基于梯度矢量C-V模型的空心叶片图像分割  

Image segmentation of hollow blade mould based on C-V model using gradient vector

作　　者：刘震昆[1] 古忠涛[1] 廖晓波[1] 蔡勇[1] 

机构地区：[1]西南科技大学制造科学与工程学院制造过程检测重点实验室,四川绵阳621010

出　　处：《计算机应用研究》2015年第12期3814-3816,共3页Application Research of Computers

基　　金：国家重大科技专项资助项目(2012ZX04007-021-07)

摘　　要：空心叶片CT断面图像存在大片同质区,还有小的气膜孔及缺陷区,正确提取气膜孔及缺陷等局部形状对判断叶片质量有重要意义。基本C-V模型利用同质灰度分布信息对图像进行分割,没有考虑局部特征;在CV模型中加入轮廓梯度矢量扩散场,使模型在最小化求解过程中,曲线始终沿最速下降方向演化,最终达到并停留在期望的轮廓上。这样既利用了目标区域内外图像灰度的全局信息,也考虑了局部边缘形状特征,以便在识别图像大块同质区的同时,实现小的局部形状的正确分割。通过对空心叶片断面图像的分割,验证了该方法能有效收敛于真实边界。A large homogeneous area exist in the image of the hollow blade section, there are also some small film hole and flaw areas. However, the film hole and flaw shape are very important to the blade quality. The essential C-V model splits image with the homogeneous grayscale information, but the local information were not used. This paper introduced the profile gradient vector to C-V model. In this way, the curve could always evolved along the steepest descent direction in the optimization procedure, and uhimately rested on the anticipant profile. So it all applied the grayscale global information and the local edge feature in the image section. The small local shape could be identified accurately along with the large homogeneous area. The image segmentation experiment of the hollow blade section proves the effectiveness of this method.

关 键 词：水平集方法 C-V模型 梯度矢量 图像分割 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]