基于Contourlet变换的连铸坯表面缺陷识别

根据连铸坯表面图像的特点,提出了一种基于Contourlet变换的连铸坯表面缺陷识别方法.通过Contourlet变换将样本图像分解成不同尺度和方向的子带,提取子带的Contourlet系数特征,并结合样本图像的纹理特征,得到一个高维的特征向量.利用监督核保局投影算法对高维特征向量进行降维,将降维后的低维特征向量输入支持向量机,对连铸坯表面图像进行分类识别.对现场采集到的裂纹、氧化铁皮、光照不均和渣痕四类样本图像进行实验,本文提出的识别方法对样本图像的识别率可达94.35%,优于基于Gabor小波的识别方法.

Recognition of surface defects in continuous casting slabs based on Contourlet transform

A new recognition method of surface defects based to the characteristics of continuous casting slabs. Sample images were on Contourlet transform was proposed according decomposed into multiple subbands with different scales and directions by Contourlet transform. The Contourlet coefficients of subbands and the textural features of sample images were combined into a high-dimensional feature vector. Supervised kernel locality preserving projection (SKLPP) was applied to the high-dimensional feature vector for dimension reduction, which resulted in a low-dimensional feature vector. The resulted feature vector was inputted to a support vector machine (SVM) for recognition of surface defects. The method was tested with sample images from an industrial production line, including cracks, scales, non-uniform illumination, and slags. The test results show that the recognition rate of these sample images is 94.35%, which is better than that by Gabor wavelet.