A multi-sensor fusion approach for real-time landslide monitoring and early warning in reservoir areas using IoT and edge computing

Landslides are considered to be one of the most destructive among all the disastrous events. Landslides cause considerable damage to infrastructure, the environment, and loss of life, especially in mountainous and rainfall-affected areas. After the feature fusion process, the fused features are fed into a Long Short-Term Memory (LSTM) network with an optimization technique called Harris Hawks Optimization (HHA) for increased accuracy in the prediction process and computational efficiency. Additionally, a Dynamic Landslide Risk Index (DLRI) computation process is used to classify the region under surveillance into Safe, Moderate Risk, and High Risk levels. The time series data collected are processed using an edge computing architecture. The performance of the model in the regression analysis was evaluated using common regression metrics: These results show a high accuracy in landslide risk prediction. Moreover, the proposed framework presents a significant reduction in inference latency, parameters, energy consumption, and memory consumption, making it suitable for deployment in edge devices with resource constraints. Also, a real-time monitoring dashboard was implemented using multi-sensor data and DLRI. The proposed framework presents an efficient solution in the development of a real-time landslide early warning system.