cause serious damage to them. If the surface defects of the hydraulic tunnel are not detected and repaired in time, water leakage or even fracture and collapse will occur, causing huge economic losses and endangering people's lives and property safety. Grouting the cracks is a common method for repairing the cracks of the hydraulic tunnel. After grouting, the water-stop needles need to be removed

[0003] The removal operation of the water-stop needles is usually completed by manual knocking. However, underwater, human operation is difficult and has certain risks, and the operation efficiency is limited. Therefore, it is very necessary to introduce a robot to replace manual operation. For a robot, it is difficult to complete the knocking action, and there are high requirements for the instantaneous speed of the robot's end. Moreover, the huge impact force generated instantaneously is not only difficult to control, but also will damage the mechanical structure of the robot, affecting the working accuracy and service life of the robot. At the same time, the underwater environment has poor lighting conditions, and the images obtained by the image acquisition device are dim and unclear. As a small target without distinct color features, the water-stop needle has problems with the accuracy of recognition in both the manual teleoperation mode and the full-autonomous mode. Therefore, for the underwater crack repair operation of the hydraulic tunnel, there is an urgent need for a semi-autonomous underwater removal method for the water-stop needle applicable to the robot to improve the automation level and emergency response ability of the hydraulic tunnel repair, so as to promote the healthy and smooth development of the water conservancy project Summary of the Invention

[0004] In view of the problems in the prior art that the removal of the water-stop needle is time-consuming, laborious, has low work efficiency and low accuracy, the present invention provides a semi-autonomous cutting and removal method for the water-stop needle for repairing underwater cracks in a tunnel. Based on the underwater small target detection method improved YOLOv5, the water-stop needle is identified from the preprocessed crack environment image. According to the recognition result, the pixel coordinates of each water-stop needle on the two-dimensional image are obtained. The three-dimensional position coordinates of the water-stop needle in the camera coordinate system are determined by the binocular parallax of the main camera. According to the hand-eye calibration result with the eye outside the hand, the three-dimensional position coordinates of the water-stop needle in the base coordinate system of the robotic arm are calculated. The water-stop needle closest to the end of the current underwater robot robotic arm is selected as the target of this operation task. Then, through the inverse kinematics solution of the robotic arm, the target pose of the end of the underwater robot robotic arm is obtained. The improved RRT algorithm is used to plan the path of the underwater robot robotic arm to output the joint space set and obtain a suitable path position. Finally, the operator uses the incremental position control method to operate the underwater robotic arm to perform the cutting action through the master-end force feedback human-machine interface. The present invention cuts and removes the water-stop needle through a bilateral teleoperation mode, improving the automation level and safety of the operation process.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: a semi-autonomous cutting and removal method for the water-stop needle for repairing underwater cracks in a tunnel, including the following steps:

[0006] S1: Obtain an image of the underwater crack environment in the tunnel through the main camera on the underwater robot, and preprocess the image;

[0007] S2: Based on the underwater small target detection method improved YOLOv5, identify the water-stop needle from the image preprocessed in step S1, and transmit the recognition result back to the master end; the underwater small target detection method improved YOLOv5 uses Swin Transformer as the basic backbone network of YOLOv5, adds an additional upsampling layer to fuse with the corresponding layer of the backbone, and sets the connection method between the feature pyramid and PANet as a fine-grained YOLO feature output layer;

[0008] The specific method for adding an additional upsampling layer to fuse with the corresponding layer of the backbone is to upsample the output of the backbone again, fuse it with the output of the previous backbone layer to generate a new feature map, and downsample the new feature map as the output of the newly added detection layer;

[0009] S3: After the master-end display receives the recognition result, display the rectangular frame of the water-stop needle recognition result on the display to assist the operator in cognition;

