Forest Fire Detection for Edge Devices

It is observed that the forest land mass was reducing rapidly from 1990 to 2020. As many plants and animals are depending on the forest, this is very alarming. Forest fire is one of the major causes of such loss. Forest fires tend to spread quickly and are difficult to control in a short time. Early detection of these forest fires is the key to mitigate the forest fire. There are many methods developed by researchers to monitor forest fire. An aerial-based detection system with unmanned aerial vehicles (U A V s) is one of the emerging methods which can provider a bird's eye view of the forest from above. Monitoring with UAVs however requires trained personnel to operate and manually monitor the forest. In this paper, we develop a fire detection algorithm that can analyzed images taken by UAVs and can be equipped into an autonomous UA V. The developed method does not require a lot computing power. It is based on YOLOv5 which is build and converted into optimized model that can run on an embedded board. Result shows that the method has a high MAP (>97%) with acceptable inference time indicating a good potential of the developed model.