A A Madhuri, Balamuralikrishna Thati, Suresh Chandolu, A Sindhura, Yalanati Ayyappa, Ruhisulthana Shaik
This study proposes a unified environmental surveillance system, which fuses IoT sensor networks and the visual analytics of multispectral remote-sensing to address the shortcomings of the conventional surveillance solutions. Edge-based LSTM anomaly detection on distributed nodes of the IoT can offer high-frequency local measurements, whereas a hybrid ResNetVision Transformer (ViT) model can improve the analysis of the satellite image. An adaptive Kalman-based temporal-spatial fusion algorithm incorporates heterogeneous streams of data towards better environmental intelligence. The system was highly performing, indicating the accuracy of the IoT sensors in 91.3-98.1% and a hybrid model at 92.4% and the fused levels at 94% and above respectively. The results were impressive on the system level, since the response time to events was improved significantly, the completeness of data improved, and the accuracy of anomaly detectors increased, as well as the network load decreased. On the whole, the suggested structure has high potential to monitor the environment in real-time, being scalable, in the fields of smart agriculture, air-quality monitoring, water-resource control, climate-risk identification, and smart urban governance.