摘要
为解决突变运动下的目标跟踪问题,提出了一种基于视觉显著性的均值漂移跟踪算法,将视觉注意机制运用到均值漂移跟踪框架中,利用时空显著性算法对视频序列进行检测,生成视觉显著图,从视觉显著图对应的显著性区域中建立目标的颜色特征表示模型来实现运动目标跟踪。实验结果表明:该算法在摄像机摇晃等动态场景下可以较准确检测出时空均显著的目标,有效克服了在运动目标发生丢失和遮挡等情况下跟踪不稳定的问题,具有较强的鲁棒性,从而实现复杂场景下目标较准确的跟踪。
In order to solve target tracking problem under circumstances of abrupt motion,a meanshift tracking algorithm based on visual significance is proposed. The visual attention mechanism is applied to the framework of mean shift tracking,the video sequence is detected by using the spatial and temporal saliency algorithm,and the visual saliency map is generated. According to the visual figure corresponding significant area,from which target color characteristics of the saliency model is established to realize moving target tracking. Experimental results show that the proposed algorithm can accurately detect obvious spatial and temporal targets in the dynamic scene,and the algorithm effectively overcomes the problem of tracking instability in the case occlusion and loss of moving objects,it has strong robustness,so as to realize accurate tracking of target under complex scene.
出处
《传感器与微系统》
CSCD
2017年第6期130-133,137,共5页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61462052
31300938)
