Abstract: To extract defect features of steel strip surface，it is required to find the edges of defect correctly at first.With the deficiency of existing method in operation of the de-noising capacity，edge detail protection in complex background and uneven light，a new method is presented to detect edges of defect in steel strip surface.This method is based on using median filtering to compute the total variation with respect to the central pixel in a filter window.The result images are sent to Canny edge detector to find the edges which are taken as preprocessing images.As the final step，the binary morphological operators are used to remove isolated points and smooth the lines in order to get connected edges.The experimental result shows that the method in this paper is characterized by high noise inhibiting ability and can overcome the influence of noise and uneven light to detection result effectively.The better edges in complex background are obtained.

Key words: order filter, edge detection, surface defects, median filter, total variation

摘要： 为了提取带钢表面缺陷特征，首先需要正确检测缺陷边缘。针对现有方法对于光照不均和复杂背景条件下噪声抑制以及保持图像细节方面的不足，提出一种新的基于序列滤波器的边缘检测方法。先将缺陷图片进行中值滤波，然后在新的滤波器窗口相对于中心元素计算总变差，结果图像作为预处理图像送入Canny边缘探测器进行边缘检测，最后采用形态学方法进行光滑处理，得到完整的缺陷边缘。实验表明该方法具有较强的抑制噪声能力，有效去除现场工况条件下粉尘、光照不均等因素对检测结果的影响，很好地提取了复杂背景条件下的目标边缘。

关键词: 序列滤波器, 边缘检测, 表面缺陷, 中值滤波器, 总变差