当前位置: X-MOL 学术J. Quant. Spectrosc. Radiat. Transf. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Characterization of particle size and shape by an IPI system through deep learning
Journal of Quantitative Spectroscopy and Radiative Transfer ( IF 2.3 ) Pub Date : 2021-03-15 , DOI: 10.1016/j.jqsrt.2021.107642
Hongxia Zhang , Zhonghao Li , Jinlu Sun , Yushi Fu , Dagong Jia , Tiegen Liu

Characterization of particle size and shape is important in studying particle field changes. Interferometric particle imaging is a reliable technique that is widely used in the characterization of droplets or bubbles. In recent years, this method has also been used to measure irregular rough particles. According to the distribution of emitting points, the projected 2D shape of particles can be obtained. In this work, we propose a convolutional neural network (CNN) to obtain the projection image of an irregular particle from its interference-defocused image through the relationship between the distributions of particle emitting points and speckles in the defocused image. Sand projections could be successfully predicted from the experimental defocused image by the trained CNN, and difference in caliper lengths between the actual and predicted particle projections was within 6%. Such a small variation proves the feasibility and accuracy of the proposed method.



中文翻译:

通过深度学习通过IPI系统表征粒径和形状

粒径和形状的表征对于研究粒子场的变化很重要。干涉粒子成像是一种可靠的技术,已广泛用于液滴或气泡的表征。近年来,这种方法也已用于测量不规则的粗糙颗粒。根据发射点的分布,可以获得投影的二维粒子形状。在这项工作中,我们提出了一个卷积神经网络(CNN),通过散焦图像中粒子发射点和散斑的分布之间的关系,从不规则粒子的干扰散焦图像中获得投影图像。训练有素的CNN可以根据实验散焦图像成功预测出沙粒投射,实际和预测的粒子投影之间的卡尺长度差异在6%以内。如此小的变化证明了该方法的可行性和准确性。

更新日期:2021-03-24
down
wechat
bug