A graph learning framework for prediction of missing landmines using airborne magnetometry in IoT environment

The continued use of landmines and the presence of already deployed landmines endanger countless lives worldwide. Detecting the landmines is a challenging task as it includes carefully surveying the target region using a suitable sensing device. One of the most trending sensory methods is the magnetometer sensing. The magnetometer sensor is used to visualize magnetic anomalies by measuring the magnetic field strength of magnetized objects especially in IoT environment. The process can be automated using a UAV-based survey by applying signal processing and computer vision techniques using the captured images. However, airborne magnetic detection fails to detect some of the landmines due to environmental conditions, thus the technology utilization is limited. On the other hand, the missing magnetic signature of some mines poses a severe risk to the mine removal team and civilians as they can be accidentally triggered. The work proposed is a further development of ongoing effort to come up with automated detection system. This paper focuses on post processing of magnetometry signature image of landmines deployed as rows in minefields, where mines are deployed in mu+++++ltiple rows patterns. The framework includes (1) a graph-based approach to identify the row clusters present in the pattern, (2) identify the row’s discontinuities and (3) localize the missing signatures via interpolation. A simulated dataset is used to test and evaluate the proposed strategies. The results signify the adequate performance of the proposed method in the identification of missing magnetic signatures of row pattern deployments which enhances reliability of the automatic airborne magnetic sensing of landmines.