摘要
机械零件形状识别是机械视觉领域的一项重要研究内容。通过傅里叶描绘子,抽取机械零件形状的轮廓信息,形成一系列的高维矢量集合;利用核主分量分析(KPCA)进行特征矢量降维;再利用支持向量机(SVM)对一些简单的机械零件(垫圈、螺母和螺栓)进行了分类实验。实验表明,对于不重叠且形状完整的机械零件具有非常高的识别率,为机械零件智能分拣、智能装配等任务的设计提供了重要的参考依据。
Shape recognition of mechanical parts is regarded as an important area in the reseach work of intelligent machine vision. The contour information of shapes of mechanical parts is extracted by Fourier descriptors, and a series of high dimensional vectors are created- These vectors are processed through the method of vector dimension reduction based on KPCA. The method of pattern recognition based on SVM is used with these vectors to classify the some simple mechanical parts ( rings, nuts and bolts) . Experiments show the methods above have very high recognition rate for non-overlapping and full shape mechanical parts. The research results Provide the imoortant reference value for mechanical intelliaent sortina, assemblv and other tasks.
出处
《机械制造与自动化》
2016年第4期132-134,139,共4页
Machine Building & Automation
基金
中央高校基本科研业务费专项资金资助项目(DC120101013)