摘要
介绍了基于机器视觉的坯布疵点实时检测系统软件部分各主要环节的具体实现。该系统运用修正的自适应邻域平均法增强坯布疵点图像的质量,用自动阈值法将疵点从背景中分割出来,并利用去除杂点法对分割后的二值图像进行滤波。以提取出的周长、面积、不变矩等为特征,运用BP神经网络分类器对常见的九种疵点进行了分类,识别正确率达到了95.7%。
An intelligent detection system of the loom stare flaw based on machine vision was introduced and the software realization of the main parts in the system was given. The self-adaptive enhancement techniques were used for improving the images of the loom stare flaw sample, the flaw image can be separated from the background by using the auto-threshold method, and the method of acnode-removed for two-value image preprocessing was applied also, The area, perimeter, invariant quadrature and ect, of a flaw were taken as its characteristics to identify the nine types of pests by using a classifier based on BP NN. The correct identifications of this classifier reached 95.7%.
出处
《计算机应用研究》
CSCD
北大核心
2007年第5期190-191,共2页
Application Research of Computers
基金
中国科学院模式识别国家重点实验室开放基金资助项目(NLPR2003)
河南省自然科学基金资助项目(200510078009)
关键词
坯布疵点
图像处理
特征提取
识别
loom stare flaw
image processing
character extracting
recognition