[0010] S4: Based on the detection results of the water stop needle targets in the images collected by the left side of the binocular main camera, obtain the pixel coordinates of the center points of the recognition result rectangles of each water stop needle in the left image. Then, match the corresponding pixel coordinates of this point in the right image, and according to the binocular vision ranging principle, calculate the depth information by calculating the disparity between the left and right images, so as to obtain the three-dimensional position coordinates of the water stop needle in the camera coordinate system. Furthermore, based on the hand-eye calibration result with the eye outside the hand, obtain the coordinate transformation relationship between the camera coordinate system and the base coordinate system of the underwater manipulator, and calculate the coordinate value of the water stop needle in the camera coordinate system converted to the coordinate value in the base coordinate system. According to the three-dimensional position coordinates of the water stop needle in the manipulator base coordinate system, select the water stop needle closest to the end of the current underwater manipulator as the target of this operation task;

[0011] S5: Through the inverse kinematics solution of the manipulator, obtain the target pose of the end of the underwater robot manipulator, and use the improved RRT algorithm to perform path planning on the underwater robot manipulator to output the joint space set, so as to approach the operation task target faster and obtain a suitable path position;

[0012] S6: When the end of the underwater robot manipulator autonomously reaches the water stop needle at the operation task target, the operator uses the incremental position control method to operate the underwater manipulator to perform the cutting action through the master-slave force feedback human-machine interface. The specific incremental position control method is as follows: the force feedback human-machine interface reads the incremental value ΔP at a certain frequency m =P m2 -P m1 , and multiplies it by the proportionality coefficient K to calculate the position increment ΔP of the end of the slave manipulator s =KΔP m , and then obtain the target position of the end of the slave manipulator as P s2 =P s1 +KΔP m . The angles of each joint of the manipulator can be obtained by solving the inverse kinematics of the manipulator.

[0013] Compared with the prior art: The water stop needle semi-autonomous cutting and demolition method for underwater crack repair in a tunnel of the present invention avoids manual underwater operations, improves the automation level of the operation process, and has the advantages of high safety for operators and high construction efficiency. Among them, the final cutting and demolition method adopts a bilateral teleoperation mode, introduces dual feedback of vision and force sense, enhances the operator's sense of presence and control ability, reduces the operator's cognitive load, and at the same time avoids damaging the concrete facilities of the hydraulic tunnel by introducing the method of virtual fixture in the prohibited area. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the water stop needle, which is the object to be cut and demolished in the present invention;

[0015] Figure 2 Schematic diagram of the working environment for semi-autonomous cutting and removal of the water stop needle for underwater crack repair in a tunnel according to the present invention;

[0016] Figure 3 Schematic diagram of the structure of the robot responsible for cutting and removing the water stop needle in the present invention;

[0017] Figure 4 Schematic diagram of the structure of the cutting tool in the robot of the present invention;

[0018] Figure 5 Flowchart of the steps of the semi-autonomous cutting and removal method of the water stop needle for underwater crack repair in a tunnel according to the present invention;

[0019] Figure 6 Flowchart of the steps of image preprocessing in step S1 of the method according to the present invention;

[0020] Figure 7 Schematic diagram of the connection mode between the feature pyramid and PANet in the underwater small target detection method based on improved YOLOv5 in step S2 of the method according to the present invention;

[0021] Figure 8 Flowchart of the RRT path planning process of the underwater robotic arm in step S5 of the method according to the present invention;

[0022] List of attached drawing reference numerals:

[0023] 1. Control computer and display; 2. Operator; 3. Force feedback man-machine interface; 4. Optoelectronic composite cable; 5. Robot; 5-1. Walking wheel; 5-2. Crawler; 5-3. Chassis bracket; 5-4. Robot box body;

[0024] 5-5. Electric lifting frame; 5-6. Main lighting lamp; 5-7. Main camera;

[0025] 5-8. Robotic arm; 5-9. Wrist lighting lamp; 5-10. Wrist camera;

[0026] 5-11. Cutting tool; 11-1. Coupling cavity; 11-2. Six-axis force sensor;

