UAV Databased Temperature Patterns Analysis with Carbon Emission Detection Using Deep Neural Network

• Published in: Remote sensing in earth systems sciences (Online) 2024-12, Vol.7 (4), p.512-523
• Main Authors: Mohanty, Sachi Nandan, Dash, Bibhuti Bhusan, Shanmugasundar, G., MGM, Johar, Aswani, Inakollu, Sundaram, Ajith, Varghese, Issac K.
• Format: Article
• Language: English
• Subjects: Earth and Environmental Science; Earth System Sciences; Geography; Monitoring/Environmental Analysis; Remote Sensing/Photogrammetry
• ISSN: 2520-8195; 2520-8209
• DOI: 10.1007/s41976-024-00143-x

Unmanned aerial vehicle (UAV) imaging methods have drawn a lot of interest lately from academics and industry professionals as an affordable option for agro-environmental uses. To improve the UAV capabilities of diverse applications, machine learning (ML) methods are specifically applied to UAV-based remote sensing data. Spatiotemporal properties were analysed and the city-level carbon emissions statistics were estimated. This research proposes novel technique in UAV-based climate temperature pattern analysis and carbon emission detection utilizing the deep learning (DL) model. Here the input is collected as UAV-based weather data which is processed for noise removal and smoothening. Processed data is extracted and classified utilizing Gaussian belief deep neural network and spatial convolutional Q-swarm colony metaheuristic optimization. A metaheuristic optimisation algorithm uses gradient information-free iterative evaluation of the objective function in order to identify a global optimum. Experimental analysis has been carried out in terms of detection accuracy, average precision, F-1 score, recall and AUC for various UAV-based weather dataset. The proposed technique attained mean average precision was 92%, recall was 94%, AUC was 87%, detection accuracy was 96% and F1-score was 93%. This work shows that remotely sensed data can be used to support more advanced evaluations that are more successful, particularly in areas with extensive selective logging and diverse forest conditions, and to help quantify carbon emissions from selective logging using conventional methodologies.