基于局部哈达玛变换的织物疵点检测  

Fabric Defect Detection Based on Local Hadamard Transform

作　　者：敦宏丽 袁晔[1] 王帅军[1] 

机构地区：[1]北京服装学院信息工程学院,北京100029

出　　处：《北京服装学院学报（自然科学版）》2016年第3期45-51,共7页Journal of Beijing Institute of Fashion Technology：Natural Science Edition

基　　金：2015年北京服装学院教育教学改革立项项目(ZDJG-1506)

摘　　要：提出一种局部哈达玛变换和互相关相结合的织物检测疵点方法.首先,选取正常织物作为模板并对模板和待检测织物图像进行直方图均衡化增强.在此基础上,对待检测织物图像进行等尺寸分割,分割得到的子图像大小与选定的模板相同.然后,分别对织物模板和待检测子图像做二维哈达玛变换.最后,分别计算模板图像与待检测子图像哈达玛变换的互相关,并选取合适的阈值作为疵点判别的依据.仿真试验表明,本方法对断经、跳花、勾丝、缺经等常见织物疵点具有良好的检测效果.A new method to detect fabric defects based on the cross-correlation of local hadamard transform （LHT） was proposed in this paper. Firstly, a normal fabric was designated as the tem- plate, followed by the histogram equalization enhancement of the template image and the fabric im- age to be detected. The fabric image to be detected is subsequently segmented into a number of sub- images whose sizes are similar to that of tl,e template image. The local Hadamard transform is then applied to the template image and the fabric image, separately. Finally, the cross-correlation value between the local Hadamard transform results of the template image and the fabric image was compu- ted. An appropriate threshold to the cross-correlation is determined by which the fabric defect detec- tion is achieved. Simulation results show that the present method is valid in the case of broken warp, skips, snagging and missing warp, etc.

关 键 词：局部哈达玛变换 疵点检测 互相关 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]