一种稳健的目标提取与跟踪算法

A Robust Scheme for Contour Extracting and Tracking

融合了GVF-Snakes算法与基于细粒度的遗传算法,提出了一种稳健的目标轮廓提取与跟踪算法.该算法通过使用边界约束替代能量计算改进了GVF-Snakes算法,降低了算法计算复杂度,提高了它的搜索速度;另外,通过引用细粒度遗传算法来筛选控制点序列,提高了算法对极端凹陷边缘和噪声干扰轮廓的提取能力.通过合成和自然图像的目标轮廓提取和跟踪实验,证明了本文提出的算法具有鲁棒性和精确性.

A new scheme is proposed to extract and track the object contour automatically in this paper, it combines the active contour based on the gradient vector flow( GVF-Snakes) and the genetic algorithm (GA) based on the fine-grained model. On the one hand, the GVF-Snakes is improved by using the edge criterion instead of the complex energy computation to reduce its complexity and speed up its search. On the other hand, the selection of the array of the reference points by the GA enhances the extracting performance of the extreme concave contours and noise-disturbing contours. Experiment on the synthetic and natural images demonstrate its robustness and accuracy.