摘要
提出了一种广义加权支持向量机(GW SVM)的焊接缺陷分类算法。首先为克服由于样本数量不平衡性引起的小样本类别精度差的问题,引入由于样本差异的权重;然后为解决不同类别的重要性要求,根据经验人工确定不同类别重要性的权重。针对样本重要性的影响,采用有监督模糊聚类方法来确定样本重要性权重。测试结果表明:广义加权支持向量机在噪声影响较大及样本类别相差较大时,能够提高重要的、数量少的缺陷检测精度。
A classification algorithm of generalized weighted support vector machine （GWSVM） is put forward. In order to improve the class accuracy of small number of samples caused by the imbalance of sample number ,weight of sample difference is introduced. According to experience,the importance weights of different classes is manually determined to solve the importance of different classes. The method of fuzzy cluster with superision is used to determine the weights of sample importance. Testing results show that GWSVM can improve the inspection accuracy of a small number of important defects when the noise and sample sort have important effect.
出处
《华东理工大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2005年第5期644-648,共5页
Journal of East China University of Science and Technology
基金
江苏省博士后科研基金资助项目