[0027] 11-3. Cutting grinding wheel; 11-4. Adapter shaft;

[0028] 11-5. Bearing gland; 11-6. Jaw fixing device;

[0029] 11-7. Device housing; 11-8. Motor

[0030] 5-12. Mechanical jaw;

[0031] 6. Water stop needle; 7. Tunnel; 7-1. Crack; Detailed implementation manners

[0032] The present invention will be further clarified below in conjunction with the accompanying drawings and specific implementation manners. It should be understood that the following specific implementation manners are only used to illustrate the present invention and not to limit the scope of the present invention.

[0033] Embodiment 1

[0034] A semi-autonomous cutting and removal method for water-stop needles used for repairing underwater cracks in tunnels. The water-stop needles for the cutting and removal operation object are as Figure 1 shown. After the grouting of the underwater crack 7-1 in the hydraulic tunnel is completed, the water-stop needles 6 installed along the crack need to be cut and removed. The working environment for the cutting and removal is as Figure 2 shown, which is composed of a main end, a slave end, and a communication link. The main end includes a control computer and a display 1, an operator 2, and a force feedback human-machine interface 3; the communication link realizes long-distance wired communication through an optical-fiber composite cable 4; the slave end is an underwater robot 5 for cutting and removing the water-stop needles 6, the water-stop needles 6 for the cutting and removal operation object, and the working environment, the hydraulic tunnel 7.

[0035] Among them, the underwater robot 5 at the slave end mainly includes a walking wheel 5-1, a crawler 5-2, a chassis bracket 5-3, a robot box 5-4, an electric lifting frame 5-5, a main lighting light 5-6, a main camera 5-7, a robotic arm 5-8, a wrist lighting light 5-9, a wrist camera 5-10, a cutting tool 5-11, and a mechanical gripper 5-12.

[0036] Walking wheels 5-1 and crawlers 5-2 are installed on both sides of the robot chassis 5-3 to provide the robot system with the ability to move to the working scene; the robot box 5-4 is installed above the robot chassis 5-3, and a battery and a control system circuit board are installed inside the robot box to provide power supply, navigation positioning, and action execution control instructions for the robot system; two electric lifting frames 5-5 are installed above the robot box 5-4, and the main lighting light 5-5 and the main camera 5-7 are respectively installed above the two electric lifting frames. The main camera 5-7 is a binocular camera, which is used to obtain the surrounding environment information and measure distances, so as to provide visual guidance for the action execution of the robot system; the robotic arm 5-8 has four degrees of freedom and is installed on the robot box 5-4. A mechanical gripper 5-12 is assembled at the end of the robotic arm. The gripper can stably hold the cutting tool 5-11 through the cooperation of protrusions and grooves, and as the action execution mechanism of the robot system, it completes the cutting action of the water-stop needle; a wrist lighting light 5-9 and a wrist camera 5-10 are assembled at the end of the robotic arm. The wrist camera provides multi-angle assistance for the operator during the execution of the final cutting action. The cutting tool 5-11 can be selected as a angle grinder, and protrusions are arranged on the surface of the grasping part of the angle grinder for cooperation with the groove of the robotic arm gripper to achieve stable grasping.

[0037] As shown Figure 4 in the figure, the cutting tool 5-11 in the robot mainly includes a coupling cavity 11-1, a six-axis force sensor 11-2, a cutting grinding wheel 11-3, a transfer shaft 11-4, a bearing gland 11-5, a jaw fixing device 11-6, a device housing 11-7, and a motor 11-8. The motor 11-8 drives the cutting grinding wheel device 11-3 to rotate through the transfer shaft 11-4 to complete the cutting action. The coupling cavity 11-1 is installed with a mechanical seal to achieve a waterproof design. The surface of the device housing 11-7 is designed with a jaw fixing device 11-6. By designing protrusions on the hexagonal housing, the cooperation with the grooves on the mechanical jaw 5-12 is realized, achieving stable clamping and avoiding slipping during underwater movement. The six-axis force sensor 11-2 is installed between the end flange of the coupling cavity 11-1 and the cutting grinding wheel 11-3. The real-time cutting force during the process of cutting and removing the water stop needle can be grasped through the force sensor. After being processed by the control computer, the force data is sent to the force feedback man-machine interface 3 to provide feedback force for the master operator 2 when performing the cutting and removing task; the force sensor data communicates with the robot box body 5-4 through wireless transmission, and then is transmitted to the master control computer and the display 1 through the wired communication of the optical and electrical composite cable 4.

