基于BP神经网络的弹壳表面缺陷分类方法

Cartridge Case Surface Defect Classification Based on BP Neural Network

为了对弹壳表面缺陷进行分类,提出一种基于BP(back propagation)神经网络的弹壳表面缺陷分类方法。针对弹壳缺陷的特点,提取了各类缺陷的灰度特征、形状特征、几何特征,建立缺陷特征数据库,并采用改进的BP神经网络算法设计了缺陷分类器。实验结果表明,该方法在枪弹缺陷识别方面具有很好可行性和有效性。

In order to classify the cartridge case surface defects, a method based on BP neural network was proposed for classification of cartridge case defects. According to the defects characters, extracted the information of gray feature, shape feature and geometric features, established the characteristic database, a defect classifier was designed base on an improved BP neural network algorithm. The experiment results show the method is feasible and effective in bullet defect detect.