In this paper, we selected seven representative sampling points in Qingdao to collect atmospheric particulate samples in different periods, and then obtained the SEM (Scanning Electron Microscope) images of particulate matters through scanning electron microscopy. By observing and analyzing the morphological characteristics of the images, we identified and selected 334 SEM images of particles with distinct morphological characteristics. Based on the obtained SEM images of atmospheric particles, the particles are divided into seven categories according to their morphological characteristics: chain particles, flocculent particles, fibrous particles, spherical particles, quasi spherical particles, irregular mineral particles, and regular mineral particles. These obtained SEM images are further enhanced using corresponding image processing methods such as horizontal inversion, color balance, brightness transformation, fuzzy processing, etc. New image samples are generated and added to the dataset, presenting a dataset of SEM images for seven types of atmospheric particles, totaling 2,672 SEM images in total. The dataset can be used for identifying, classifying, segmenting research on atmospheric particles using machine learning and other related methods. Moreover, it can provide fundamental data for other related research on atmospheric particles.
Read full abstract