Research On UAV Crossing Moving Obstacles Based On Visual Positioning

UAV platform is an important tool for aerial monitoring,surveying and identification,and it is of great significance to use UAV to carry out target detection and tracking for unknown environment environment.Especially in the limited space environment,it has important application value that using the UAV platform equipped with visual sensors to carry out obstacle detection,tracking and traversal for visual technology.However,in the indoor environment,the navigation signal of the GPS device may be weak or missing,so that it is unable to provide navigation and positioning information for the UAV.In addition,due to the random movement of obstacles,it is easy to cause flight failure and collision in indoor environment.To this end,this paper considers the scenario of the GPS-free UAV crossing a random moving doorframe and proposes a set of autonomous crossing methods in an indoor unknown environment,it includes five links: UAV autonomous positioning,obstacle detection,obstacle positioning,obstacle tracking and obstacle crossing.This article first discusses the overall strategy design of drones crossing the moving obstacle,and then compares and applies each link method separately,including vision-based autonomous positioning,random moving doorframe target detection,target positioning,and target tracking.Finally,the UAV platform is used to verify the method and autonomously traverse moving obstacles in the indoor environment.Regarding the doorframe as the target obstacle,we study the problem of traversing mobile doorframe by the UAV without GPS.The specific content is as follows:First,the overall strategy design of UAV crossing moving obstacles.We disscuss the design idea and composition of the overall strategy,analyze the key technologies of each component module and give solutions,and propose a set of solutions based on visual positioning for UAV to cross moving obstacles.Second,UAV platform detectes random moving target based on visual positioning.The UAV does not rely on GPS equipment to obtain navigation information and uses the visual SLAM method to accurately determine its position and posture in the current environment to achieve navigation and positioning.Then we use a monocular camera to obtain environmental information,extract the target imaging position through the image detection method,and determine the target imaging area.Finally,we use the target geometric constraints to establish the equation of target space position and solve the monocular scale uncertainty problem,and fuse the UAV pose data to get the target absolute position.Third,UAV platform tracks and traverses random moving doorframe.After determining the position of the target space,the mobile doorframe is tracked,we use Kalman filtering to predict its movement position,and control the UAV to follow the mobile doorframe.Then combining the target speed,the current position,the UAV’s own speed and position to estimate the target’s moving position in the future and controlling the UAV to complete the moving doorframe.We carry out experiments on autonomous flying,tracking and traversing obstacles in moving doorframes for indoor environment drones to evaluate the effectiveness of algorithms in all links.Simulation and actual scene tests show that this method can effectively achieve autonomous positioning and navigation of the UAV platform,and continuous tracking and traversal of random moving doorframes in a complex indoor background environment.This method is of great significance for the research of GPS-free drones using visual autonomous positioning to complete mission flight,from visual SLAM mapping to image recognition technology,from target detection and tracking to drones crossing obstacles,whether applied to civilian applications,or the military field,these have great application value.