织物疵点检测与识别的方法研究

Method Research On Inspection and Recognition of Defects in Fabric

将织物图像分成大小相同的局部窗口,选取局部熵最小的窗口为待研究的感兴趣区域,在此区域内分割出疵点图像并用数学形态学中的开运算滤除噪声,计算疵点形状因子等作为识别参数,研究表明此方法因能避免对整幅图像进行复杂运算,在提取特征参数时对图像的全局搜索,具有识别正确率高、检测速度快等优点.

By dividing fabric image into many windows with the same size, choosing a window with minimum local entropy as a region of interest, thresholding the region and filtering out noise through opening function of mathematical morphology, calculating the defect shape factors as recognition parameters, a algorithm and method of detection of defects in fabric is studied. Which has advantages of high identification, correctness, and fast inspection speed mainly because it can avoid complicated calculation of whole image and global searching when the feature parameters are extracted.