Optimization of Feature Point Detection and Matching Algorithm Based on Quadcopter Platform

Abstract

Aiming at the problems of poor tracking accuracy, inadequate anti-jamming ability and unavoidable monitoring unreachable angle in the traditional tracking method of four-rotor unmanned aerial vehicle (UAV), based on compressed sensing theory and combined with FAST (Features From Accelerated Segment Test) and SURF (Speeded Up Robust Features) algorithm, this paper proposes a fast matching algorithm for multi-target detection and single-target tracking of UAV. We built a matching onboard hardware system, collected data through the camera, and then used feature point detection and matching algorithms to detect multiple moving objects. Finally, we used compressed sensing theory to quickly locate the tracked objects. Compared with the traditional algorithm, this algorithm needs much less time to achieve tracking in the same scene than the general tracking algorithm, reaching the millisecond level, and the tracking loss rate is only 5% for the object whose area is less than 256*256 pixels in the image, which greatly improves the tracking accuracy and antijamming performance.