A particle filter tracking algorithm based on adaptive feature fusion strategy

Abstract

The design and implementation of a particle filter tracking algorithm based on adaptive feature fusion of color histogram and edge orientation histogram is introduced. Experimental results show that the feature fusion tracking algorithm is more robust, especially when the target is moving in a varying environment, compared to that of single feature tracking algorithms. The adoption of two features increased the computational complexity inevitably. To avoid degeneracy of tracking speed, integral edge orientation images are built up. The final algorithm, running on a Pentium IV computer, can track pedestrians walking at normal speed effectively.