A particle filter algorithm for real-time multiple objects tracking based on color local entropy

To achieve accurate and real-time visual multi-object tracking, overcome the difficulties brought by the object deformation, occlusion, and illumination variations, a novel particle filter for multi-object tracking algorithm based on color local entropy (CLE) is proposed. This paper starts from improving the description ability of the object feature model. First, the traditional histogram weighted function is optimized. Second, for the shortcoming that the color feature is sensitive to illumination and environmental interference, a new color local entropy object observation model is constructed by mapping the object from color feature space to local entropy space which is stable to deformation and illumination. In addition, in order to make the algorithm better adjust to the object deformation and environmental interference, an adaptive updating strategy of the object model is designed and the number of particle is adjusted dynamically. The experimental results show that the proposed algorithm is effective for the real-time multiple objects tracking and it is applicable to both rigid and non-rigid object.