Research On The Mosaic Algorithm For Fiber Images

Exports of textiles and clothing have always occupied an important position in the export trade in China. Cotton, wool and cashmere is most commonly used in the textile industry. Not only in China cotton production, imports, consumption, and exports of textiles is among the first countries in the world, China is also a major producer of wool and cashmere. Fiber length as well as blending ratio and fiber fineness, is important indicator of textile production. Quality of import and export products is judged and graded according to them. Detection result directly impacts on enterprise and national interests.Parameters such as fiber length, fineness and maturity reflect fibers and their yarn forming quality, which directly affect fiber's spin ability and process validation. They must be tested in quality inspection of textile material, worsted process.We developed a computer image fiber detection system of detection to improve the automation of textile fiber examination, which gives direct, accurate and fast measurement of cotton fiber length, fineness, maturity, and other indicators. The system obtains all the fiber indicators by doing several processes such as image fusion, image mosaic, fiber thinning, fiber separation and so on.In order to get fiber characteristic parameters of vertical, we place the fibers cut into short pieces on a glass slide, and use a digital camera connected with a optical microscope which has a magnification of 100～200. Limitations of microscope objective numerical aperture and microscope view, make it is impossible for targeted fiber cross section to be collected as a whole in only one frame. Therefore, we need to capture some local images, and then pose a panoramic image by mosaic.The main content of this article is to determine the concatenation template and stitching area based on adjacent relation of the images, and to propose an algorithm for high-accuracy, which can reduce the interference caused by noise and gray value difference to ensure precise positioning and correct stitching between different images.Considering the particularities of fiber images contrast to natural images, a fiber image mosaic algorithm is proposed for panoramic display of fiber images. Firstly, when people consider two images on align; they see cross-regional area is the smallest after overlap. We propose a mathematic to measurement accuracy of matches between the two images---alignment AM. Registration robustness and precision is improved because blocks are suitable for optimization algorithms compared with fibers that are slender with smooth appearance. Thus, our image mosaic is implemented with block objectives that are obtained with fibers of different obliquities.Image mosaic based on image feature is divided into two processes:feature extraction and feature matching. We choose the polygon boundary information as features of block objectives. Since the fibers are used as the dividing lines of blocks, the contour of the fibers naturally become important. We extract characteristic parameters of overlap region easily by tracking and segmenting the contour. Then, linear dependent and iterative arithmetic are respectively introduced to realize match process. Iterative matching carried out with polygon approximation is more efficient.Images fusion is proposed to eliminate the seam in mosaic. Anglicizing both fusion algorithm based on weighted average and fusion algorithm based on wavelet transform, we choose the better one to realize seamless splicing. Experimental results demonstrate the proposed algorithm, which achieve a satisfactory splice results in the majority of cases, can effectively reduce the matching complex and improve the stability of matches.