摘要
提出了一种新的基于无监督聚类木材缺陷识别方法。该方法借助于木材图像颜色矩特征提取,有效实现数据降维,利用K-means算法对产生的特征数据集进行聚类,产生不同木材表面类型类别标签,自动找出并正确标识木材表面缺陷位置。分析了不同木材表面缺陷类型识别效率。仿真实验结果表明,该方法能快速有效地进行木材表面缺陷自动识别,平均运行时间为0.442 s,平均识别查准率约为86.5%,平均识别查全率约为81.1%。
The paper addresses the non-supervised clustering method for wood defect inspection . To reduce the dimension of the image data, the method of color moments is used to extract the texture features from the color wood images. Based on the generated characteristics of datasets, K-means clustering algorithm is used to produce different types of labels and to locate the wood surface defects automatically. The practical usefulness and high scalability of the proposed method are demonstrated through extensive simulation studies. The experimental results show that this method can be carried out automatically and efficiently, and can receive the average runtime of 0. 442 s as well. The averaged accuracy is about 86.5% and the averaged recall ratio can reach 81.1%.
出处
《江南大学学报(自然科学版)》
CAS
2009年第5期520-524,共5页
Joural of Jiangnan University (Natural Science Edition)
基金
国家自然科学基金项目(30671639)
江苏省自然科学基金项目(BK2005134
BK2009393)
南京林业大学创新基金项目(163070036)