A Stochastic Model for Active Fire Detection Using the Thermal Bands of MODIS Data

Abstract

Active fire detection using satellite thermal sensors usually involves thresholding the detected brightness temperature in several bands. Most frequently used features for fire detection are the brightness temperature in the 4-m wavelength band ( 4 ) and the brightness temperature difference between 4-and 11-m bands (1 = 4 -11 ). In this letter, the task of active fire detection is examined in the context of a stochastic model for target detection. The proposed fire detection method consists of applying a decorrelation transform in the ( 4 1 ) space. Probability density functions for the fire and background pixels are then computed in the transformed variable space using simulated Moderate Resolution Imaging Spectroradiometer (MODIS) thermal data under different atmospheric humidity conditions and for cases of flaming and smoldering fires. The Pareto curve for each detection case is constructed. Optimal thresholds are derived by minimizing a cost function, which is a weighted sum of the omission and commission errors. The method has also been tested on a MODIS reference dataset validated using high-resolution SPOT images. The results show that the detection errors are comparable with the expected values, and the proposed method performs slightly better than the standard MODIS absolute detection method in terms of the lower cost function.

A Probabilistic Approach for Vision-Based Fire Detection in Videos

“…Many fire detection systems are based on satellite images or thermal analysis of satellite sensors [13]- [15]. However, this type of application is out of the scope of this paper.…”

“…When edited newscast videos are considered, the decision block "Edited Video?" indicates how the algorithm should compute the PFM: "Color Threshold"-corresponding to (13) or "Position Weighted Color Threshold"-corresponding to (13).…”

Detection of vegetation fires and burnt areas by remote sensing in insular Southeast Asian conditions: current status of knowledge and future challenges

“…Accuracy assessments of global active fire products conducted in this region have not only indicated high commission and omission error levels (∼30%), but also highlighted remarkable differences in the quantity and type of fires detected by different active fire products (Liew et al 2003;Stolle et al 2004 other local data to constrain the non-burning background, this tighter constraint permitting better detection of marginal fires than could not be captured by a global product without significantly increasing false-positives. In insular Southeast Asia, at least three local active fire detection approaches utilizing knowledge of local ground conditions, fire type, and varying atmospheric humidity levels have been developed (Nakayama et al 1999;Fuller and Fulk 2000;Liew, Lim, and Kwoh 2005a). Both Nakayama et al (1999) and Liew, Lim, and Kwoh (2005a) reported slightly better detection accuracy than the standard global algorithms, while Fuller and Fulk (2000) did not compare their results to any globally used algorithm.…”

“…In insular Southeast Asia, at least three local active fire detection approaches utilizing knowledge of local ground conditions, fire type, and varying atmospheric humidity levels have been developed (Nakayama et al 1999;Fuller and Fulk 2000;Liew, Lim, and Kwoh 2005a). Both Nakayama et al (1999) and Liew, Lim, and Kwoh (2005a) reported slightly better detection accuracy than the standard global algorithms, while Fuller and Fulk (2000) did not compare their results to any globally used algorithm.…”

Optimum systems for satellite fire detection

“…We assume a gaussian model to simulate the two fire temperatures T sm and T fl [8], with mean values μ sm = 600K, μ fl = 1000K, and standard deviations σ sm = 50K, σ fl = 100K. In order to simulate the natural variability of the background temperature T b we have chosen, as well, a gaussian model with appropriate parameters.…”