基于粒子光散射场成像的颗粒物特征传感方法研究

Air pollution has raised the widespread concern of all governments and publics around the world. The sensing of particulate matter in the atmosphere is priority for monitoring, tracing and reducing the particulate matter pollution. At present, many aerosol measurement instruments are often too large. Moreover, their sensitivity, service life and working ability could not fully meet the requirements of the sensing of particulate matter in the atmosphere. The optical sensing methods of particulate matter have certain advantages in this field. The open problem of the optical sensing methods is how to predict the particle size distribution, refractive index and other parameters of the particulate matter accurately and efficiently based on the optical signals measured directly or indirectly, in order to obtain the statistical characteristics of the particulate matter. However, the particle size distribution, refractive index and other high-dimensional parameters could not be predicted accurately only by the signals measured at a few observation angles because the light scattering signals of particulate matter are radiated toward all directions in space. In this project, the spatial signals model of light scattering field of particulate matter is obtained from a series of single particle light scattering fields weighted by the particle size distribution. A two-dimensional imaging method based on the semi-spherical optical projection of the spatial signals of light scattering field is proposed, which is a lossless transformation from the high-dimensional spatial information of light scattering field into two-dimensional image signal. Finally, the characteristic parameters of particulate matter are inverted from the light scattering field image based on deep learning, in order to realize the accurate sensing of the statistical characteristics of atmospheric particulate matter in a simple and efficient way.

大气污染问题已引起政府和民众的广泛关注，对大气颗粒物的传感，是对大气悬浮颗粒物污染进行监控、溯源和治理的前提。目前的气溶胶测量设备过于庞大，其灵敏度、使用寿命和工作能力都不能完全满足实际要求。采用光学原理的颗粒物传感方法具有一定优势，但存在的问题是，如何利用直接或间接测量得到的光学信号，准确高效地预测颗粒物的粒径分布函数、折射率等参数，并得出颗粒物的统计特性。颗粒物的光散射信号向空间各个方向辐射，包含粒径分布函数和折射率等高维数信息，仅靠有限观测角的测量信号难以准确求解。本项目建立单粒子的光散射场，由粒径分布函数加权获得粒子群的光散射场空间信号模型，提出基于半球面光学投影的颗粒物光散射场空间信号的二维成像方法，将光散射场的高维数空间信息转化成二维图像信号，利用深度学习方法由光散射场图像反演颗粒物的特征参数,以简单高效的方式实现大气颗粒物统计特性的精准传感。

本项目面向颗粒物精准传感需求，针对现有光学颗粒物特征传感中面临的如何利用直接或间接测量得到的光学信号，准确高效地预测颗粒物的粒径分布函数、折射率等参数的科学问题，提出了基于粒子光散射场成像的颗粒物特征参数传感方法。基于Mie散射理论构建了颗粒物的光散射场空间模型；通过设计光学构造与搭建光散射场成像光学系统，将粒子光散射场中半球面空间信号投影成二维图像，提出了基于半球面光学投影的颗粒物光散射场空间信号的二维成像方法；利用颗粒物光散射场空间信号的二维图像信息，采用深度学习方法获得了颗粒物的粒径分布函数和折射率等直接特征参数，实现了颗粒物的主要间接统计参数，包括数量浓度、表面积浓度、体积浓度的解析；进一步基于对数正态模型从粒子光散射场中仅提取少量关键光散射信号，实现了颗粒物粒径和浓度信息的准确传感。在本项目支持下，项目组在国际上发表高水平学术论文9篇（按照2022年中国科学院文献情报中心期刊分区表划分，一区2篇，二区5篇，三区2篇），国际会议报告4篇，授权与申请发明专利4项，1名博士研究生毕业，2名博士研究生即将毕业论文答辩，3名硕士研究生毕业。本项目提出并实现的传感方法为包括大气颗粒物、烟雾粒子在内的气溶胶特征精准传感提供了新的技术支撑。