Established health hazards posed by dust microparticles require automated and mobile devices for their assessment. Such devices should provide an opportunity to analyze component and disperse structures of the solid component in ambient air pollution operatively and in real time. In future, they will replace labor-consuming sampling and separate identification of fraction structure and chemical composition of dusts. The aim of this study was to develop and test new methodical, procedural and instrumental approaches to monitoring of solid particles in ambient air. We suggest a hardware and software complex that implements a two-stage scheme for identifying solid particles sampled in ambient air according to the from-coarse-to-fine principle. The first stage involves identifying the total concentration of solid particles by laser diffraction. Microphotographs are taken with iMicro Q2 mini portable microscope with magnification x800. The microscope lens is connected to a camera, which is linked to nVidia Jetson Nano micro PC. The micro PC classifies particles, identifies their contours by using a neural network and deals with image segmentation. The second stage relies on using computer vision that makes it possible to automate routine recognition of particle images created by the microscope in order to calculate levels of different substances in a sample. All the data are analyzed by the second neural network that performs preset calculations in accordance with mathematical logic (model). The network is trained using a library that contains attributed microphotographs of dusts with different qualitative and disperse structures. The algorithm has been tested with some promising results. Identified disperse structures and chemical composition of dusts turn out to be quite similar to those identified by conventional approaches and measurement methods. The method has been shown to offer wide opportunities to identify dust composition and structure, to create dust pollution profiles, and to estimate a contribution made by a specific source to overall pollution. The study results ensure more correct and precise health risk assessment under exposure to dusts in ambient air.
Read full abstract