Vision-based Object Following And Landing For Uavs In Unknown Environments

Target following and landing is one of the important contents of the autonomous flight of UAVs.It can not only be used to monitor the target in unknown environments but also guide landing of UAVs on the place which may dynamically change in unknown environments.Vision-based target following and landing of UAVs is one of most popular research topics because of its low-cost,high-precision and informative characteristics.In this paper,we use the monocular camera as the main sensor and AR.Drone 2.0 as the platform to study the target following and landing of UAVs in unknown environments.This research includes two aspects: target following and landing.In the target following stage,we use projective geometry to establish the relative coordinate between the camera and target,and obtain the 3D position of the target in this coordinate system.Then we use the PTAM algrithom to obtain the 6DOF pose of the UAV and establish the world coordinate system,so we can use the relationship between the camera coordinate system and the world coordinate system to obtain the 3D position of the target in the world coordinate system.We use the kalman filter to estimate the velocity of the target and design an object searching method to deal with the problem that the target moves out of the field of view of the camera.In the landing stage,we combine the advatages of projective geometry and visual servo to control the landing of UAVs.We solved some of the problems in the process of target following and landing.The first one is the problem of UAV position missing when the target moves out of the field of view of the camera when using relative coordinate system.We solved it by combining projective geometry and PTAM algrithom.The second problem is target missing.We use the kalman filter to estimate the velocity of the target so when the target is missing,the UAV can search it efficiently according to the moving direction before it is lost.The third problem is that when the target is incomplete in the image,the control algorithm based on projective geometry will not work.We solved it by combining visual servo control.At last,we realize autonomous target following and landing on it for UAVs.Furthermore,we analysed the ambiguity problem cased by projective geometry when using monocular camera.Then we analysed the localization accuracy of the vision system and the estimation accuracy of the target velocity based on the kalman filter.At last we analysed the efficiency of the controller for UAV landing.