Research for the measurement and treatment of yarn blend ratio

The method for measuring the yarn blend ratio with image treatment is studied. which are divided into two major parts. The former part mainly introduces the whole process of pretreatment of the original image, which is dotted with noises and other mayuseless things. The techniques of image enhancement and shape filtering have been applied. Good result is achieved with proper algorithms. The latter part emphasis on the choice of characteristic parameters that determine the fiber cross-sectional character from the treated image. Conventional parameters, which are used to discriminate the fiber characters, have been extracted from the treated image. After studying and comparing the traditional methods of pattern recognizing, a new recognizing measure is presented. Artificial neural network (ANN), with capabilities of tolerance and learning, can be used to recognizing and classifying the target parameters. Many types of network mode have been experimented. It was proved that right type of network could be in good performance in identifying the target signals, and then sorts the different fibers. Experiment shows that using ANN can improve the efficiency of recognizing process.