Research On Monocular Vision Method For Ground Targets’ Pose Measurement With The UAV

With the rapid development of imaging technology, the demand for measurementof ground targets using the UAV with imaging equipments is more and more extensive.But compared with the observation distance, limited by factors of the baseline’s length,the UAV’s own size and the installation location, the UAV and the other space-basedplatforms are not suitable for installing multiple visual imaging equipments, andgenerally installed only one imaging equipment, and form the monocular vision system.The monocular vision measurement method is that using only one imagingequipment to capture images to measure the targets’ geometric dimensions, pose. Themonocular vision system not only has no need to solve the optimal distance betweentwo cameras, feature points’ matching problem in the binocular vision system, but alsodoesn’t bring about serious aberrations like the omnidirectional sensors. It has moreadvantages than the binocular vision system in terms of camera installation, field ofview adjustment, calibration of the camera parameters. More importantly, limited byfactors of installation platform, site on many occasions, the binocular vision systemcann’t work, the monocular vision system becomes the only option. In the field ofcomputer vision, how to use the monocular vision system to solve the actual targets’pose has become an important direction.This thesis aims to study the monocular visionmeasurement method for ground targets’ pose problem. Finished works are as follows:1. systematically search and summary the current research situation of the monocularvision measurement method, and detailed introduce9kinds of common methods.2. Be aimed at the problem of ground moving targets’ position with monocular of theUAV, a new method based on the ground static scene around the target is designedand realized. The method’s measurement error is analyzed by the theory of ErrorAnalysis of Covariance, and proved to be correct by the simulation experiment, andthe author has analyzed the sources of error that affect the measurement accuracy,and the results are consistent with the theoretical error analysis. At last, the methodis proved to be feasible by the3groups of physical experiment.3. Be aimed at the ground targets like radar base, and so on, that has special shape onits surface(like circle, cylinder, and so on), a monocular vision measurement methodfor round face targets’ pose and shape measurement is researched and analyzed. Themethod fully makes use of the Geometry Constraints and the Similar TrianglesPrinciple, and get the targets’ pose only using one image shot by only one camera,and we can get the targets’ real-time pose by the image sequences shot by the UAV.The thesis deduce the fundamental and get the corresponding simulation.