Researches On Pose Estimation Problem For UAV Vision-based Landing Guidance

Autonomous landing guidance is an urgent issue to be conquered for Unmanned Aerial Vehicle (UAV) development, and the Vision-Based technology is considered to be one of the most potential solutions to the specific issue. This thesis aims to study the pose estimation problem, which is the core problem of UAV Vision-based Landing Guidance. Finished works are as follows:(1) A systematically review of pose estimation is presented, which focuses on introduction of 5 classical algorithms of 3 categories-analytic algorithms, iterative algorithms and globally optimal algorithms.(2) Specific to the issue of pose estimation in practice, the applicable cases are discussed about some classical algorithms with deterioration of landmarks coplanarity, meanwhile, the orthogonal iteration algorithm and EPnP algorithm are improved to suit the coplane case. Specific to the complication of classical analytic algorithms, the special condition of landmarks coplanarity is adopted to put forward a analytic algorithm of pose estimation based on area ratio, which is of slightly lower accuracy but much less complication. In addition, the possible problems of "pose ambiguities" are rationally explained from geometry and algebra perspectives, two rational solutions are introduces as well.(3) Theoretical error lower bound of pose estimation were deduced by taking advantage of relative theories of Cramer-Rao lower bound. And the influence of pose estimation was analyzed accordingly in the category of the inner parameters calibration error, image extraction error and landmarks position error. The basic conclusions are as follows: First, image error has bigger influence on the measurement accuracy than landmarks position error; Second, errors of the same level have obvious different influence on the measurement accuracy in different measuring conditions.(4) The mathematical model with designing questions of optimal scheme was established for Vision-based Landing Guidance. Taking "CNAH Industrial Cup" International UAV competition as application background, an optimized scheme is presented in ways of simulated annealing algorithm, and the designing results are tested and analysed.