王杉 朱亚涛 胡建辉



摘 要: 针对我国目前气象观测站点分散不均匀导致实时得到的空气质量信息不准确的问题,文中提出一种基于光学成像的空气质量定期性制定方法。该文结合通过数字图像去雾领域中基于光学成像原理去雾得到的大气透射率和通过气象观测领域中得到的空气质量与散射系数间的相关性,并考量同一景深下大气透射率和散射系数之间的指数关系,利用BP神经网络模型学习空气质量和图片透射率间的隐性联系,做到通过图片来定性判定其空气质量。实验结果显示,所提空气质量判定方法的整体识别率為83.72%,优类和良类识别率达到了90%以上。
关键词: 空气质量判定; 定性判定; 光学成像; 图像去雾; 实验设计; 实验结果分析
中图分类号: TN911.73?34; TP391 文献标识码: A 文章编号: 1004?373X(2020)08?0113?04
A methodof air quality qualitative judgement based on optical imaging
WANG Shan, ZHU Yatao, HU Jianhui
(College of Information Engineering, East China Jiaotong University, Nanchang 330022, China)
Abstract: As the real?time information of air quality is inaccurate due to the uneven distribution of meteorological observation stations in China, a method of air quality qualitative judgement method based on optical imaging is proposed. The atmospheric transmittance obtained by defogging based on optical imaging principle in the digital image defogging domain, and the correlation between air quality and scattering coefficient obtained by meteorological observation are combined in this paper. And also, the exponential relationship between the atmospheric transmittance and the scattering coefficient at the same depth of field is considered, and the implicit relation between air quality and image transmittance is studied by means of BP neural network model. On the bsasis of the above method, the air quality is judged qualitatively by pictures. The experimental results show that the overall recognition rate of the proposed air quality judgement method is 83.72%, in which the recognition rate of superior and good classes is over 90%.
Keywords: air quality judgement; qualitative judgement; optical imaging; image defogging; experiment design; experiment result analysis
0 引 言
雾霾是当今一个越来越严重的环境问题。学者们不仅研究它的成因并解决其造成的问题,如影响各识别系统等,而且还研究自行观测的数据分析不同空气质量情况下各因素的变化情况[1?2],大众则关心外面的雾霾污染程度是否适宜外出活动。
雾或雾霾对数字图像的影响是一样的,而图像去雾的研究已经有几十年的发展。文献[3]里系统介绍了图像去雾的研究进展。数字图像去雾,整体来讲主要有两个方面:一是基于图像增强去雾[4];二是基于大气散射模型去雾[5]。第二种目前已经成为主流的去雾方法,其思路是研究光学成像[6]的原理,将大气粒子对光线的散射应用到雾天成像中,故此方法是专门针对图像中的雾提出的解决方案,尤其是He等基于此模型提出的暗原色先验理论,使得图像去雾的研究达到了一个新高度[7]。……