Study On Key Algorithms Used In Automatic Identification System For Fabric Weave Pattern

With the development of Computer Aided Design technique of woven fabric, intelligentization of fabric imitative design and redesign is required urgently to be improved, in order to meet the new market needs of little batch and large types. The key problem in computer-aided weave identification for woven fabric is: how to automatically identify the weave patterns and parameters of the woven fabric sample by using digital image analysis. Some efforts have been performed on several key issues in this dissertation. The main contributions of this dissertation include:On the study of fabric image preprocessing, this thesis proposes an algorithm for skew detection and weft yarn density recognition based on Hough transform. Since the skew is inevitable during scanning fabric sample, the scanned image should be adjusted. To solve this problem, we provide a fast skew detection and correction algorithm for fabric image. Based on the weft direction information extracted from the image, hierarchical Hough transform is used to estimate the skew angle to get a satisfying precision. In terms of the previous result, a novel algorithm to identify the weft density is also introduced, which is not affected by the image skew. This algorithm pickes up the projection profile of the skew angle which represents the arrangement of weft yarns, and then calculates the weft density. The experimental results show the high accuracy of our algorithm.On the study of identifying weave patterns and parameters, this thesis proposes a method to semi-automatically identify double-layer weft woven fabric. The traditional image-based algorithms only aim at single-layer fabric and gray image. The strength of our algorithm is to identify double-layer fabric by color fabric image, which regards the double-layer weave pattern as a combination of several single-layer weaves. For the single-layer weave identification, a series of color image pre-processing techniques are adopted firstly, including color reduction and morphology operation. Then the weft and warp yarns are segmented respectively by analysis of their different distribution and arrangement features. The weave pattern is obtained from the correct yarn locations finally. Since the nonideal arrangements of weft and warp are considered especially, this algorithm is more practical than the traditional ones.On the study of weave pattern segmentation, this thesis firstly proposes a segmentation algorithm based on color and spatial information. In this algorithm, the...