文献类型：期刊文献

中文题名：基于CNN的晶体生长状态图像重建与仿真研究

英文题名：Research on Image Reconstruction Algorithm and Simulation of Crystal Growth State Based on CNN

作者：王坤[1];李钰[1];毛颖杰[1];殷姸[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2019

卷号：36

期号：4

起止页码：300

中文期刊名：计算机仿真

外文期刊名：Computer Simulation

收录：CSTPCD;;北大核心:【北大核心2017】；

语种：中文

中文关键词：熔体状态;卷积神经网络;图像重建;多通道输入;祛伪影

外文关键词：Melt state;Convolutional neural network;Image reconstruction;Multi-channel input;Artifacts reduction

摘要：针对晶体生长过程中熔体状态图像对比度低、纹理细节不清晰的问题,在分析基于卷积神经网络的图像超分辨率重建算法的基础上,结合多通道输入以及去噪技术,提出一种基于多通道输入-祛伪影卷积神经网络的图像重建算法,用于提高熔体状态成像质量。以蓝宝石晶体为研究对象的重建仿真结果表明,所提算法的重建结果在主观判断和客观评价上都优于传统重建算法,对提高晶体生长设备的自动化水平具有十分重要的意义。
In view of low contrast and unclear texture details problem of the melt state image during crystal growth,with analysis on super-resolution reconstruction algorithm of convolutional neural network,and combined with multi-channel input and denoising technique,an image reconstruction algorithm of multi-channel and artifacts-reduction based on convolutional neural network is proposed to improve the quality of melt state image.The reconstruction simulation results of sapphire crystal show that the reconstruction result of the proposed algorithm is superior to the traditional reconstruction algorithm in subjective judgment and objective evaluation,which is of great significance for improving the automation level of the crystal growth equipment.