The Research On The Evaluation Of Pilling Grade Base On Sequential Projected Image

Fabric pilling refers to the surface flaw with fluff tangled caused by abrasion, which affects the appearance of garment and consumer choice. In recent years, consumer pays more attention to pilling problem of fabric, especially of thin and high tex fabric. Pilling grade of fabric is an important index to assess fabric quality. The common method that people apply for is to compare the sample with the standard photos. This method belongs to subjective evaluation with results affected by many man-made factors. This paper aims to study the objective evaluation of fabric pilling grade based on sequential projected images to attempt to replace subjective evaluation with objective method. The main contents includes experimental verification of theoretical model of the moving pilling based on sequential projected images. the detection of fabric thickness, the classification of fabric thickness into thin and light fabric, medium thick fabric, and heavy fabric, image segmentation of fabric of different thickness. and the establishment of objective grading standards using the nearest neighbor algorithm. This objective method is verified by experiments with15fabric sample. The main contents are described as following:1、 Supposing a small ball is taken as one pill, a sequential projected images of the ball are obtained through an image collection device. The profiles of these projected images are detected by edge detection and joined together to obtain image of fabric pilling indirectly. The pilling image is pretreated, segmented. Then features are extracted. The experiments results show the correctness and reasonableness of the moving pilling theory model based on the sequential projected images.2、 Considering the thickness of fabric, an optimal threshold is selected to segment the sequential projected images based on double gaussian fitting with different segmentation parameter for fabrics of different thickness. The contour lines of projected image are joined into pilling image that is segmented by single-peak gaussian threshold. On image of sever pilling fabric, two or more pills are likely adhered together. The pilling area of adhesive pills is segmented again by a method based on point neighborhood ratio to get the threshold to separate the adhesive pills3、the number of pilling, the total area of pilling. the average area of pilling,the maximal area of pilling, the total volume of pilling, the average volume of pilling, the maximal volume of pilling, the average height of pilling, the maximal height of pilling are extracted. The relationship between these nine pilling features and pilling grade is analyzed and discussed in detail. Four pilling features (the number of pilling, the total area of pilling, the average volume of pilling and the maximal height of pilling) are selected to be normalized to give a basis for evaluating pilling degree.4、 Nearest neighbor algorithm is chosen to establish the standard between the four pilling features and pilling grade. Then three sets of images of actual pilling fabrics including15samples are evaluated. The objective result is compared with the subjective evaluation with consistent rate of86.7%, indicating the feasibility of objective evaluation base on sequential projected images.