多特征融合的自适应性粒子滤波跟踪算法

为提高粒子滤波视觉目标跟踪算法的实时性与鲁棒性，提出了一种基于多特征融合的自适应性粒子滤波跟踪算法。该算法利用颜色和结构特征表示目标，将两者融合于粒子滤波的框架中，利用融合后的信息计算粒子的权值，以降低算法受目标形变及复杂环境的影响。同时，根据跟踪预测的准确程度动态计算跟踪所需的粒子数目，对采样粒子集进行自适应调整，以提高粒子质量，降低粒子数量，减少算法运算时间。实验结果表明，所提算法对于每帧图像的平均计算时间相对于传统混合跟踪算法缩短了将近一半，而且算法的鲁棒性较强。

Adaptive particle filter tracking algorithm by fusing multi-features

To improve the real-time and robustness performance of particles filter algorithm for tracking vision objects, an adaptive particle filter tracking method based on multi-feature fusion is proposed. The proposed method uses the color and structural features to present the interested target. These features are integrated in the frame of particle filter, and the weights of particles are calculated by this integration, in order to conquer the distractions from the target deformation and the complex background. Meanwhile, particle number is calculated dynamically according to the tracking accuracy, and the particle-set is also adjusted adaptively, in order to promote the quality of particle and reduce its quantity, and then the cost of calculation is reduced. The experimental results show that the average of each frame’s operation time of the pro-posed method is nearly half of classic hybrid algorithm, and the proposed method is of higher robustness.