基于显著性检测和超像素分割的本色织物疵点检测系统

Detection System for Grey Fabric Defects Based on Saliency Detection and Superpixel Segmentation

针对本色织物生产过程中出现的断经、断纬、污渍、擦伤和破洞等表面缺陷,课题组设计了一种基于显著性检测和超像素分割的本色织物疵点检测系统。课题组首先对输入的图像进行双边滤波,保持图像边缘的同时去除织物纹理;然后将图像分成n×n个大小相同的图像块,对每个图像块使用基于全局对比度的图像显著性检测生成显著图;再对整张粗定位显著图进行超像素精细分割,以及二值化和图像形态学处理剔除孤立点,定位出疵点区域。实验结果表明:与3种常见的显著性检测算法相比,新系统对本色织物疵点检测的准确率更高,时间更短且疵点轮廓的分割更精确。

In view of the surface defects such as broken warp,broken weft,stains,scratches and holes in the production process of natural fabric,a detection system of natural fabric defects based on saliency detection and super pixel segmentation was designed.First,bilateral filtering was used for the input image to keep the image edge and remove the fabric texture;then,the image was divided into n×n image block with the same size based on the global contrast of image significant testing generates significant figure;again,the entire coarse positioning significant figure was eliminated isolated point positioning the defect area,by pixel fine segmentation,and binarization and morphological image processing.The experimental results show that compared with the three common saliency detection algorithms,new system has higher accuracy,shorter time and more accurate segmentation of the defect contour of this color cloth.