Ground-Based Target Localization Using Monocular Vision on a Quadcopter

Vision-based target localization in the air-to-ground scenes of quadcopter is a hot topic in the military and civilian applications. The current localization approaches mostly rely on the intrinsic parameters of camera which are needed to calibrate in advance. But if the calibration is inaccurate, the localization based on the focal length of camera will result in inaccuracy. Moreover, due to the autonomous flight of quadcopter, the pose changes of the quadcopter will lead to the erroneous locating results. In this paper, we propose an accurate ground-based target localization method by means of monocular vision on a quadcopter. Based on the field of view (FOV) angle of camera, we present a localization algorithm requiring no calibration. Meanwhile, we introduce the quadcopter pose superposition strategy and the IMU quaternion denoising method to eliminate the impact of quadcopter pose changes in flight. The simulation experimental results demonstrate the high performances of our locating method in both accuracy and robustness.