The Research On Feature Recognition Of Raw Cotton Defects And Impurities Based On Image Processing Technology

Raw cotton defects and impurities have a major influence on the quality of cotton textiles,they are easy to form yarn, dyeing and fabric defects in the process of textiles, then they makethe textile dye uneven or form holes and reduce the economic benefits of cotton textileenterprises, so it is very important to purge these defects and impurities before spinning.Nowadays, we often sort these defects and impurities rely on workers in China, this method isinefficient and wasteful, and is easy to be affected by human factors. Based on this reason,this paper proposes a view which about using the computer image processing technology todetect these raw cotton defects and impurities.The main research content of this paper is the computer detection and feature recognitionabout raw cotton defects and impurities, the paper studies and analysis the method ofcomputer image processing technology, and complete the task of detection and identificationabout these raw cotton defects and impurities by testing and debugging many detectionalgorithms.First, this paper discusses the background and significance of the research subject, andthen describes each process and calculating method of image processing. In the part of imageacquisition, the paper puts forward a lighting system design plan about multi-angle andmulti-light source. Then the paper puts forward a kind of adaptive threshold imagesegmentation algorithm which based on Mean_Shift by experimental measurement andanalysis; and in the part of profile extracting about flaw and impurity, it introduces a scanningmethod based on morphological, and make the break point in the contour line link. Last, forthe description of characteristic value, this paper extracts four kinds of shape features ofimages, and analysis their correlation, and ultimately determines their matching method, andfinish the task of detection and identification about these raw cotton defects and impurities.Compare the final test results of this study with real object, the coincidence rate canreach90%, proved in this paper, we prove that the image processing method which is taken inthis paper is feasible.