基于形态学的硬币镜面区域缺陷检测算法研究

Research on Defect Detection Algorithm Based on Coin Mirror Surface

在机器视觉领域中金属产品表面缺陷检测是生产加工中的重要环节,而对于具有兼有镜面特性和纹理特性的硬币进行检测,则更加复杂.为了更有效地检测出硬币表面的缺陷,可将硬币表面分为2部分检测:一个是镜面部分,另一个是纹理部分.针对硬币镜面部分的检测提出了一种基于形态学配准的算法:首先利用无缺陷的标准硬币作为模板图像进行二值化、中值滤波、边缘检测以及形态学处理,然后再将待处理的硬币图像进行除形态学外的相同处理,最后对2幅图像的特征配准算法和差分将匹配错误的位置作为缺陷进行提取.实验结果表明,本方法可以有效检测出硬币镜面部分的缺陷.

In machine vision, the defect detection of metal surface is a very important step; meanwhile, the detection of coin with mirror section and texture section is difficult. In order to achieve an efficient detection of surface defects of coins, the surface of a coin can be divided into two parts： one is mirror area; the other is texture area. In this paper, algorithm based on morphology is proposed for the detection of the coin mir- ror. Firstly, a coin without defect, as the template, is preprocessed by binary, reed filter, edge detection and morphology; then the coin with defect is preprocessed in the same way without morphology. At last, registra- tion algorithm and the difference of two different images can locate the defects which can be extracted. The experiment shows that the defects of the mirror surface of the coins can be effectively detected.