关键词: 二维主成分分析, 双向二维主成分分析, 目标跟踪, 增量学习

Abstract: An object tracking algorithm based on bidirectional two-dimensional principle component analysis（Bi-2DPCA） is proposed. Object representation based on Bi-2DPCA is used to generate the object image subspace. To increase the speed of algorithm, an incremental learning method based on proposed adaptive incremental factor according to the object image match degree is adopted to update the related mean matrix and covariance matrices. The comparative experiments on classical image sequences containing dynamic backgrounds have been carried out, the results show the proposed algorithm is capable of tracking object accurately even in case of partial occlusion, and more efficient than the algorithm based on two-dimensional principle component analysis.

Key words: two-dimensional　principle component analysis, bidirectional two-dimensional principle　component　analysis, object　tracking, incremental　learning