Multi-Zone Fence Perimeter Surveillance: A New Edge-FOG Architecture for Efficient Detection and Classification of Intrusion

In this paper, we propose a geophone based fence surveillance system with a FOG architecture for detection of intrusion and classification in spatially separated zones. In the proposed decentralized architecture − for edge layer we propose an efficient spectral-energy-comparison detector; and at FOG layer, we propose a highly accurate supervised machine learning algorithm in the form of linear support vector machine for classification of mode of intrusion; lastly, the FOG layer updates the intrusion status of respective zones to the cloud layer. Extensive analysis of field experimental data acquired with an ad hoc geographically distributed fence setup indicates that the proposed detector renders \(99\%\) $99\%$ accuracy with very low false alarm rate and also outperforms known detectors. We perform feature engineering and demonstrate that the proposed classifier achieves \(97.9\%\) $97.9\%$ accuracy for both man-made intrusions and natural events even with reduced feature set. We also show that the proposed classifier outperforms known fence perimeter surveillance schemes. Lastly, we validate the performance of proposed system through real life experiments and analysis therein.