IoT and deep learning-inspired multi-model framework for monitoring Active Fire Locations in Agricultural Activities

Abstract

This paper proposes an Internet of Things (IoT) and deep learning-inspired multi-model system for detection, dissemination, and monitoring of Active Fire Locations(AFL) in agricultural activities. The IoT module of the proposed system works on the fusion of IoT sensors-based detectors and deep learning-based detectors. Fuzzy logic is used for the fusion of various sensors and providing real-time detection and location of AFL. The deep learning detector implements IP camera-based MobilenetV2 architecture for accurate and long-distance detections trained on a novel self-created dataset. The proposed framework also provides a software module for monitoring and tracking of various AFL. The software comes with several features like automatic extraction of fire locations from remote sensing sites, assigning active fire locations to multiple stakeholders, extracting farmers' names indulged in burning, automatic sending a notification to government agencies, and provisions for citizens centric participation. The results of the proposed framework are quite encouraging.