Vision-Based Guidance Research On Intelligent Control Of Underwater Hydraulic Robotic Arm

As modern technology continues to advance,underwater robots are widely used in production.The combination of computerised image processing technology is an important way of enhancing their intelligence.With the increasing exploration of the oceans,underwater operations are often required.However,due to factors such as low light,turbid water and turbulent currents underwater,it is difficult to achieve fast and accurate grasping of underwater targets,resulting in inefficient underwater operations.This thesis introduces a vision-guided underwater hydraulic robotic arm,which is capable of accurate and efficient target grasping by means of vision recognition for ranging and acquiring the external outline of the target.The structural design of the robotic arm is completed,a suitable camera and vision system are selected,an improved image restoration algorithm is used to restore the image,based on the processed image and using binocular ranging,the accuracy of ranging is achieved,a greedy triangulation algorithm is used to smoothly construct the target object in 3D,and a suitable position is selected for grasping according to the outline of the target object,effectively improving the grasping success rate.The aim of this thesis is to extend the research on land-based binocular stereo vision to underwater,and to develop an underwater image ranging and 3D reconstruction algorithm,which provides a visual guide for the accurate identification and grasping of underwater hydraulic arms.The main elements are as follows:(1)The overall scheme of the master-slave hydraulic robotic arm interactive control system was completed.The structural design of the robotic arm was carried out,and the design analysis of its hydraulic system was carried out.The design of the human-machine interaction system of the master-slave hydraulic robotic arm was completed,and the visual simulation of the robotic arm working process was realised.(2)The coordinate system transformation method of the ROV binocular vision robotic arm positioning system was studied.The transformation from the world coordinate system to the camera coordinate system was achieved by determining the internal and external parameters of the camera through the Zhang Zhengyou calibration method,and the transformation from the camera coordinate system to the end-effector coordinate system was achieved by determining the position relationship between the camera and the end-effector through the hand-eye calibration.(3)An algorithm for fast de-fogging of fogged images based on the dark channel priority principle is used.The algorithm addresses the drawback that the HE algorithm takes a lot of time to optimise the transmission map t.It uses a combination of adaptive median filtering and bilateral filtering to calculate the dark channel,overcoming the slow speed of the HE algorithm.Fast and high-quality de-fogging of underwater turbid images is achieved without basically affecting the de-fogging effect.A parallax map is obtained on the basis of the defogged image using the SAD matching method,and the scene is reconstructed in three dimensions based on the parallax estimation results,which in turn leads to accurate ranging.Then an improved greedy projection triangle mesh reconstruction algorithm was proposed to perform greedy projection triangle reconstruction of the point cloud by the normal vector estimation of the MLS algorithm.After a smoother and more complete object surface was reconstructed,a suitable area was selected for grasping,and an improvement in the grasping success rate was achieved.(4)The robot model was imported into MATLAB/Simscape as the controlled object and two controller designs were carried out.The fuzzy PID algorithm was used to track and control the desired trajectory,and the robot’s operation process was clearly seen through the Simscape visualisation window.The simulation results show that the trajectory tracking performance under the fuzzy PID control algorithm is good,and the trajectory tracking error of each joint is kept within 1%.The success rate of grasping different areas of the target object was counted through several grasping experiments,and the results verified the feasibility of the idea.