Objective evaluation of fabric pilling based on multi-view stereo vision
Lulu Liu, Na Deng, Binjie Xin, Yiliang Wang, Wenzhen Wang, Yan He, Shuaigang Lu
Abstract
Fabric pilling evaluation is very important for the quality control of textile industry. Traditional image analysis based methods have the disadvantages of 2 D imaging and color sensitivity, this paper presents a new method based on stereo vision to solve the problem of 3 D imaging of fabric pilling. One set of self-developed mobile camera system is established to capture a group of images for the 3 D reconstruction of the fabric surface, the point cloud model of the fabric surface is generated by the self-developed stereo vision algorithm, including structure from motion (SFM) and patch-based multi-view stereo (PMVS) algorithm. One 2 D gray-scale image is obtained from the 3 D point cloud model by mapping to the 2 D image plane, which contains the depth information of fabric pilling. The segmentation of fabric pilling could be done by accurate positioning of edge detection, adaptive thresholding and morphological analysis. Four feature parameters including pilling number, pilling area, pilling density and coverage ratio are extracted for the determination of fabric pilling grade objectively. Experimental results show that the new developed system and method is effective and reliable for the fabric pilling evaluation, which is consistent with the subjective pilling evaluation. It is workable for the color printed or yarn dyed fabrics, the proposed imaging system could be a good solution for the digital intelligent quality control of textile products.