EKF based object detect and tracking for UAV by using visual-attention-model

Abstract

Visual-attention-model (VAM) is a kind of model with good robustness of bionic vision. For ground object sensed on UAV (Unmanned Aerial Vehicle) platform, in this paper object detection and tracking algorithm based on VAM and extended Kalman filter (EKF) is proposed, and applied to the ground surroundings for object detection and tracking. In order to quickly extract the ground objects in aerial images, visual saliency map was calculated. Based on the robust detection of ground objects and EKF optima estimation, the proposed algorithm based on VAM and EKF could robustly detect and track object for UAV. Experimental results show that the algorithm in this paper is capable to be adapting to complex ground surroundings for object detection and tracking, the designed visual saliency map is not only de-noise images effectively but also can help reserve original information as possible. In addition, calculation of the proposed algorithm is pretty simple, and so suitable for engineering application.