中文摘要：

针对现有跟踪主流算法对目标机动性、目标遮挡和目标背景干扰综合性能不强的现状，改进算法利用组合分片模型和粒子滤波算法的结合来提升综合性能，提高跟踪算法准确性。改进算法采用粒子滤波算法，同时通过优化组合重采样算法提高算法的跟踪性能；组合分片模型结合水平竖直分片模型和环形分片模型的优点，通过Bhattacharyya系数进行模型相似性度量，高效克服人脸跟踪中遮挡问题和背景干扰问题。实验通过改进算法和对比算法在多变化人脸视频集进行跟踪，证明改进算法提高了对人脸目标的跟踪成功率，实现了算法对目标机动性、目标遮挡和目标背景干扰三种因素综合性能的提升。

英文摘要：

There are mainly three factors that affect the tracking accuracy of the human face, namely, motility, occlusion and background interference of the target. However, existing mainstream algorithm usually concentrate on improving only one fac- tor＇ s tracking performance. This paper intended to improve the comprehensive performance of all the three factors by combining combination fragment model and filter particle. Particle filter algorithm worked well on non-linear mobile targets and boosted tracking performance of the algorithm by applying optimization combination re-sampling algorithm in re-sampling. Combination fragment models that were made up of horizontal ＆ vertical fragment model and circular fragment model had these two models＇ advantages, conducted similarity measure with Bhattaeharyya coefficient and improved tracking accuracy when human face was occluded and interfered by background. The experiments applied optimization combination re-sampling algorithm into particle filter algorithm and applied combination fragment model into target modeling tracks changing human face. Aiming at solving motility, occlusion and background interference of the target, this paper greatly improves tracking effect, increases tracking success rate and achieves desired goal by enhancing the algorithm.