PMConcentration Measurement Based on Natural Scene Statistics and Progressive Learning

Published: 01 Jan 2023, Last Modified: 13 Aug 2024IFTC (2) 2023EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: To ensure the well-being of residents and ecological balance, it is essential to conduct continuous and precise PM\(_{2.5}\) concentration monitoring. To this end, we have developed a visual-based PM\(_{2.5}\) concentration estimation algorithm consisting of two models: a visual feature perception model based on natural scene statistics and a progressive learning-based PM\(_{2.5}\) concentration prediction model. The proposed visual feature perception model, drawing from natural scene statistics, pinpoints visual characteristics in the structure and saturation domains, accurately quantifying the loss of color and structural information caused by particulate matter. To build a comprehensive PM\(_{2.5}\) concentration prediction model, we integrate multiple sub-PM\(_{2.5}\) concentration prediction models using a ’decision-fusion’ approach, resulting in our final PM\(_{2.5}\) concentration measurement model. Through rigorous testing, we confirm that this method outperforms existing photo-based PM\(_{2.5}\) monitoring techniques in terms of both accuracy and efficiency.
Loading