Multiple Objects Tracking Based on Snake Model and Selective Attention Mechanism

In this paper we present a novelty multiple objects tracking (MOT) algorithm that relies on Snake and selective attention mechanism (SAM) for aerial image sequences. Snake model with five energy functions is used to segment contours of targets precisely. At the same time, in order to decrease the computational complexity, a SAM model is designed. In our method we view attention selection as a process of feature extraction and classification. So unlike most of other SAM models, which compute in pixels level absolutely, our approach focuses on the process of classifying high-level information about color, size, speed and relative location to introduce SAM. Binary BP classifier and linear criterion function are used to decide whether a target is noticed by camera or not. Simulation results show this approach can discriminate targets and segment their contours in different computational complexities under the assumption that no occlusions occur.