棉纺织技术2024，Vol.52Issue(4)：59-64,6.
基于图像处理技术的纤维直径测量方法
Fiber diameter measurement method based on image processing technology
摘要
Abstract
In order to measure fiber diameter more quickly and accurately,a fiber diameter measurement method based on image processing technology was explored.Firstly,according to the characteristics that fibers were light in colour and small in area under the optical microscope,the attention mechanism module and the dilated convolution kernel pair model were introduced into the UNet + VGG16 model for optimization,more accurate segmentation of fiber images was achieved.Secondly,the edge lines and median axis of the segmented binary image were extracted using the orthogonal skeleton method containing Canny operator and median axis transform algorithm,then the fiber diameter was measured by using the linear fitting method.Finally,the developed image processing measurement method was verified by direct microscopic measurement.The results showed that the improved model converged faster,the pixel accuracy was 96.5%and the pixel accuracy was 0.14 μm/pixel.The relative errors between two samples fiber are 1.13%and 2.07%respectively,the maximum difference of the same batch of specimens for five measurements was 0.06 μm.It is considered that the method has the advantages of faster testing speed,higher automation and better stability.关键词
纤维直径/图像预处理/UNet+VGG16/空洞卷积/注意力机制Key words
fiber diameter/image processing/UNet+VGG16/dilated convolution/attention mechanism分类
计算机与自动化引用本文复制引用
常帅才,魏菊,林素存..基于图像处理技术的纤维直径测量方法[J].棉纺织技术,2024,52(4):59-64,6.基金项目
辽宁省教育厅基本科研项目(LJKZZ20220064) （LJKZZ20220064）
