关键词: 目标跟踪, 光流法, 尺度不变特征, 模板更新

Abstract: This paper presents an object tracking algorithm that combines the optical flow estimation algorithm and the Scale Invariant Feature Transform（SIFT） with a new template update strategy for the change of scene and object. The SIFT feature is local feature which is invariant to the scale and rotation change of the image. The optical flow is a velocity field and the whole feature which represents the change of intensity of the pixel. The SIFT feature satisfies the condition of the optical flow estimation method. The experimental results show that this improved method can be used in partial covering tracking, and can achieve more accurate tracking than the traditional method.

Key words: object tracking, optical flow, Scale Invariant Feature Transform（SIFT） feature extraction, template update