中文摘要：

针对现有带钢表面缺陷检测方法检测率低,难以满足高质量带钢生产的需求。本文融合Gabor小波变换和加权马氏距离阈值化方法,提出了一种新的带钢缺陷检测算法。首先通过实验获取Gabor滤波器一组最优参数,对采集到的样本图像做Gabor模板卷积,得到边缘图像。然后对图像像素点进行加权处理,重新估计马氏距离的协方差矩阵,增大感兴趣区域像素点权重,得到任意灰度值样本与总体样本加权的马氏距离,增强了目标缺陷的边缘特征。最后利用连通区域标记法,搜索并标记带钢缺陷位置,完成了缺陷检测。实验对比表明本文提出的带钢缺陷检测新算法检测率为94.13%,漏检率为4.87%,验证了算法的有效性。

英文摘要：

According to the low detection rate for existing strip surface defect detection methods,which are difficult to meet the demand of high quality steel strip production,a new detection algorithm based on Gabor wavelet transformation and improved threshold for weighted Mahalanobis distance is proposed in this paper. Firstly,a set of optimal Gabor filter parameter is gained via the result of experiment analysis,and the Gabor template convolution for sample image is employed to obtain the edge image. Then,the image pixel data is weighted,so as to reevaluate covariance matrix of Mahalanobis distance. Furthermore,the Mahalanobis distance between arbitrary gray value sample and total sample is obtained,and the edge feature of image is enhanced. Finally,the defect position of the strip is searched and marked by employing the connected component labeling method,and the defect detection is finished. The experimental results illustrate that the defect detection rate of strip steel detection is94. 13% and the missing rate is 4. 87%. So the effectiveness of the algorithm is verified.