Multi-Feature Adaptive Target Tracking Algorithm Based on Rotational Inertia

Abstract

In motion target tracking algorithms based on video, traditional mean shift target tracking algorithms use color as the target feature, which is easily affected by homochromatic interference, and the adaptive adjustment of kernel function bandwidth is insufficient. A multi feature video object tracking algorithm based on rotational inertia is proposed in this paper, which uses target color and texture features to jointly create the target sample model. The target sample is projected to generate a probability density distribution map. In the density distribution area of the target, the method of rotational inertia is used to calculate length, width, and angle of the target, thereby adaptively adjusting kernel function bandwidth, and locking the size and angle of video target. The experimental results show that the algorithm can adaptively adjust video tracking window according to the size and angle of the target, and can resist interference from similar colors, with good robustness.