基于变分水平集的目标轮廓分割方法  

Segmentation of object boundary based on variational level set method

作　　者：付增良[1] 秦永明[1] 张江[1] 张尚彬[1] 

机构地区：[1]中国航天空气动力技术研究院,北京100074

出　　处：《计算机工程与设计》2013年第7期2512-2516,共5页Computer Engineering and Design

摘　　要：针对风洞试验摄影测量技术中标识点图像边缘模糊,难以快速、准确分割的难题,设计了一种基于变分水平集的目标轮廓自动分割方法。分割模型根据图像域内零交叉曲线两侧二阶微分值符号相反的特性,构造了新的曲线演化引导函数,位于标识点内部、外部,甚至与标志点轮廓相交叉的初始曲线,在该引导项的指引下均可自动收敛到标识点边界处。同时,模型中引入新的边界函数,解决了标识点弱边缘分割过程中演化曲线边界泄露的问题。在合成图像和实际风洞模型图像上的仿真实验结果表明了该方法的准确性与优越性。In view of the difficulty in accurately and fast segmentation of low-quality marker image that obtained in wind tunnel test with photogrammetric technique, a novel automatic segmentation scheme based on variational level set method is proposed. According to the property that the sign of the second derivative is opposite on either side of zero-crossing curve, a curve evolution indicating term is given. The evolving contour, located inside, outside or intersected with object boundary, can correctly converge to the desired position guided by the evolving term. Moreover, an edge stopping function is also introduced to avoid the intractable boundary leakage problem, Experiments on both synthetic and the real image of wind tunnel model demonstrate the feasibility and superiority of the proposed method.

关 键 词：风洞试验 轮廓提取 零交叉 变分水平集 二阶微分 边界停止函数 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]