[0038] The force feedback man-machine interface 3 in this embodiment is a force feedback hand controller, which is used for the master operator 2 to control the slave manipulator 5-8 to reach the final cutting point in the teleoperation mode, and use the buttons on the hand controller to control the start and stop of the cutting tool to complete the cutting action; the control computer and the display 1 receive the image data transmitted back from the slave end, perform image processing and generate a visual interface, and the operator 2 obtains the working environment information of the slave end through the display screen.

[0039] The semi-autonomous cutting and removal method for the water stop needle used for underwater crack repair in the tunnel is as Figure 5 shown in the figure and includes the following steps:

[0040] Step S1: The main control computer performs image enhancement on the working environment image collected by the left side of the main camera binocular using the contrast-limited adaptive histogram equalization algorithm and performs image sharpening using the USM method.

[0041] Among them, the process of using the contrast-limited adaptive histogram equalization algorithm and the USM method for image preprocessing is as Figure 6 shown in the figure: First, convert each frame of the returned RGB format image to the HSV format; secondly, perform contrast-limited adaptive histogram equalization processing on the H, S, and V channels respectively and then synthesize to obtain image A; use USM sharpening to obtain image B; then fuse the above two result images through simple weighted calculation, that is, I enhanced= α * A + (1 - α) * B, where 0 < α < 1.

[0042] Step S2: The master control computer uses an improved YOLOv5-based underwater small target detection method to identify the water stop needles in the preprocessed images. Based on the classical YOLOv5 target detection method, Swin Transformer is used as the basic backbone network of YOLOv5 to be applicable to underwater images with blurred targets; and by adding an additional upsampling layer to fuse with the corresponding layer of the backbone, the connection method between the feature pyramid and PANet is set as a fine-grained YOLO feature output layer to be applicable to small target detection. The connection method between the feature pyramid and PANet is as Figure 7 shown, where C1, C2, C3, C4, and C5 are the outputs of the backbone layers after convolution. By resampling the output of the backbone again and fusing it with the output of the previous backbone layer to generate a new feature map, the new feature map is downsampled to be used as the output of the newly added detection layer. The underwater small target detection method uses the Mosaic data augmentation method and is first trained in the land water stop needle dataset, and then transfer learning is used to train in the underwater water stop needle dataset.

[0043] Step S3: According to the above target detection results, the recognition result rectangle of the water stop needle is displayed on the monitor to assist the operator in recognition.

[0044] Step S4: According to the water stop needle target detection results of the images collected on the left side of the binocular vision of the main camera, the pixel coordinates (X left , Y left ) of the center point of each water stop needle recognition result rectangle in the left image are obtained. Then, the corresponding pixel coordinates (X right , Y right ) of this point in the right image are matched, and according to the binocular vision ranging principle, the disparity d = X left - X right is calculated, and the depth information is calculated to obtain the three-dimensional position coordinates of the water stop needle in the camera coordinate system:

[0045]

[0046] where B is the baseline distance of the camera and f is the focal length of the camera.

