基于恒定曲率变化的主动轮廓模型

随着主动轮廓模型（又称snake模型）被广泛应用于无人参与的自动化任务,人们对模型的鲁棒性和自适应能力提出了更高的要求.而传统内部力模型的收缩效应,过平滑作用,及缩放可变性,导致了内部力模型的参数调整困难,很难由程序自动进行调整.据此,提出了基于恒定曲率变化的内部力模型.该模型不仅具有缩放不变性,而且在保证轮廓光滑连续的同时,未引入其他的副作用（如收缩,过平滑等）,提高了模型参数的鲁棒性.实验结果表明,该模型精确提取目标轮廓的能力得到增强,能够成功提取存在高曲率位置的凹陷轮廓,而且内部力大小衰减迅速,对轮廓点的增删不敏感,保证了模型快速稳定地收敛到期望轮廓.

Constantly Changing Curvature Based Active Contour Model

As the active contour model(Snake model) is widely used in automatic missions,higher robustness and adaptability of the model are required.Due to the contraction,over-flatten effect on contour and scale variance of the traditional model,the adjustment of its parameter is quite difficult and hard to be automatically decided by program.Our constantly changing curvature based internal force model is scale invariant.Besides keeping the smoothness and continuity of the contour,no side effect is introduced and the robustness of model parameters is increased greatly.Experimental results show that the proposed model greatly improves the ability of detecting the accurate contour of the target,and even works well for the contour with high curvature position and concavity.Because of fast attenuation of internal force and insensitivity to addition and deletion of contour points,the model is guaranteed to converge quickly and steadily to the expected contour.