Multiscale Optical PM2.5 Particles Recognition and Sorting System in Dust Probes
Andrey N. Kokoulin, Aleksandr A. Yuzhakov, Andrey N. Kokoulin · 2020 5th International Conference on Smart and Sustainable Technologies (SpliTech) · 2020
Authors propose the novel approach for optical PM2.5 and PM10 particles recognition and sorting based on Super-resolution neural networks in the research on dust emissions from industrial enterprises. The objective of the dust emissions analysis is to determine their component composition and the fine particle size distribution (PM10 and PM2.5). The scanning electronic microscope of high resolution was used to obtain the set of large-scale images of dust particles. We use the images of particles in different scales as the entire imagery data to create the high quality image suitable for reliable particles recognition and use quality metrics PSNR, MSE, SSIM to ensure that the created image is close to ground truth.