A Novel Hierarchical Wildfire Alarm System Based on Vegetation Features

A Novel Hierarchical Wildfire Alarm System Based on Vegetation Features

In recent years, frequent wildfires have not only severely damaged the environment and ecology, but also threatened the environment of human life. In order to enable fire alarms to detect fire disasters earlier, many related works have proposed image fire detection methods using global positioning systems (GPS) and unmanned aerial vehicle (UAV). These related work intends to make the fire detection method accurate and send early to reduce the threat posed by wildfire. The related work of fire detection is to use fire with high energy as the detection target. However, the work that only uses fire features to detect when used in fire alarms often produces false alarms. In order to understand the cause of the false alarm, we found one of the possible reasons from the fire-related report. The fire related report pointed out that the traditional fire triangle has been extended in time and space, and the original oxygen, fuel and heat sources have become climate, vegetation and ignition. Information related to climate features can already be obtained from existing equipment other than images. In related studies, the most frequently detected fire is the ignition that extends the fire triangle. The last fuel-extended vegetation is a feature that has not been used for image fire detection in related work. Therefore, based on the features of deep learning convolutional neural network (CNN), this article proposes a flame detection task that uses both vegetation and ignition features at the same time. And compare the experimental results with related work. The results of the comparison show that the features we proposed allow the fire alarm to achieve 100% true positive rate (TPR) as in the past, but also reduce the false positive rate from 40.47% to 4.15%. This experimental result shows that the features we propose can effectively assist the flame alarm system to reduce the chance of false alarms. At the same time, a method is proposed to further support the automated fire alarm system to effectively respond to wildfire threats.