摘要
摩托车轮毂表面缺陷可能会造成摩托车在行驶过程中发生交通事故,危及驾驶者的生命,因此,对于摩托车轮毂的表面缺陷检测就显得尤为重要。针对摩托车轮毂表面缺陷检测,提出使用支持向量机(SVM)的方法,在特征选择上使用不同的灰度特征和纹理特征相结合。剖析了样本数量的多少与不同特征维度对支持向量机分类结果造成的影响。当提取灰度特征和纹理特征作为特征变量的时候相较于单独的灰度特征分类精度具有明显的提高,当样本数量为100时,识别精度可以高达93%。研究结果表明,在灰度特征的基础上加入纹理的特征可以大幅度提高支持向量机在摩托车轮毂表面缺陷检测的识别精度。
Motorcycle wheel surface defects may cause motorcycles in traffic accidents,endangering the driver’s life,thence,For the motorcycle wheel surface defects detection is particularly important. To detect the surface defects of motorcycle hubs,a support vector machine(SVM)method is proposed,which uses different gray and texture features in feature selection. The effects of the number of samples and the different feature dimensions on SVM classification results are analyzed. Compared with the single gray-scale feature classification accuracy,the gray-scale feature and the texture feature are obviously improved when the feature is extracted. When the sample size is 100,the recognition accuracy can reach as high as 93%. The results show that adding texture features based on the gray feature can improve the recognition accuracy of the support vector machine on the surface of the motorcycle wheel.
作者
李飞
唐亚健
苑玮琦
LI Fei;TANG Ya-jian;YUAN Wei-qi(School of Information Science and Engineering,Shenyang University of Technology,Liaoning Shenyang110870,China)
出处
《机械设计与制造》
北大核心
2020年第8期296-299,共4页
Machinery Design & Manufacture
基金
国家自然科学基金资助项目(61271365)
辽宁省教育厅科学研究一般项目(LQGD2017029)。
关键词
摩托车轮毂
缺陷检测
机器视觉
支持向量机
Motorcycle Wheel
Defects Detection
Machine Vision
Support Vector Machine