摘要： 对双前沿活动轮廓模型进行了改进，提出一种鲁棒的视频目标轮廓跟踪新方法。通过限制支撑区域的大小，提高了演化速度。拟气球力的引入，既解决了因支撑区域限制导致的均匀区域内演化停滞的问题，又使参数调整更加容易，跟踪更加灵活。能较好地跟踪运动速度变化较大和形变较严重的目标，处理速度较快、并能自动适应目标拓扑的变化，适用于动态背景的视频。实验结果证明了该算法的高效性和鲁棒性。

关键词: 目标跟踪, 曲线演化, 双前沿活动轮廓模型, 水平集

Abstract: A robust and novel method for video object contour tracking by improving the dual-front active contour model. is presented With the metho, the speed of the algorithm is increased greatly by support region restriction, and the adopted quasi-balloon force can both solve the convergence problem results from support region restriction and make tracking more flexible. The method can be adapted to track objects with abrupt speed change and large degree of deformation. It is fast and topology change adaptive. And it can be used even the background is dynamic. Experiments demonstrate its effectiveness and robustness.

Key words: object tracking, curve evolution, dual-front active contour model, level-set

中图分类号: 

• TP311