基于复域Contourlet特征向量的连铸板坯表面特征分类方法  

Continuous casting slab surface feature classification method based on complex Contourlet feature vectors

作　　者：于基睿 汪泽浩[1] 李培玉[1] 

机构地区：[1]浙江大学机械工程学系,杭州310027

出　　处：《计算机应用》2014年第12期3660-3664,共5页journal of Computer Applications

摘　　要：针对连铸板坯表面缺陷特征检测的复杂性,提出了应用复域Contourlet分解特征图像的一种表面特征分类方法。其相对于传统方法具有更优的移不变性、方向选择性和图像特征提取率。通过将图像特征区域进行复域Contourlet分解,再经由方向滤波器组,提取Contourlet域的方向子带系数矩阵,再求取系数矩阵的能量、方差、偏斜度,作为一种量化图像的特征向量。将特征向量输入支持向量机进行图像特征识别训练,之后可以进行图像的各类特征识别和分类。工业现场测试的效果表明,通过该方法进行的表面特征分类准确率在90%左右,能够应用于连铸钢板坯的图像特征提取和板坯缺陷检测。Concerning the problem that it is complex to detect the surface feature of continuous casting slab, a surface trait extraction method based on complex Contourlet decomposition was developed. Compared to conventional method, the new one has the characteristics of shift invariance, excellent directional selectivity and higher retrieval rate. The image was decomposed in Contourlet domain and then the image＇ s sub-bands directional matrices were extracted by using directional filter bands, then the feature vector was constructed by the energy value, stand deviation and skewness. The support vector machine was trained by the feature vectors, to classify images. Industrial test result shows that the accurate rate of classifying surface features is about 90%, and it can be used in image feature extraction and slab flaw detection.

关 键 词：复域Contourlet分解 支持向量机 特征分类 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]