A Study On The Method Of Obstacle Avoidance Through Navigation For Indoor Dynamic Complex Environment

Patrol robot,as an important branch of robot,has developed rapidly in recent years.In the development of patrol robot,the research of navigation and obstacle avoidance is an indispensable part.Combined with the task requirements of indoor patrol robot,this paper puts forward a set of navigation obstacle avoidance scheme,aiming at obstacle positioning method,pedestrian detection method and map construction method,aiming at solving the above problems with computer vision,and finally realizing automatic patrol of mobile robot.The main contents and work of this paper are as follows:(1)on the detailed analysis as well as the principle of binocular distance measurement,image correction if calibration method for visual range,based on the theory of information and visual feedback is proposed based on image depth robots avoid obstacles strategy,can be achieved under the environment of indoor dynamic obstacle avoidance function,and compared with obstacle recognition algorithm,this paper strategy to lower demand for hardware,has faster response speed.(2)in HOG detector and detailed analysis,on the basis of KCF tracking algorithm,based on patrol robot patrol indoors problems encountered in the process of pedestrians,the RGB-D detector to detect the pedestrian,and puts forward the factors in local path planning to join the pedestrian,make robot patrol in the indoor environment pay more attention to the safety of pedestrians,in does not affect the patrol robot system under the condition of stability and real-time performance,making more intelligent robots.(3)based on the detailed analysis of the principle of map construction using computer vision,this paper proposes the method of robot active map construction aiming at the problem that manual control robot is required to walk in the indoor environment to construct the map in the construction of point cloud map.Meanwhile,in view of the problem that dynamic targets in camera field are not considered in the construction of point cloud map,this paper proposes a new method,focusing on pedestrians in camera field,to improve the accuracy of inter-frame matching in the process of visual map construction.In order to solve the problem of pedestrians affecting map availability in point cloud map construction,this paper proposes a new method to remove pedestrians from point cloud map and make point-cloud map more reasonable.