Abstract: A parallel particle filter object tracking algorithm is given out,which is based on multiple feature fusion with the help of GPU (Graphic Processing Unit) technology. Due to the limitation of the model representation based on single visual feature, two complementary visual features, which are gray histogram and gradient histogram, are used in the algorithm to improve the tracking stability and accuracy. Moreover, to handle the large amount computation cost of the particle filter, a GPU parallel optimized scheme is designed to improve the algorithm speed. and can meet the real-time application requirement.