Study On The Detecting System Of Cotton Web Quality Based On Computer Vision

Cotton web quality is a comprehensive indicator to reflect its structure condition. At the same time, it largely influences the quality indicators of the subsequent product, such as the yarn evenness, the content of neps, the breaking tenacity, etc. So far, major national textile enterprise assess the quality of cotton web by visual assessment, so as to track down the carding machine condition and the rationality of its process parameter in time. But this traditional assessment methodology has some shortcomings, which include less sample amount, insufficient representative, longer sampling interval time and low efficiency. Applying online detection system can detect the full sample in real time, and efficiently control the quality of products. This method could increase the accuracy and improve the carding efficiency. With the introduction of the advanced computer graphical and visual technology, this approach might be developed as a potential on-line method to detect the cotton web quality based on computer vision in textiles.. Currently, only Germanic TRUTZSCHLER's TC-NCT device can achieve the cotton web on-line detection, but the equipment is expensive and the high cost of repair and maintenance is other important factor to restrict its popularization and application. So it is pressing to develop innovative the detecting system of cotton web quality with freedom knowledge property right and excellent performance. and this detection equipment have important social and economic meaning.A system based on computer vision was built to detect the cotton web quality in this paper. Firstly, the main design, constitution, and principle are described about the system in detail. This system using high-performance linear scanning CCD camera and LED backlight mode to capture web images on-line. The next step is image preprocessing. After analyzing illumination compensation methods, image enhancement methods and de-noising algorithms, the web images are preprocessed adopted homomorphic filtering, transformation of histogram and wavelet. Nep and trash in carded web are usually difficult to be detected as they often superimpose on dense fiber layer. In order to solve this problem, one adaptive threshold segmentation method was proposed for obtaining the information about nep and trash from card web image, based on Otsu threshold method and linear regression model. This paper makes a elementary discussion about the detection of cotton web evenness and the web grade estimation. The computation and statistic results show that CV (%) value can be truly reflecting the cotton web evenness, so the vertical, transverse and the overall CV values are chosen as its representative indicators. If patchiness or broken hole was appeared in the web, it is classified as the lowest grade. Finally, with the help of graphical user interface (GUI) provided by the MATLAB, the design of the detecting system of cotton web quality is achieved. In addition to the visual presentation of test results is directly displayed to the user, the carding machine condition is presented respectively.The experimental results and data analysis show that this detecting system of cotton web quality is effective and feasible. It can more accurately detect the neps, trash, patchiness and broken hole in carding web. On the other hand, according to the test results, the carding machine condition can be presented respectively. The requirements of real-time is taken into account when designing algorithm, which provide a good basis for further studies of the real-time detecting and control system of cotton web.