基于离散傅里叶变换的织物纹理信息在线检测  

Fabric Texture Information Online Detection Based on DFT

作　　者：楼越升[1] 祝成炎[2] 郭振荣 郑鹏飞[4] 

机构地区：[1]义乌工商职业技术学院机电信息分院,浙江义乌322000 [2]浙江理工大学材料与纺织学院,浙江杭州310018 [3]浙江蓝宇数码科技有限公司,浙江义乌322000 [4]华东理工大学机械与动力工程学院,上海200237

出　　处：《东华大学学报（自然科学版）》2016年第5期732-737,744,共7页Journal of Donghua University(Natural Science)

基　　金：浙江省教育厅科研基金资助项目(Y201534823);义乌工商职业技术学院科研基金资助项目(2015026)

摘　　要：通过对离散傅里叶变换幅度谱中不同像素点位置所对应原图织物纹理特征的分析,提出根据幅度谱次亮点的位置来计算织物纹理密度及角度的次亮点分析方法.阐明纹理角度检测的精度由原图像包含的纹理数目所决定,并提出3种方法来提高检测精度:增大图像规格、增大拍摄物距、进行边缘检测;在此基础上,对3种织物进行了检测试验,进一步验证了提高精度的3种方法;最后,与投影分析法检测结果进行比较,证明次亮点分析方法更适合解决纹理角度的在线检测问题.By analyzing the texture characteristics in the original fabric image and the associated pixel position in the DFT（discrete Fourier transform）magnitude spectrum, the analytical method that the texture direction and density can be calculated according to the position of the second brightest pixel pair is proposed. The detection accuracy is determined by the number of textures included in the original image and three ways to improve detection accuracy are pointed, including enlarge the image size, increase the shooting distance, and conduct edge detection. Based on these, detection experiments are taken on three kinds of fabrics to verify the three ways in improving accuracy. Finally, comparing trial on projection analysis, it is proved that the second brightest pixel analysis is more appropriate for online detection of texture direction.

关 键 词：离散傅里叶变换 织物纹理 检测精度 次亮点分析 投影分析 

分 类 号：TP202[自动化与计算机技术—检测技术与自动化装置] TP29[自动化与计算机技术—控制科学与工程]