Random Feature Point Selection Method for Landing Position Detection Based on Vector Constraint

Abstract

The invention discloses a random feature point selection method for landing position detection based on vector constraints, comprising the steps of: 1. Acquisition and synchronous upload of landing area images; 2. Landing area image processing, the process is as follows: 201. Initial sampling Landing area image processing at time; 202. Image processing of landing area at the next sampling time, including steps: Harris corner point extraction, Harris corner point matching, SIFT feature extraction, SIFT feature point matching, feature point fusion, feature point combination generation and Optimal feature point combination screening; 203, return to step 202, and process the image of the landing area at the next sampling time. The method of the invention has simple steps, reasonable design, convenient implementation, strong real-time performance, good use effect, can simply and quickly select feature points for pose estimation from feature points extracted from images of unknown landing areas, and can effectively improve pose estimation precision.

Description

technical field [0001] The invention belongs to the technical field of visual navigation, in particular to a random feature point selection method for landing position detection based on vector constraints. Background technique [0002] When UAVs perform rescue and search tasks, they are faced with the problem of emergency landing with unknown terrain in the drop zone, complex and no ground auxiliary navigation equipment guidance. Due to the cumulative error of inertial navigation and the susceptibility of GPS to interference, there are huge safety hazards when UAVs land in complex and unknown environments. Visual relative navigation has the advantages of simple equipment, large amount of information, strong concealment, and good autonomy. It is widely used in the fields of UAV autonomous landing / landing, autonomous refueling in the air, and spacecraft rendezvous and docking. Visual relative navigation has the advantages of strong autonomy, passiveness, and high guidance ac...

AI Technical Summary

Problems solved by technology

[0003] Compared with the relative pose estimation in the known landing area, there are two main problems in the relative pose parameter estimation of the UAV in the unknown landing area: one is the geographic coordinates of the feature points used to solve the pose equation It is impossible to calibrate in advance, so how to calibrate the geographic coordinates of the feature points online is a difficult problem; second, the number of feature points extracted in the unknown landing area is large and random, and how to select the feature points for pose estimation to make the pose estimation accuracy high, Good real-time performance is also a problem

Images