Abstract: A novel tracking method based on PSO（Particle Swarm Optimization） algorithm and Uniform LBP（Local Binary Pattern） feature is proposed for robust fragment tracking using exhaustive search strategy, sensitive to light and other issues. On the basis of the robust fragment tracking method using gray integral histogram as a feature, add the Uniform LBP feature descriptor, which has computational efficiency and is invariant to illumination. PSO algorithm is applied to the search for candidate target, because of its high accuracy, and fast convergence characteristics. Compared to robust fragment tracking, experimental results show that the proposed tracking method has satisfactory tracking precision and good real-time performance in complex backgrounds with illumination variations and completely occlusions.

Key words: target tracking, Particle Swarm Optimization（PSO）, fragment, Uniform Local Binary Pattern

摘要： 针对鲁棒分块跟踪采用穷举的搜索策略以及对光照敏感等问题，提出了一种基于粒子群优化算法和Uniform LBP特征的分块跟踪方法。利用统一的局部二值模式（Uniform Local Binary Pattern）特征对光照的不变性以及计算效率高的特点，在原鲁棒分块跟踪方法以灰度积分直方图作为特征的基础上，添加了Uniform LBP特征；利用粒子群优化算法具有精度高，收敛快的特点，将PSO算法运用到对候选目标的搜索中。实验结果表明，在不降低算法运行速度的情况下，以及光照变化较大，短时间目标完全遮挡的跟踪环境下，该算法鲁棒性显著增强。

关键词: 目标跟踪, 粒子群优化算法, 分块, Uniform LBP特征