Interactive Visualization of Hyperspectral Images in Sustainable Environments Using Machine Learning

Abstract

In the endeavour to achieve sustainable development objectives, the notion of smart cities has gained significant importance in recent times. New ideas for more sustainable cities have generally surfaced as a result of smart-city services addressing prevalent urban issues. One of the most important topics in realm of remote sensing is categorization of hyperspectral images (HSI). For conventional machine learning (ML) models, the classification process is quite difficult since HSI typically needs to cope with complicated features and nonlinearity among the hyperspectral data. Aim of this research is to propose novel technique in urban area green region analysis based on hyperspectral imaging in geographical information system using machine learning and smart grid for sustainable city application. Here the IoT based environmental monitoring sensor integrated with smart grid has been used in urban area green region analysis. Then through IoT module the monitored sensor hyperspectral images has been collected and processed for analysing the green region using graph convolutional U-net adversarial neural network. experimental analysis has been carried out based on various hyperspectral images in terms of training accuracy, precision, sensitivity and Normalized square error. Then the analysis of smart grid module is carried out in terms of throughput, end-end delay, packet delivery ratio, QoS. Proposed technique training accuracy 97%, precision 96%, Normalized square error of 58% and sensitivity of 95% for hyperspectral image analysis. The proposed GCU-netANN obtained accuracy of 97%, packet delivery ratio of 95%, QoS of 91%, end-end delay of 75% for IoT smart grid network analysis.