Online Recognition of Foreign Fibers in Lint

Online Recognition of Foreign Fibers in Lint ĦĦĦĦAimed at online recognition of foreign fibers in lint, an approach for recognition of foreign fibers in lint based on automated visual inspection was proposed. To construct an efficient image acquisition system for detection and classification of foreign fibers in lint using automated visual inspection, an optimal band selection method for detecting foreign fibers in lint using spectral analysis was proposed. A fast image processing method was developed to segment the foreign fiber objects out of the background. The captured image was firstly segmented according to the mean and standard deviation of R, G and B value of each pixel in the image. Then noises were removed using area threshold method. To obtain features to describe a foreign fiber object accurately, an improved genetic algorithm was developed to select the optimal feature subset effectively and efficiently. To classify the foreign fiber objects fast and accurately, a novel classification method based on multi-class support vector machine (MSVM) was proposed. To test the abovementioned methods, a prototype of automated visual inspection system for detection of foreign fibers in lint was designed and implemented. The experimental results indicate that the ultraviolet band is the optimal band to detect the fluorescent foreign fibers, the visible light band is the optimal one to detect color foreign fibers, and the infrared band is fit for inspecting those foreign fibers such as polypropylene and polyethylene materials, candy wrappers, hairs and feathers. The feature selection results show that IGA gets improved searching ability and accelerated convergence speed compared with SGA. The optimal feature subset selected by IGA has smaller size than that selected by SGA. The classification results indicate that one-against-one diagram-based MSVM is the fastest and is fitter for online classification.