[0047] When performing hand-eye calibration, the AprilTags tags are fixed at the end of the robotic arm, so that the camera can obtain the three-dimensional position of the end of the robotic arm in the camera coordinate system with high precision. The relationship between the camera coordinate system, the robotic arm base coordinate system, and the robotic arm end coordinate system can be described as follows:

[0048]

[0049] wherein is the transformation matrix between the base coordinate system and the camera coordinate system; is the transformation matrix between the base coordinate system and the end of the robotic arm, which can be obtained through the forward kinematics of the robotic arm; is the transformation matrix between the end of the robotic arm and the camera coordinate system, which can be obtained by calculating the AprilTags tag information through the camera.

[0050] By moving the position of the end of the robotic arm multiple times, multiple groups of results can be obtained, and the transformation matrix between the base coordinate system and the camera coordinate system can be obtained using the least squares method

[0051] According to the hand-eye calibration result with the eye outside the hand, the coordinate transformation relationship between the camera coordinate system and the base coordinate system of the underwater robotic arm is obtained, and the three-dimensional coordinate value of the stop water needle in the camera coordinate system is converted to the coordinate value in the base coordinate system. According to the three-dimensional position coordinates of the stop water needle in the base coordinate system of the robotic arm, the stop water needle closest to the current end of the robotic arm is selected as the target of this operation task.

[0052] Step S5: Through the inverse kinematics solution of the robotic arm, the target pose of the end of the robotic arm is obtained. Taking the end pose as the input, an improved RRT algorithm combined with artificial potential field guidance is used for path planning to output the joint space set. Introducing the artificial potential field can increase the step size of the random tree expanding in the free space to approach the target faster, and reduce the step size in the obstacle area to obtain a suitable path position.

[0053] The path planning process is as Figure 8 shown. First, set the current position X current of the end of the robotic arm as the root node, the final target position is X goal , set the step size as ε, and then calculate the random probability p rand that follows a uniform distribution. If 0 ≤ p rand ≤ 0.2, then select the random sampling point X nearest with the closest Euclidean distance to the current node on the random tree as the direction of the node to be grown, and grow from X current to X nearest in the direction with step size ε to obtain a new node X new ; if 0.2 < p rand ≤ 0.6, then select the final target position X goal as the node growth direction, and grow from X current to X goal in the direction with step size ε to obtain a new node X new ; if 0.6 < p rand ≤ 1, then both the direction and step size of the new node are based on the random sampling point X nearest with the closest Euclidean distance to the current nodePerform calculations with the artificial potential field. The generation formula for the new node is where F total is the sum of the gravitational and repulsive forces of the artificial potential field acting on the end of the robotic arm at position X nearest . Then, determine whether X new collides with an obstacle or is less than the threshold distance from an obstacle. If a collision occurs or the distance is less than the threshold, discard it; otherwise, retain it. Continuously repeat the above steps until the random tree grows to the target area. Finally, perform redundant node removal and maximum curvature constraint operations to make the path smoother.

[0054] Step S6: When the end of the robotic arm autonomously reaches the expected cutting point of the water-stop needle, the system enters the bilateral teleoperation mode. The operator uses the incremental position control method to operate the slave underwater robotic arm through the master-end force feedback human-machine interface, thereby manually selecting the cutting point, enabling the slave underwater robotic arm to reach the final cutting point and perform the cutting action. During this process, the wrist camera provides multi-view assistance.

[0055] During the operator's operation, the force feedback human-machine interface reads its incremental value ΔP m = P m2 - P m1 at a certain frequency, multiplies it by the proportionality coefficient K, calculates the position increment ΔP s = KΔP m of the end of the slave manipulator, and further obtains the target position of the end of the slave robotic arm as P s2 = P s1 + KΔP m . By solving the inverse kinematics of the robotic arm, the angles of each joint of the robotic arm can be obtained, and then sent to the slave robotic arm through the communication link, enabling the robotic arm to receive the command and complete the corresponding action. The proportionality coefficient K affects the accuracy of the robot's movement. To meet the flexibility requirements of the robotic arm in different situations, the proportionality coefficient K uses variable proportional control. When |ΔP m | < 0.001 m, the proportionality coefficient K = 0 to eliminate the error caused by the operator's jitter; when 0.001 m ≤ |ΔP m | < 0.005 m, the proportionality coefficient K = 0.5 to achieve high-precision operation in a small range of the slave robotic arm; when 0.005 m ≤ |ΔP m | < 0.01 m, the proportionality coefficient K = 1; when |ΔP m | ≥ 0.01 m, the proportionality coefficient K = 2 to achieve rapid approach when the distance between the current position of the end of the robotic arm and the target is large.

