关键词: 图像处理, 聚酯纤维识别, 纤维分离, BP神经网络, 智能化测定

Abstract:

The most effective method to determine polytrimehylene terephthalate (PTT) fiber and polybutylene terephthalate (PBT) fiber blending ratio, the melted micro projection method is difficult to achieve intelligent identification and automatically count number of fibers. To solve this technical problem, the microscopic image of PBT fiber and PTT fiber blended was preprocessed by image processing technology to obtain the ideal thresholding image, and PBT and PTT fibers was intelligently separated and automatically counted; then the three eigenvalues of fitness, chromaticity and fiber straightness of each fiber was extracted, and three eigenvalues matrix of 200 fibers were used as learning samples to establish a BP neural network, and twenty fiber were used to verify that. The results show that the recognition rate of PTT fiber and PBT fiber are higher than 99%. The mixed fibers are intelligently identified by this method successfully. The images of 1 000 blended fibers are processed in the same way. The blending ratio of PTT fiber and PBT fiber is calculated by the melted micro projection method to be 38% and 62% or 32% and 68%. Compared with the corresponding actual blending ratio (40% and 60% or 30% and 70%), the error was within ±3%.

Key words: image peocessing, polyester fiber identification, fiber separation, BP neural network, intelligent determination