摘要
提出一种多线索动态融合和目标区域划分的粒子滤波视觉跟踪算法。在粒子滤波框架基础上,选取颜色、纹理、边缘线索于目标模型中,采用带权重的乘性融合策略自适应计算粒子权重,并实时更新目标模型。为增强在遮挡时的跟踪能力,采用局部目标而非整个运动目标作为粒子目标模型。实验结果表明,改进后的算法比简单的线索融合、传统的粒子滤波模型选取方法更能鲁棒并实时地跟踪目标。
This paper presents visual cues fusion and tracking local object properties for object tracking in video sequences using particle filtering. The visual cues, color, edge and texture, form the likelihood of the developed particle filter, a method for self-adaptively weighted product fusion strategy is proposed, and the cues real-time is updated. By using local object properties instead of the global ones, the performance of the tracker is greatly improved when the object undergoes partial occlusion. The results show that the proposal is more robust than simple cue fusing or conventional particle filter, and fast enough for real-time applications.
出处
《计算机工程与应用》
CSCD
2013年第19期186-190,共5页
Computer Engineering and Applications
关键词
粒子滤波
多线索
融合策略
遮挡
particle filter
multiple cues
fusion strategy
occlusion