[0056] During the process of the robotic arm performing cutting and demolition operations, cutting force data is collected from the six-axis force sensor in the end cutting tool and sent to the control computer to calculate the feedback force. Then, the feedback force is applied to the operator through the force feedback human-machine interface. At the same time, the hydraulic tunnel wall is set as an unreachable area at the end of the robotic arm, and the virtual fixture method is used to calculate the feedback force. When the end of the end manipulator approaches the prohibited area, the force feedback human-machine interface provides the feedback force to the operator, and as it gets closer to the prohibited area, the feedback force increases exponentially to prevent the operator from accidentally damaging the wall during operation.

[0057] In summary, the semi-autonomous cutting and demolition method of the water stop needle for underwater crack repair in tunnels of the present invention avoids manual underwater operations, improves the automation level of the operation process, has the advantages of high operator safety and high construction efficiency, is suitable for the actual situation of underwater crack repair in tunnels, has a high utilization rate, and greatly improves the working effect and working efficiency.

[0058] It should be noted that the above content only illustrates the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. For those of ordinary skill in the art of this technology, without departing from the principle of the present invention, several improvements and refinements can still be made, and these improvements and refinements all fall within the protection scope of the claims of the present invention.

Claims

1. Semi-autonomous cutting and removal method for water-stop needles used in underwater crack repair of tunnels, Characterized in that, It includes the following steps: S1: Obtain images of the underwater crack environment of the tunnel through the main camera on the underwater robot, and preprocess the images; S2: Identify the water-stop needles in the preprocessed images of step S1 based on the underwater small target detection method of improved YOLOv5, and transmit the recognition results back to the main end; the underwater small target detection method based on the improved YOLOv5 uses SwinTransformer as the basic backbone network of YOLOv5, adds an additional upsampling layer to fuse with the corresponding layer of the backbone, and sets the connection method between the feature pyramid and PANet as a fine-grained YOLO feature output layer; The method of adding an additional upsampling layer to fuse with the corresponding layer of the backbone is specifically to upsample the output of the backbone again, fuse it with the output of the previous backbone layer to generate a new feature map, and downsample the new feature map to be used as the output of the newly added detection layer; S3: After the main end display receives the recognition results, display the rectangular frames of the water-stop needle recognition results on the display to assist the operator in cognition; S4: According to the recognition results, obtain the pixel coordinates of the center point of each water-stop needle recognition result rectangular frame on the left image of the binocular camera, obtain the depth information through the binocular disparity of the left and right cameras of the main camera, so as to determine the three-dimensional position coordinates of the water-stop needle in the camera coordinate system; according to the hand-eye calibration result with the eye outside the hand, obtain the transformation matrix between the camera coordinate system and the robotic arm base coordinate system, calculate the three-dimensional position coordinates of the water-stop needle in the robotic arm base coordinate system, and select the water-stop needle closest to the end of the current underwater robot robotic arm as the target of this operation task; S5: Solve through the inverse kinematics of the robotic arm to obtain the target pose of the end of the underwater robot robotic arm, and use the improved RRT algorithm to perform path planning on the underwater robot robotic arm to output the joint space set to approach the operation task target and obtain a suitable path position; in the process of path planning using the improved RRT algorithm for new node expansion, three expansion strategies are selected according to random probability: expanding in the free space, expanding along the target direction, or expanding according to the guidance of the artificial potential field; Set the current position X of the end of the robotic arm current as the root node, and the final operation task target position is X goal . Set the step size as ε, and calculate the random probability p that follows a uniform distribution rand . If 0 ≤ p rand ≤ 0.2, then select the random sampling point X that is the closest to the current node in terms of Euclidean distance on the random tree nearest as the direction of the node to be grown. From X current to X nearest direction, grow a new node X with a step size of ε new ; If 0.2 < p rand ≤ 0.6, then select the final target position X goal The node growth direction is from X current to X goal and grow a new node X in steps of ε new ; If 0.6 < p rand ≤ 1, then both the direction and step size of the new node are calculated based on the randomly sampled point X nearest that is closest to the current node in terms of Euclidean distance and the artificial potential field. The generation formula for the new node is where F total is the sum of the gravitational and repulsive forces of the artificial potential field acting on the end of the robotic arm at X nearest . Then determine X again new Whether it collides with an obstacle or the distance from an obstacle is less than the threshold. If a collision occurs or the distance is less than the threshold, discard it; otherwise, keep it. Continuously repeat the above steps until the random tree grows to the target area of the operation task; S6: When the end of the underwater robot manipulator autonomously reaches the water stop needle of the operation task target, the operator uses the incremental position control method to operate the underwater manipulator to perform the cutting action through the master-side force feedback human-machine interface. The six-axis force sensor on the manipulator collects the cutting force and sends it back to the master-side control computer. The specific incremental position control method is as follows: The force feedback human-machine interface reads the incremental value ΔP at a certain frequency m = P m2 - P m1 , and multiplies it by the proportionality coefficient K to calculate the position increment ΔP at the end of the slave manipulator s = KΔP m , and then the target position at the end of the slave manipulator is obtained as P s2 = P s1 + KΔP m . The angles of each joint of the manipulator can be obtained by solving the inverse kinematics of the manipulator.

