Deep-learning-based moving target detection for unmanned air vehicles

Abstract

In this paper, a deep learning network is investigated to detect moving targets for a UAV equipped with monocular camera. An algorithm based on fully convolutional network is proposed to obtain the position and moving direction of targets. A Kalman filter is incorporated into the proposed algorithm to increase the accuracy of target position information acquisition. The experimental results show the effectiveness of the proposed algorithm with a relatively low hardware resource consumption.