An Optimal Serverless Cloud Computing Framework for Large-Scale Remote Sensing Data Analysis

Abstract

Translate Article

With the exponential growth of remote sensing data, efficiently analyzing large-scale satellite imagery remains challenge due to computational complexity and resource constraints. This paper in turn suggests a superior serverless cloud-computing architecture to overcome such constraints with Hunger Game Optimization (HGO) to introduce live resource scheduling and Deep Reinforcement Learning (DRL) to introduce intelligent workflow. The suggested architecture processes cloudless Landsat images between the years 1991 and 2022 in both dry and rainy seasons. Data will be obtained on the Landsat series, namely <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$5,7,8$</tex> and 9, with a 30meter spatial resolution, therefore providing a vast temporal and spatial coverage to be used in the analysis to follow. Experimentally, the framework has been shown to provide an average <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$7.8 x$</tex> speedup compared to traditional server-based models, and at the same time have a 28% higher resource utilization. Notably, the classification accuracy is maintained at a high rate, with the overall F1-score of 92.3%. The findings demonstrate the effectiveness of combining HGO and DRL in a serverless cloud, which is a scalable, cost-efficient, and accurate solution to long-term remote sensing data analysis, and environmental monitoring.