基于色彩相关直方图与SIFT特征融合的粒子滤波跟踪算法

Particle Filer Tracking Algorithm Based on Fusion of Color Correlogram and SIFT Features

论文提出了基于色彩相关直方图与尺度不变性特征(SIFT)相融合的粒子滤波跟踪算法,该方法克服了在跟踪过程中运动目标的遇到形变以及目标被短暂遮挡丢失的问题。特别是提高了当运动目标与背景相近的情况下的跟踪效果。该算法在粒子滤波算法基础上将目标色彩相关直方图作为目标的一种描述特征,用于评价不同预测状态与观测状态之间的色彩相关性。依靠尺度不变性特征保证克服运动目标的仿射变化的特性,将色彩相关性特征与尺度不变性特征进行特征融合,从而提高目标描述的可靠性及跟踪的准确性。实验结果表明,该算法优于传统粒子滤波的跟踪算法。

This paper proposes a tracking algorithm based on particle filter algorithm with color correlogram and SIFT features.In the process of tracking,this algorithm has attractive result to resolve the problem of target losing no matter when the target was obscured or when target deformation occurred.The tracking effect is especially improved when moving target is similar to background in color.To evaluate different color correlation between prediction and observation state,the algorithm adopts color correlogram as one describe feature.To overcome the affine change characteristics of target,the SIFT feature is also adopted.With the color correlogram and SIFT features fusion,both the reliability of features description and accuracy of tracking is improved.This algorithm is superior to the traditional particle filter tracking algorithm.