基于局部纹理特征的钢卷侧面缺陷检测方法研究

Study of side image of steel coil defects detection based on local textural feature

针对钢卷侧面图像的纹理结构,提出了一套有效的钢卷侧面图像检测方法和一种新的Gabor滤波器参数设计方法,并将Gabor滤波器与LOG滤波器相结合对图像做预处理,有效的滤除了噪声并增强了图像纹理结构。另外,改进了基于"散度"的Zernike矩特征选取方法并结合SVM对预处理后的钢卷侧面图像子块分类,从而检测目标图像中是否含有缺陷。

Accroding to the texture structure of the side image of steel coil（SISC）,an effective scheme of SISC detection is put forward.A new design method of Gabor filter parameters is proposed.The pretreatment based on the Gabor filter and the LOG filter which can filter noise and enhance the structure of texture effectively.In addition,a feature extraction method of Zernike moment based on a con-ception of scatter is proposed,and combine with SVM to classificate the SISC.So we can detect whether there are defects in the image.The experimental results show that the whole scheme works well in defects detection of the SISC.