Anomaly Detection and Visual Perception for Landslide Monitoring Based on a Heterogeneous Sensor Network

The paper seeks to develop a landslide monitoring system based on wireless sensor networks, which consist of geological sensor nodes and camera sensor nodes. The architecture and work mode of the heterogeneous sensor network are first introduced. Then anomaly detection based on geological sensor nodes is analyzed in terms of self-learning threshold, information fusion and sleep scheduling. Once geological sensor nodes in the lower layer confirm a slope anomaly, camera sensor nodes in the upper layer will be activated to perform visual analysis involving object detection and image compressive transmission with automatic image quality control. Finally, we build a tiny simulation platform and the test results obtained from preliminary laboratory trials show that the landslide monitoring system is feasible and effective.