Design And Implementation Of UAV Companion System In Outdoor Complex Environment

With its advantages of wide field of vision,long distance,high mobility and high flexibility,UAVs have a high application potential in man-machine collaboration and service fields.However,most of the existing UAVs cannot be controlled without human control,so the intelligent and autonomous research of UAVs has attracted much attention.The key to turn on the intelligence of UAV is to accurately identify the pedestrian position and have the ability of autonomous accompaniment.Therefore,aiming at the application of unmanned aerial vehicle accompanying pedestrians,this paper carries out the research on autonomous safe flight and target tracking of unmanned aerial vehicle,realizes the autonomous flight and target tracking of unmanned aerial vehicle based on vision in the complex outdoor environment,and proposes a set of lightweight unmanned aerial vehicle autonomous flight tracking system,which is of great significance for improving the application value of unmanned aerial vehicle.The main work contents are summarized as follows:In this paper,a multi-camera fusion based attitude estimation and tracking method for accompanying targets is proposed to solve the problem of UAV attitude estimation and tracking of accompanying targets in flight.Due to the scale irrelevance of monocular camera,the large amount of binocular camera calculation,the difficulty of target recognition and other problems,this paper creatively uses binocular camera to complete the calibration of the real distance of human bones,combined with the assumption that human shoulders and hips are in the same plane,the pose recovery problem is converted into a PNP(3D-2D projection)problem.The pose information of the adjoint target is recovered by monocular camera,and the adjoint target is tracked by Kalman filter.The experimental results show that the average error of the proposed method is less than 0.2m within 5m,which is more than 50% less than binocular.In this paper,real-time prediction of the trajectory of accompanying targets is carried out,research on accompanying strategy of UAV is carried out,and the problems of temporary loss of accompanying targets due to obstructions and autonomous accompanying flight of UAV under obstructions are solved according to environmental real-time path planning.Experiments show that the average success rate of trajectory prediction relocation is 69.6%,and the UAV can accompany the target efficiently and smoothly at a speed of 3m/s.After verifying the feasibility of visual positioning and perception on UAV,this paper independently designs mechanical hardware and builds a UAV accompanying system based on visual perception.Since the load capacity and computing capacity of the UAV are very limited,the binocular camera is selected as the main perception sensor of the UAV in this paper.Through the visual inertia mileage calculation(VIO)and binocular stereo matching algorithm,the real-time positioning information of the UAV and the depth information of the surrounding environment can be obtained.Finally,through the experimental flight verification,the UAV can reach the highest flight speed of 3m/s and depth perception of 20 m in the environment without GPS obstacles.All the above contents were verified by experiments on the self-designed and built UAV platform.The self-designed UAV greatly reduced the load pressure.The method proposed in this paper efficiently and accurately calculated the position and pose information of the accompanying target,and finally formulated an efficient tracking strategy and planned a reasonable path to complete the tracking of the accompanying target.