Template Matching Algorithm Based on Shape-Varying Object Model in IR Image Sequence

摘要

This paper presents a template matching algorithm with an object model for a shape-varying object in IR image sequence. We develop the object modelling method and adopt each model to the object which is being matched. The refined object model is more characteristic of the shape-varying object and therefore increases the robustness of the track. We also update the reference template adoptively using the object model when the object changes its shape or size largely during the track. Each of these refinements contributes in their own way to better overall performance of the tracking algorithm. Experiments on IR image sequences show that the proposed method could track a shape-varying object in real time.