Traditional particle filter-based tracking methods often fail to achieve precise tracking in complex video backgrounds．This article combines traditional particle filtering algorithm and Gaussian-Hermite algorithm to preprocess the video first. To improve tracking speed, this article uses an integral histogram to describe facial features, greatly reducing the feature calculation time and weight update time of particles. Additionally, dynamic fusion of color and edge features is implemented, combined with real-time adjustment of template parameters based on the dynamic evolution of facial morphology. The experiment verified that the method proposed in this paper can effectively enhance its stability and reliability when dealing with skin color, different lighting conditions, and other complex environmental factors.