Pedestrian tracking for infrared image sequence based on trajectory manifold of spatio-temporal slice

摘要

The research of pedestrian tracking in infrared image sequences is a curial part of video surveillance. Considering the particular characteristics of the infrared image, such as low contrast, fuzzy edge and unknown noises interference, the study of infrared pedestrian tracking algorithm becomes a great challenge. Spatio-temporal slice method is effective due to considering both spatial and time scale. It can extract the trajectory of moving targets, reflecting the trajectory manifold variations of targets along the time, to provide ways to depict the regions of targets. However, traditional spatio-temporal based methods only consider the horizontal slice analysis and usually require a large amount of calculation time; this paper proposes a spatio-temporal tracking algorithm to infrared image sequences, using both horizontal and vertical multi-layer slices to obtain the integral trajectory manifold. The integral trajectory is analyzed to obtain the target boundary and position information, with which the target can be tracked in each frame. The experimental results show that the proposed method has a relatively high tracking accuracy with a fast computing speed. Moreover, it can perform effectively in different infrared image sequences with various motion modes by single pedestrian from OTCBVS/05 Terravic Motion IR Database.