Research On Vision Servo Strategy Of Mobile Robot With Unexpected Image

Visual servoing of mobile robots is a very hot research topic in the field of robotics at present.The desired image is the important information for the mobile robot to complete the visual servoing stabilization task.In general research methods,the desired images are indispensable.Without the desired image,the desired pose will not be defined,which can lead to the failure of the visual servoing stabilization task.Moreover,in the process of capturing feature points and tracking,the camera may lose some feature points due to the occlusion of some unknown obstacles.Thus,the correct visual information can not be obtained.It also can lead to the failure of visual servoing stabilization task.Therefore,this thesis will study the visual servoing stabilization system without desired image.The main work is as follows:(1)To verify the algorithms proposed in this thesis,the experimental platform of visual servoing stabilization system for the mobile robot is built.Firstly,the hardware platform is built,including the connection and communication mode between various parts of hardware.Secondly,the software platform is built,including the establishment of the robot operating environment and a camera display environment,and realizing the precise control of the mobile robot and the camera.Finally,to improve the efficiency of experimental development,the functions commonly used in experiments are secondary encapsulation.(2)To solve the problem of the unknown desired image in the visual servoing stabilization task for the mobile robot,in this dissertation,a method about the visual servoing stabilization without desired image is proposed based on arbitrary desired pose.Based on the 3D coordinates of the measured feature points and the corresponding definition of coordinate system,and by using the imaging model of the feature points,the transformation of the coordinate system and the least square solution of the homography matrix,the pose relationship between the current pose of the mobile robot and the desired pose set arbitrarily is finally obtained.Finally,the visual servoing stabilization control is realized by polar coordinate controller.In addition,this section analyzes two difficulties encountered in the simulation to make the simulation closer to the real environment when the mobile robot is running.Simulation results and experimental are provided to verify the feasibility and effectiveness of the proposed strategy.(3)When the mobile robot carries out the visual servoing stabilization task without desired image,to solve the problem of visual information loss arising form some feature points occlusion,a pose estimation method of partial feature points occlusion is proposed in this dissertation.To be specific,it can use the data of the current frame to determine which feature point is occluded,and calculate several groups of possible pose relationships based on the three feature points that are not lost.Based on the data of the previous frame,it can judge the correct pose relationship and recover the data of the current frame.Similarly,the data lost in the next frame can also be recovered,and the rest can be done in the same manner.Therefore,we can recover the correct data for each frame and complete the visual servoing stabilization task.In this section,two sets of experiments are performed,and two different feature points are occluded respectively.The results verify the feasibility of the method.