基于PCNN和FCM的钢带表面缺陷检测

Surface defects inspection of metal strips based on PCNN and FCM

针对目前还没有较好的方法正确的检测金属钢带表面缺陷,提出一种结合耦合神经网络(PCNN)和模糊C-均值(FCM)的钢带表面缺陷检测算法,首先通过有效性指数求得聚类中心,其次用PCNN最短路径法确定目标函数极小值,最后通过改进的FCM分割目标。通过对比实验表明,该算法能够快速的分割出缺陷目标,正确率在95%以上。

For there is no better way to detect the defects of metal strip accurately, it presents a new algorithm based on im- proved pulse coupled neural network （PCNN） and Fuzzy C-means （FCM）. Firstly it can be obtained a clustering center through the effectiveness index. Secondly, through the PCNN shortest path algorithm, the objective function minimum values was deter- mined. Finally, it can be detected the defects of metal strips by improved FCM. By the comparison Experiments, it shows that this algorithm can rapidly detect the defects, the proposed method can increase segmentation accuracy rate to 95%