粒子滤波是适用于非线性非高斯系统下目标跟踪的强有力工具.MiroSot足球机器人系统可以作为研究机动目标跟踪问题的平台.对此,在分析MiroSot系统目标特征的基础上,提出一种基于目标特征约束的均值漂移粒子滤波算法,利用约束和优化的思想提高粒子的质量并减少其数量.对比实验表明,该方法有效地克服了传统粒子滤波的计算量和粒子退化问题,保证了多机动目标跟踪的准确性和实时性. Particle filter is a powerful tool for target tracking under non-linear non-Gaussian system.MiroSot soccer robot system can be used as a platform to study maneuvering target tracking problem.On the basis of analyzing the target characteristics of MiroSot system, The feature-constrained mean-shift particle filter algorithm, which uses the idea of ​​constraint and optimization to improve the quality and reduce the number of particles.Comparison experiments show that this method effectively overcomes the computational complexity and particle degeneration problems of traditional particle filter, and guarantees the multi-maneuvering target Tracking accuracy and real-time.