Feature Extraction On Cotton Foreign Fibers And Study On Method Of Measurement

The quality of cotton products depends heavily on the amount and types of foreign fibersin the raw matterial. Traditional assessment of cotton quality was performed by mannualidentification and weighing of foreign fibers. Limited by the ability of human eyes to detectthe subtle differences between contton and foreign fibers, this method became unsuitable asthe demand for highquality textile products grew. Though the identification and quantitationof impurities have been greatly facilitated by the application of computer-aidded device, areliable model for gauging the cotton quality loss caused by foreign fibers.Targeting seven typical foreign fibers (feathers, hemp, colored threads, plastic pieces,hair, polypropylene fiber and colored paper), this study intended to construct an apraisalmodel for the extent of cotton contamination by foreign fibers on the basis of computerizedimage processing and statistics. The study consisted of the following key steps:1. Image segregation: Machine-aquired images were segregated through threeconsecutive wavelet transforms, resulting in target images;2. Foreign fiber feature extraction, classification, and quantitation: Features of color,texture, and shape were extracted and eigenvectors generated; cohensive hierachicalclustering algorithm and exhaustive attack method were then used to determine the optimaleigenvectors. Classification and quantitation were performed using principal componentanalysis and BP neural network. An identification rate of at least95%was achieved for eachforeign fiber studied.3. Evaluation of foreign fiber contamination: Taking into account the types and numbersof occurance of foreign fibers, a cotton quality apraisal model was established. Cottonsamples were graded according to the amount of foreign fibers, rather than to the total weightof impurities. This model is superior as it reflects the varied quality-damaging consequencesof foreign fibers.Overall, this study set up the foundation for further development of more efficientprocesures of identifying and removal of cotton foreign fibers.