摘要
提出一种基于聚类分析和支持向量机(SVM)的布匹瑕疵分类方法.该方法充分利用瑕疵的几何特征,首先使用迭代自组织数据分析技术算法(ISODATA)对其进行聚类,在聚类形成的子空间内再根据瑕疵的纹理特征利用SVM进行分类.根据布匹瑕疵的特点提出一种新的几何特征,并使用各类瑕疵的几何特征均值作为初始聚类中心,提高ISODATA算法的聚类效果.实验表明,该方法有效地提高了分类准确性,降低了训练的复杂度,分类准确率可达90%.
Presents an efficient method of fabric defect classification based on cluster analysis and support vector machine (SVM). The iterative self-organizing data analysis technique algorithm(ISODATA) is applied to cluster the defects,SVM is then used to classify each cluster. The paper presents a new geometric feature according to the characteristics of fabric defects, and use the mean geometric feature value of each defect class as the initial clustering center to improve the result of clustering. Experimental results show that the method improves the precision of classification effectively and reduces the complexity in training. The overall classification precision reaches 90%.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2004年第8期687-690,共4页
Transactions of Beijing Institute of Technology
基金
中国纺织品进出口总公司资助项目(HK0109-05)
关键词
瑕疵分类
聚类
支持向量机
特征提取
defect classification
clustering
support vector machine
feature extraction
