Research On Autonomous Grasp System Of EOD Robot Based On Deep Vision

In recent years,with the increasing frequency of terrorist bombings,public safety has become more and more important.More and more attention has been paid to the role of robots in the field of public safety.The EOD robot system based on the depth vision can replace the human to work in the dangerous environment and ensure the safety of the workers,so it has been widely developed and applied.However,the traditional grab system of eod robot based on fixed vision usually fixes the camera in a certain position,when the target object or manipulator is moving,the vision of the fixed camera is easily blocked by the manipulator arm,and for cameras with a certain focal length and pixels,the sharpness cannot be guaranteed when shooting from a distance,which has certain limitations;In addition,the traditional follow-up vision system often requires hand-eye calibration,resulting in more complicated and complicated solving process.Aiming at the above problems,a self-grasping system of explosion-ejector robot with a depth camera at the end of the manipulator and moving with the manipulator is proposed in this dissertation,and the system improves the grasp rate by improving the grasp method.This dissertation first studies the acquisition of the position information of the target object,places the depth camera at the fourth joint of the eod robot,analyzes the imaging principle of the camera,allocates the depth camera with the color camera,and calculates the 3d coordinates of the target object through the pixel ratio.Secondly,the motion control method of eod robot is studied in this dissertation.A coordinate conversion method based on a following camera is proposed to transform the position coordinate information of the target object in camera coordinates into the robot global coordinate system in real time,thus realizing the dynamic mapping relationship among the camera coordinate system H,the robot global coordinate system R and the tool coordinate system E of the end-effector claw,and then compare with the traditional fixed vision method.Finally,this dissertation studies fetching method of autonomous grasp system,and an improved grasping method based on GVF snake algorithm is proposed.The color map extracted by RGB camera was preprocessed,the target object contour was extracted by GVF snake algorithm and the pixel coordinates of the center point of the contour were obtained,which were converted into 3d coordinates,and the success rate of grasping the center point of the contour was improved.The experimental results show that the autonomous grasping system of the eod robot realized by the algorithm in this dissertation can reach the grasping point of the target object accurately within the tolerance range of 2cm.Moreover,the grasping success rate is high when the distance is between 100 cm and 150 cm,and the maximum grasping success rate is 85% at 130 cm.For the improved system in this dissertation,two kinds of objects of different shapes are used for grasping experiment to verify the performance of the system,and the maximum grasping success rate can reach 92%.In the same condition,compared with the traditional fixed vision,combined with the coordinate transformation method and the improved grasping method,the EOD robot grasping system proposed in this dissertation can not only meet the requirements of the EOD robot grasping explosives,but also improve the success rate of the claws of the system reaching the grasping point and grasping success rate.This system provides a basis for the work of explosive ejection robot and it has certain theoretical and practical significance.