Improved algorithm for adaptive fire detection using MODIS data

TL;DR: In this paper , a new adaptive fire detection algorithm was proposed, which focuses on two methods to improve the accuracy of fire detection: single-channel and multi-channel test conditions were added and new contextual algorithms were adopted; second, a method for weighting the fire test conditions based on the test conditions for differences in the sensitivity to fire was proposed.

Abstract: The Moderate Resolution Imaging Spectroradiometer (MODIS) has 36 channels covering the visible to far infrared band range with a resolution of 250 m to 1 km, and is important for detecting fires in large areas. Traditional fire detection algorithms mainly rely on thermal infrared channels using threshold or contextual methods. Such methods usually fail to detect small fire and often misidentify high-temperature objects on the surface. This paper proposes a new adaptive fire detection algorithm, which focuses on two methods to improve the accuracy of fire detection. First, single-channel and multi-channel test conditions were added and new contextual algorithms were adopted; second, a method for weighting the fire test conditions based on the test conditions for differences in the sensitivity to fire was proposed. This method reduces the issue of small fires being overlooked because they do not satisfy certain test conditions. In addition, a priori database was built using twelve-year US wildfire reference records and highest confidence fire data in MODIS fire products, adaptive thresholds suitable for fires were selected using the bubble sorting method based on the radiation characteristics of global fires. Testing results show that the improved algorithm improved the accuracy of small fire identification and reduced the false detection rate of pseudo-fires.