摘要
针对玻璃瓶质量检测系统缺陷分类难的问题,选取气泡、结石、裂纹、污点、皱纹这五种常见的缺陷作为分类目标,从研究每种缺陷的图像特征入手,提出了七个统计特征作为分类器的输入特征向量,根据该分类问题的特点构建SVM分类器,采用现场采集的缺陷图像样本对SVM分类器进行训练和测试。实验结果表明:设计的SVM分类器识别率较高,适合玻璃瓶缺陷图像分类。
Aiming at the defects classification of glass bottle inspection system, five common defects（blister, stone, crack, dirt and crinkle）are chosen. Based on the image features of every defect, seven statistical features are put forward as input feature vector of classifier. According to the characteristic of such classification problem, SVM classifier is constructed, trained and test. The experimental result shows that the SVM classifier has high identification accuracy, and is suitable for the defect classification of glass bottle image.
出处
《机电产品开发与创新》
2015年第1期23-25,共3页
Development & Innovation of Machinery & Electrical Products
