Hotspot Detection in Nighttime Landsat Data

We propose a statistically based method to detect hotspots in nighttime short-wave infrared Landsat data. The method assumes an independent and identical Gaussian distribution on the background data and looks for parts of the images with abnormally high values. For this, the variance of the background model is estimated using a robust estimator, being able to provide a good estimation even in the presence of outliers (hotspots). Then, a region growing algorithm is used to extract 4-connected regions with high values. Finally, a statistical test is used to decide whether the sum of the values of each region is significantly higher than expected on the background model. Only regions detected in the two shortwave infrared bands are validated. The test level is selected in order to control the number of false detections. Compared to classical pixel-based methods, our approach allows the detection of hotspots with lower radiance while keeping a low commission error rate. Experiments on a time-series of acquisitions over an oil and gas-producing region showed that this greatly increases the number of detections.