Research Of Feature Selection For Cotton Foreign Fiber Objects Based On Dynamic Cotton Flow

Related to important resources to the people’s livelihood, cotton plays a decisive role inthe national product and economic. Foreign fiber has serious impact on textile products eventhe quantity is very small. Once the fiber is mixed into cotton fiber, it will not only affect thetextile spinning capacity, but also form color spots in dyed fabric which may seriously affectthe appearance of the fabric and make a great damage to the cotton textile industry.For finding the excellent feature set, lots of features are extracted from the cotton foreignfiber objects, which include color, texture, shape features. Automated visual inspection, imageprocessing, feature extracting, and pattern recognition are main techniques used in thisresearch. The main research works are as follows:(1) Reaserch of a fast feature selection for cotton foreign fiber objects based on binaryparticle swarm optimization. It is for the current feature selection of cotton foreign fiberhaving more iteration times and slow speed. This method uses the binary particle swarmoptimization algorithm as a feature selection algorithm. It can effective balance the searchingfor local optimal solution and global optimal solution by using dynamic parameter to solvethe problem and introducting the intertial weight coefficient and the contraction factor tocontrol the velocity of the particle. And the support vector machine algorithm was used toverify the optimal feature set. Experimental results shows that the running time can reduce1/4,in the condition that the classification accuracy is almost with other algorithms.(2) Design and realization of cotton foreign fiber detecting system based on the dynamiccotton flow, in order to accord with actual production. The system includes seven parts of thelint opener, carding fan, lose cotton pipelines, the image acquisition device, cotton collectionboxes, electrical cabinets and industrial computer. Powered by carding fan can be improve theefficiency of the cotton foreign fiber detection. Lint opener with rack rollers can avoiddamage the cotton fiber. The foreign fibers can be exposure in the cotton surface which canadvantageous to detective, identife and measure of foreign cotton fibers. Experimental resultsshows that cotton foreign fiber detecting system based on the dynamic cotton flow can meetthe demand for cotton dynamic testing.