2. The semi-autonomous cutting and removal method for water-stop needles used in underwater crack repair of tunnels according to claim 1, Characterized in that: The preprocessing in step S1 at least includes image enhancement using the limited contrast adaptive histogram equalization algorithm and image sharpening using the USM algorithm. Step S1 specifically includes: S11: Convert each frame of the acquired RGB format image to HSV format; S12: Perform limited contrast adaptive histogram equalization processing on the H, S, and V channels respectively, and then synthesize the processed three-channel images to obtain image A; S13: Use USM sharpening on the HSV format image obtained in step S11 to obtain image B; S14: Perform weighted linear fusion on the image A obtained in step S12 and the image B obtained in step S13, that is, I enhanced = α * A + (1 - α) * B, where 0 < α < 1; S15: Convert the processed HSV image to RGB image.

3. The semi-autonomous cutting and removal method of the water stop needle for underwater crack repair in a tunnel as described in claim 2, characterized in that: during the process of path planning by combining the improved RRT algorithm guided by the artificial potential field in step S5, after the random tree grows to the target area of the operation task, it further includes an operation step of removing redundant nodes and an operation step of maximum curvature constraint to make the path smoother.

4. The semi-autonomous cutting and removal method of the water stop needle for underwater crack repair in a tunnel as described in claim 3, characterized in that: In the step S6, when |ΔP m | < 0.001 m, the proportionality coefficient K = 0; when 0.001 m ≤ |ΔP m | < 0.005 m, the proportionality coefficient K = 0.5; when 0.005 m ≤ |ΔP m | < 0.01 m, the proportionality coefficient K = 1; when |ΔP m | ≥ 0.01 m, the proportionality coefficient K = 2.

5. The semi-autonomous cutting and removal method of the water stop needle for underwater crack repair in a tunnel as described in claim 4, characterized in that: it further includes setting the tunnel wall as a prohibited area and restricting the movement of the end of the manipulator of the underwater robot by using the virtual fixture method of the prohibited area.

Images