Removing temperature drift and temporal variation in thermal infrared images of a UAV uncooled thermal infrared imager

Abstract

A thermal infrared (TIR) imager mounted on an unmanned aerial vehicle (UAV) has been widely used to obtain land surface temperature (LST) at very high-spatial resolutions. The TIR images are valuable for various applications, including mapping fine-scale surface evapotranspiration and monitoring crop water stress. However, UAV-borne uncooled TIR imagers generally suffer from temperature drift and temporal variation impact due to a long data acquisition period, significantly decreasing the initially acquired brightness temperature (BT) reliability and hindering subsequent applications. Here, a so-called DRAT (Digital number probability density function fitting and RAdiative Transfer simulation-based) method is proposed for simultaneously removing such temperature drift and temperature variation on the temporal scale. The DRAT is post-processing-based and only needs very few ground BT observations for calibration, thus significantly reducing dependence on auxiliary data. Test results show that the visual effect of the normalized TIR mosaics has been significantly improved, displaying more reasonable temperature distribution and good consistency with the land cover. The corrected BT has a root mean square error of 1.51 K and a mean bias error of 0.19 K by in situ data. The DRAT is valuable for obtaining high-accuracy LST and is feasible for other UAV uncooled TIR imagers with similar sensor structures and working principles. Therefore, it can significantly enhance the applicability of UAV TIR remote sensing.