基于颜色特征与SIFT特征自适应融合的粒子滤波跟踪算法

Particle filter tracking method of color and SIFT features based adaptive fusion

针对序列图像中的运动目标在跟踪过程中发生运动模糊以及部分遮挡的问题进行了研究,提出一种将改进的颜色直方图特征模型与尺度不变特征(SIFT)模型相融合的粒子滤波跟踪算法。采用基于模糊逻辑的方法,根据当前跟踪环境自适应调节两种特征信息的权重,从而实现特征信息间的融合,提高描述目标观测的可靠性。实验结果证明,该算法优于传统的单特征或采用固定权值的多特征目标跟踪算法。

In order to solve the problem of motion blur and partial occlusion during the object tracking in the sequence images, this paper proposed a particle filter tracking method based on the improved color histogram model and the scale invariant feature transform （SIFT） model. According to the track environment, it adaptively allocated the weights of the two types of target information by using fuzzy logic method, thus improved the reliability of target observation description by feature fusion. Experimental results show that the proposed method is superior to the traditional methods which is based on only one feature or fixed weight．