Classification Processing Method of Cotton Foreign Fibers Based on Probability Statistics and BP Neural Network

Abstract

A classification processing method of cotton foreign fibers was proposed based on probability statistics and BP neural network. Due to the origin was wide, the type was complex and the characteristic of cotton foreign fibers was different, it was difficult to build a model on classification and identification of cotton foreign fibers in the detection process of cotton spinning enterprises. This method could solve this question elegantly. Firstly, obtained the sample data by extracting the mean value of R, G, B in the fiber image and built a model about BP neural network. Then, classified 2-types cotton foreign fibers by calculating absolute value and variance of the feature vector based on probability statistics. Finally, processed the extraction features according to the different types image. The cross-validation experiment, the results show that the method combining the probability statistics and BP neural network can classify the cotton foreign fibers efficiently, and the effect is better when the types of cotton foreign fibers corresponding to the different features extraction methods The cross validation experiment, results showed that the combination can effectively identify the classification of foreign fibers and BP network based on probability and statistics, and different types of contton foreign fiber used different feature extraction methods, the effect is remarkable.