With the development of artificial intelligence,a large number of robots have been put into use to replace human beings to complete dangerous tasks,especially when robots work together to solve complex tasks well.In this paper,unmanned aerial vehicle(UAV)and ground mobile robot are taken as the research object to study the target detection and navigation obstacle avoidance technology in cooperative robot environment.The following is the main research content of this paper:Firstly,the land-air cooperative robot rescue system is designed.Based on the complementary characteristics of UAV and mobile robot,the work flow of cooperative robot search and rescue is determined,so as to determine the tasks of UAV and ground mobile robot.Then the kinematics model of land-air robot is established respectively.Secondly,the identification method of aerial robot to ground target is studied.First,the characteristics difference between visible images of high-altitude and ground targets is analyzed,and the target recognition performance of SSD and YOLO target detection frameworks in different scale features in land and air scenes is tested,and the detection performance and practicality of the two are analyzed in detail.Secondly,the detection algorithm suitable for UAV scene and equipment platform is selected through performance comparison,and the ground target recognition model based on spatial pooling Darknet53 is constructed.The target detection model is trained by uav data set,and the test results of the detection model at high altitude are given.In order to estimate the target position,the binocular vision positioning method is used to estimate the target position,and the detection accuracy is tested.Thirdly,the obstacle avoidance algorithm of ground robot is studied.Differences of environmental barriers are classified according to the environmental characteristics,selection of grid maps describe the mobile robot working environment,in view of the kinds of obstacle detection,using the depth sensor detect obstacles,environmental perception method based on visual point cloud,the spatial point cloud information dimension reduction,the screening potential impassable regions and projected onto a two-dimensional plane form the obstacle map,The detection method is tested in a real environment with positive and negative obstacles.Finally,set the unmanned aerial vehicle(uav)and mobile robot experiment platform,using aerial drones the acquisition of video for ground target recognition algorithm based on spatial pooling Darknet53 is verified,and set up kinds of obstacles in the laboratory environment,use of obstacle avoidance method for mobile robot based on visual point cloud experimental verification,and to verify this air-ground robot coordinated rescue plan together. |