基于小波理论的铸件缺陷纹理分类分析

根据纹理特征，提出了铸铝及铸铁缺陷的自动分类方法，通过类人眼对铸件缺陷的感知方式，运用Ｇａｂｏｒ小波函数分析了缺陷的纹理。为满足高速在线检测的要求，提出了启发式滤波器参数选择原则及Ｇａｂｏｒ函数在空域中快速而简便的运算方法。

TEXTURE CLASSIFICATION ANALYSIS FOR CASTING DEFECTS BASED ON WAVELET THEORY

This paper proposes a automatic classification scheme for aluminum and magnesium casting defects according to texture features. A set of Gabor wavelets is employed to thoroughly analyze the defect textures, analogous to visual perception mechanism of human being. A heuristic filter selection approach and the easy and fast implementation of Gabor functions in the spatial domain are presented, so as to meet the high-speed on-line inspection requirement. A new method for measuring the areas of texton blobs is proposed with which the areas of texton blobs can be estimated from the filtered response I-mages. Combined with conventional energy feature analysis, this method can be adopted to discriminate casting defects on both type and grade. Experiment results show that the proposed method is effective.