基于机器视觉的金属边缘细微缺陷检测方法的研究

Research on detection method for subtle defects of metal edges based on machine vision

针对金属表面质量人工检测工作量大、效率低等情况,提出一种机器视觉检测金属零件边缘细微缺陷的方法。先根据金属零件表面反光的特点,在亮场下垂直照射,运用形状模板匹配定位,对前景区域膨胀处理,截取包含零件边界信息的图像,缩小处理区域。接着锐化和滤波边缘区域,线性拟合边缘轮廓,提取拟合线段的方向向量,并以此为特征进行区域类划分,提取边缘坐标点。以提取的坐标点为圆心作圆领域,求取每个领域的灰度平均值并线性插值迭代剔除边界干扰点。最后提取符合要求的坐标点排序,重构多直线段,结合背景差分法提取缺陷。实验结果表明,该检测方法能够有效检测出金属边缘细微缺陷。

In view of the heavy workload and low efficiency of the surface quality manual inspection of metal parts,a method of detecting the slight defects of the edge of metal parts by machine vision is proposed.First,according to the characteristics of the surface reflection of metal parts,it use shape template matching to locate the part which is vertically irradiated under the bright field.Then,it expands the foreground area and captures the image containing the part boundary information,and reduce the processing area.After that,it sharpen and filter the edge region,and linearly fit the edge contour.It extracts the direction vector of the fitted segment which is classified by direction vector,and extracts edge coordinates.It uses the extracted coordinate points as the center of a circle area,and calculates the average value of gray in each area and eliminate boundary noise points by linear interpolation iteration.Finally,it extracts coordinate points that meet the requirements,and reconstructs the edge of lines with them,which are sorted,then the defect is extracted by simulating the background difference method.The experimental results show that the detection method can detect the fine edges of metal edges effectively.