Abstract:

A novel method for infrared target tracking is proposed based on particle filtering and hierarchical shape representation. Under the theory framework of particle filtering, the establishment of observation probability models is an important factor affecting the infrared target tracking performance. Infrared target representation is the foundation to build observation probability models. For the characteristics of infrared imaging, the hierarchical histogram of gradient and spatial pyramid technologies are applied as a basis to implement the hierarchical shape representation of infrared targets. Through computing the Bhattacharyya distance between the  hierarchical  histogram of gradient of reference targets and the one of target samples, the observation probability model is constructed. Two infrared target tracking experiments with different scene characteristics and comparative analysis demonstrate that the method is effective and robust.