Development of a Vision-Based Ground Target Detection and Tracking System for a Small Unmanned Helicopter
Feng Lin,Kai-Yew Lum,Ben M. Chen,Tong H. Lee
DOI: https://doi.org/10.1007/s11432-009-0187-5
2009-01-01
Abstract:It is undoubted that the latest trend in the unmanned aerial vehicles (UAVs) community is towards visionbased unmanned small-scale helicopter, utilizing the maneuvering capabilities of the helicopter and the rich information of visual sensors, in order to arrive at a versatile platform for a variety of applications such as navigation, surveillance, tracking, etc. In this paper, we present the development of a visionbased ground target detection and tracking system for a small UAV helicopter. More specifically, we propose a real-time vision algorithm, based on moment invariants and two-stage pattern recognition, to achieve automatic ground target detection. In the proposed algorithm, the key geometry features of the target are extracted to detect and identify the target. Simultaneously, a Kalman filter is used to estimate and predict the position of the target, referred to as dynamic features, based on its motion model. These dynamic features are then combined with geometry features to identify the target in the second-stage of pattern recognition, when geometry features of the target change significantly due to noise and disturbance in the environment. Once the target is identified, an automatic control scheme is utilized to control the pan/tilt visual mechanism mounted on the helicopter such that the identified target is to be tracked at the center of the captured images. Experimental results based on images captured by the small-scale unmanned helicopter, SheLion, in actual flight tests demonstrate the effectiveness and robustness of the overall system.