摘要
介绍一种计算机皮革缺陷检测与排样切割系统的构成原理与设计思想。该系统对皮革纹理图像采用小波包分解去噪,并用共生矩阵提取皮革图像的纹理特征向量,用改进模糊聚类方法对样本特征样本向量进行聚类分析。提出了一种基于类间方差和类内方差的自适应确定分割区域数的方法。该算法可以准确的检测出皮革表面缺陷信息,其在线缺陷检测结果与人工判断缺陷类别的一致性在89.1%以上,实现了对皮革缺陷的自动检测,优化排样。
The compositional principle and design concept of a computerized defects detecting nesting and cutting system for leather were introduced. The noise of leather texture images was wiped off by adopting wavelet packet decomposition, and the characteristic rectors of images were extracted using co-occurrence matrix, then the characteristic vectors were clustered by modified FCM. A valid function based on infra-class and inter-class variance to decide the number of regions was presented. The algorithm can exactly detect the defects of leather, and the consistence of the results of on-line detection and artificial detection is above 89.1%. The automatic detection and nesting were achieved.
出处
《皮革科学与工程》
CAS
2006年第4期24-26,56,共4页
Leather Science and Engineering
基金
陕西省自然科学基金资助项目(2004F44)
关键词
图像检测
皮革
排样
缺陷检测
共生矩阵
image detection
leather
nesting
defects detecting
co-occurrence matrix