Landing-trajectory Estimation For The UAV Guided By Multi-view Ground-based Vision

The research on the safe landing guidance technology of light and small UAVs in GNSS-denied environment and the target monitoring and trajectory estimation technology of low-altitude small UAVs has important academic value and practical significance.However,in application scenarios such as guiding UAVs to land safely and small UAV target surveillance based on visual means,problems such as low trajectory estimation accuracy and insufficient data stability have become difficult problems.In addition,during the execution of missions,UAVs may encounter navigation sensor failures,especially in landing missions,where traditional navigation methods may have potential safety hazards.Therefore,the emergence of a reliable UAV trajectory estimation scheme is particularly important.This paper studies a fixed-wing UAV landing technology based on multi-view ground-based vision guidance.On the basis of evaluating the UAV trajectory estimation scheme based on a single vision sensor,this paper discusses,designs and evaluates a decentralized multi-view lowaltitude UAV target trajectory estimation scheme.Compared with the single vision sensor solution,this research expands the perception distance and field of view of the groundbased vision system by distributing multiple low-cost cameras in the front,rear,left and right sides of the scene,so as to meet the needs of full-range,continuous and stable landing trajectory estimation.Improve trajectory estimation accuracy and reliability.This research mainly includes the following two aspects of work:1.Aiming at the problem of insufficient stability of position estimation data in the single vision sensor scheme,a target occlusion detection algorithm and information complementary strategy based on multi-view scenes are designed.By using multiple ground-based vision cameras to work together for data mapping and replacement,this design scheme can effectively reduce the probability of frame loss of UAV position estimation information,thereby improving the continuity of trajectory data.2.In order to solve the problem of low overall trajectory estimation accuracy of the vision-guided UAV landing system,a trajectory fusion algorithm based on a multi-view decentralized deployment scheme is further designed to improve the accuracy of UAV trajectory estimation.First of all,in the scenario of multi-directional deployment of ground-based vision cameras,the perception distance of UAVs in a relatively long range is improved.Secondly,through the trajectory fusion algorithm based on multi-camera,the overall trajectory estimation effect is improved.The simulation verification results show that under the improved scheme,compared with the single-view scheme.In the overall trajectory estimation results,the lateral and lateral errors are reduced by 27%,the longitudinal errors are reduced by 56%,the altitude direction errors are reduced by 41%,and the spatial distance errors are reduced by 18%.Furthermore,the similarity between the estimated trajectory and the true trajectory increases by 47%.This thesis is based on the ground-based vision-guided landing system developed by the research group earlier,and has been improved and optimized.The module of occlusion detection and information complementation is developed,and a multi-view trajectory fusion algorithm is designed to improve the accuracy of the overall trajectory estimation of the UAV.These modules support ground-based vision systems for data collection and evaluation with multiple cameras in a dispersed arrangement.On this basis,this paper comprehensively verifies the accuracy of UAV trajectory estimation under various schemes.