基于特征组合的视觉目标跟踪

针对大多目标跟踪算法仅关注目标特征的问题,提出同时采用目标特征及背景特征对目标进行跟踪.针对场景中灰度和梯度特征在跟踪过程中其分类性能动态变化,首先计算各特征似然度,然后根据各特征分类性能计算特征权重,求取加权似然图.根据该似然图,结合粒子滤波算法,进而确定待跟踪目标位置与尺寸.仿真结果表明,该方法对环境光照变化、目标尺寸变化、局部遮挡等均具有一定的鲁棒性.

Visual object tracking based on feature combination

Most of the object tracking algorithms only concentrate on the object features.In this paper,the object features and background features are utilized together during tracking process.As the classification ability of color and grads cue of the scene is varying during tracking,map image of every feature is calculated first,then the weights of every feature are acquired based on its classification ability,and the integrated likelihood image is obtained.Based on this likelihood image,together with particle filter,the position and size of the target is achieved.The experimental results show that this algorithm is robust to the partial occlusions,scale-change illumination and variations.