Fusion of Millimeter Wave Radar and Monocular Vision for Obstacle Detection Based on Rotor-Wing UAV

This paper presents a framework for detecting the obstacles from a rotor-wing UAV by fusing the information from millimeter-wave radar and monocular camera. In order to implement the angle measurement function of ranging radar, we designed a dual radar network system to locate the target position, and through calibration we achieved correspondence between target orientation and image area. In order to deal with the complex movements of UAV (such as yaw and pitch), this paper proposes a robust algorithm, which can calculate the accurate height of target position points in the image as well as the bottom boundary of the target area in the image, thus avoiding over segmentation of the target. Finally, the target is segmented from the background of the partition after fusion through the threshold algorithm according to the mutational gray value of the bottom boundary of the target position. Through extensive experiments, the effectiveness and efficiency of proposed algorithm have been proved and the experimental results show that the proposed method is novel, simple and feasible.