Object Tracking Combined Spatio-temporal Context Learning with Illumination Invariant Feature

Abstract

For improving the accuracy, robustness and real-time in object tracking such as surveillance and human actions recognition, this paper presents a novel object tracking method which combines the spatio- temporal context learning (STC) with the illumination invariant feature (IIF). We cast the object appearance model based on the locality sensitive histogram (LSH) and the IIF, and then matched the similarity between the target and the template by the STC algorithm which models the statistical correlation between the target and its surrounding regions. The proposed method makes full use of the advantages of the STC algorithm and the IIF, and it can diminish the influence of illumination change, pose variation and rotation. The experimental results demonstrate that the tracking precision and robustness is obviously improved when the proposed method is used in the two classic object tracking video databases. Extensive experimental results show that this method can be applied to the recognition of humans or vehicles.