Abstract: Aiming at the visual tracking drawbacks of standard particle filters under the conditions of complex environment and illumination transformation, a new particle filter was proposed based on color and gradient orientation feature to get over the problem of low robustness of the particle filter tracking method through unique feature of color.By designing the model for features fusion, the performance of only color feature was improved when the environment changed. In addition, the proposed target pattern update algorithm improved the adaptability to complex scene diversification. The experiment indicates that the proposed method is effective and robust for the visual tracking under complex backgrounds.

Key words: Target tracking, Particle filter, Features, Color, Gradient orientation