DEEP Learning-based air quality monitoring model via BM-KMC using seasonal images
R. Udaya Shanmuga,G. Tamilpavai
DOI: https://doi.org/10.1007/s10668-024-05745-5
IF: 4.08
2024-12-06
Environment Development and Sustainability
Abstract:Nowadays, numerous cities have seen severe air pollution (AP) that affects the daily lives of people and produces harmful health impacts. As a result, researchers have shown considerable interest in monitoring AP. However, accurate monitoring of air quality remains a significant challenge. To overcome this problem, an image-based analysis framework is proposed here to monitor air quality. For this purpose, initially, input images relevant to diverse seasons are collected from publicly available sources. Then, the input images are converted into a hue, saturation, and intensity (HSI) color transformation model for easily analyzing and processing the colors independently. Afterward, the HSI images are enhanced by employing the HMFBHE algorithm for extracting information effectively in night scene images. Next, the sky segmentation is performed on the contrast-enhanced images by using the gibbs generalized gradient vector flow (GGGVF) algorithm, which is done by means of pixel values. Then, to evaluate the linear relationship's strength betwixt the variables, the correlation calculation is performed on the segmented sky images. Afterward, the correlated images are clustered as strong, medium, and low by employing the broyden model-based K-means clustering (BM-KMC). After that, the low correlation valued features are extracted from the clustering phase. In this way, from the other objects of the segmented sky image, the features, including buildings, trees, and so on are extracted. Thereafter, the optimal features are chosen from the extracted features by employing the ratio nambie beetle optimization (RNBO) technique. At last, from the selected features, the air quality is monitored as good, satisfactory, moderate, poor, very poor, and severe by using the LPKF-PSELU-ResNet algorithm. Thus, the experimental outcomes proved that the proposed model attained a high accuracy of 0.98677, which outperformed the prevailing techniques.
environmental sciences,green & sustainable science & technology