基于自适应状态转移的混合跟踪算法

为提高粒子滤波视觉目标跟踪算法的准确性和实时性，提出一种基于自适应状态转移的混合跟踪算法。首先采用零阶自适应变化模型来获取目标的可能状态，然后利用均值漂移算法的局部优化特性找到后验概率的最大值。在多峰值情况下由粒子滤波随机产生粒子，用新的粒子集来确定目标的最终位置。实验结果表明，这种改进的算法在保证准确性的同时，降低了系统的计算时间。

Tracking algorithm based on adaptive state transition

To improve the accuracy and real-time performance of particle filter algorithm for tracking vision object, a tracking algorithm based on adaptive state transition is proposed. Firstly, it employs a zero-order adaptive model to obtain the state of the target, and then uses local optimization characteristics of the mean-shift algorithm to find the maximum value of posteriori probability. Particle filer is used to produce more samples at the case of multi-peaks and determine the final goal set position with the new particles. Experimental results show this combined tracker provides comparable accu-racy and reduces computation complexity.