摘要
烧结断面火焰图像中蕴含大量烧结工艺信息,通过分析图像特征信息的变化能够有效判断烧结终点位置,帮助钢铁企业节能减排。由于烧结生产处在浓烟尘、高热辐射的环境下,导致采集的火焰图像易出现边缘模糊、噪声退化等问题,影响图像特征信息的提取。为提升图像质量,针对多种图像退化因素,提出一种Real-ESRGAN网络与形态学多结构算子相结合的图像复原方法,利用Real-ESRGAN网络对图像进行去模糊化,增强图像的视觉效果,再使用4种不同方向的多结构算子并结合权重自适应算法,抑制图像噪声,达到提高烧结火焰图像质量的目的。实验结果表明,该方法对于图像复原具有显著效果:PSNR均值达34.3018 dB,SSIM均值达0.9260,为通过图像信息判断烧结终点位置奠定了基础。
The flame image of the sintered section contains a lot of sintering process information.By analyzing the change of image feature information,the sintering end point can be effectively judged,which helps steel enterprises save energy and reduce emissions.Due to the sintering production in the environment of dense smoke and high heat radiation,the collected flame images are prone to problems such as blurred edges and noise degradation,which affects the extraction of image feature information.In order to improve the image quality,an image restoration method combining Real-ESRGAN network and morphological multi-structure operator was proposed for a variety of image degradation factors.Real-ESRGAN network was used to deblur the image and enhance the visual effect of the image,and then four multi-structure operators in different directions were used in combination with the weight adaptive algorithm so as to suppress the image noise and improve the image quality of the sintered flame.Experimental results show that the method has a remarkable effect on image recovery,with an average PSNR of 34.3018 dB and an average SSIM of 0.9260,which lays a foundation for judging the sintering endpoint position through image information.
作者
安金铭
梁秀满
王雁
曹晓华
曾凯
王福斌
AN Jin-ming;LIANG Xiu-man;WANG Yan;CAO Xiao-hua;ZENG Kai;WANG Fu-bin(College of Electrical Engineering,North China University of Science and Technology,Tangshan Hebei 063210,China;College of Metallurgy and Energy,North China University of Science and Technology,Tangshan Hebei 063210,China)
出处
《华北理工大学学报(自然科学版)》
CAS
2023年第2期60-68,共9页
Journal of North China University of Science and Technology:Natural Science Edition
基金
河北省自然科学基金项目(E2021209037)
河北省高等学校基本科研业务费研究项目(JYG2020004)。
作者简介
第一作者:安金铭,女,硕士研究生。研究方向:图像处理;通讯作者:曾凯,男,博士研究生,讲师。研究方向:机器视觉。