一种改进的颜色粒子滤波目标跟踪算法

An Improved Color-Based Particle Filter Algorithm for Target Tracking

为实现运动目标精确跟踪,克服跟踪过程中目标的非线性运动以及由目标形变、遮挡和光照等因素带来的影响,本文提出了一种改进的颜色粒子滤波方法.算法从提高目标模型描述能力入手,首先对直方图加权函数进行了改进,使模型对区域特征描述更加合理;然后针对颜色直方图特征对光照明敏感、易受环境干扰等缺点,将目标由颜色特征空间映射到对光照稳定、抗几何失真能力强的局部熵特征空间,构建了颜色局部熵观测模型;同时设计了目标模板的自适应更新策略,当目标受到严重干扰的时候动态调节粒子数目.实验结果表明相比传统的颜色粒子滤波算法,本文算法具有更好的鲁棒性,能够在存在遮挡、光照变化、非线性运动等情况下实现稳定跟踪.

To overcome the disadvantages that the traditional particle filters based on color histogram is susceptible to environmental interference and illumination variations, an improved particle filter algorithm was proposed. This article starts from improving the description ability of the target feature model. First, the histogram weighted function was optimized. Second, for the shortcoming of the color feature, a new color local entropy target observation model was constructed by mapping the target from color feature space to local entropy space. In addition, in order to make the model better adjust to environmental interference and target deformation, an adaptive updating strategy of the target model was designed and the number of particle was adjusted dynamically. Experimental results demonstrate that the proposed algorithm is effective.