基于频域滤波的织物疵点检测  

Fabric defects detection based on frequency domain filtering

作　　者：管声启[1] 石秀华[1] 

机构地区：[1]西北工业大学航海学院,西安710072

出　　处：《计算机应用》2008年第10期2673-2675,共3页journal of Computer Applications

基　　金：西安市科技攻关资助项目(GG04039)

摘　　要：根据织物纹理和疵点的频谱不同分布,提出了织物疵点检测的新方法。首先在研究织物纹理分类的基础上,通过傅立叶变换获得频谱图像;然后设计频域滤波器抑制正常纹理频谱信息,通过重构灰度图像,把灰度图像分割成若干子窗口;最后将统计子窗口方差作为特征值与设定的阈值比较,从而判别是否存在疵点。实验结果表明该方法是有效的,检测正确率达到90%以上。According to the spectrum of fabric texture and defects, a new method for defect detection was presented. Firstly, on the basis of the fabric texture classification, a spectrum image was derived from the Fast Fourier Transform （ FFT）. Then, the frequency filter was designed to remove normal texture information, and the acquired reconstruction images were segmented into many sub-windows, in which variances were calculated as characteristics. Lastly, the characteristics were compared with the standard threshold to determine whether there were defects, Experimental results show that this method is valid, and the detection average accuracy is over 90%.

关 键 词：周期性性纹理 方向性纹理 频域滤波 疵点检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]