摘要
提出了一种将支持向量机(SVM)用于机械零件识别的方法。实验采用了97张零件图片,9类零件其中一部分作为训练样本,另一部分作为测试样本。提取零件的 Hu 矩作为特征向量,并将 BP神经网络与 SVM 进行了比较。实验结果表明,以多项式为核函数的 SVM 有较高的识别率。
A method for application of SVM in recognition of mechanical parts is presented in this paper. Hu moments of 97 samples of the mechanical parts in 9 types are taken as the feature vectors for their classification using SVM. Comparing with BP neural networks, the results indicate that SVM is of higher accuracy of pattern recognition.
出处
《电子技术应用》
北大核心
2008年第11期108-110,114,共4页
Application of Electronic Technique
关键词
支持向量机
零件识别
HU矩
BP神经网络
support vector machine
parts recognition
Hu moment
BP neutral network.
