Normalization Of UAV Thermal Infrared Images And Land Surface Temperature Retrieval

Land surface temperature（LST）is a fundamental parameter in the earth’s surface and atmospheric system,which can be used to describe the inherent physical processes of energy and material exchange.Compared with satellite remote sensing,unmanned aerial vehicle（UAV）is light and flexible to quickly obtain remote sensing images,and has the characteristics of high timeliness,high spatial resolution and large scale.Besides,its images are not easily obscured by clouds,which makes up for the deficiency that satellite remote sensing images are easily affected by clouds.Since UAV thermal imagers are typically uncooled sensors,the temperature-drift of the thermal infrared images can seriously reduce the reliability of final data.Therefore,to study the method of removing the temperature-drift and to normalize the thermal infrared images is the basis and key to further enhance the value of the data.Meanwhile,due to the wide range of the spectral response functions of UAV thermal imagers,the LST retrieval algorithms currently applied to satellite thermal infrared sensors may induce significant uncertainty if they are directly applied to UAV thermal imagers,thus severely deteriorating the application effect of the LST data.Yet,few studies have considered the adverse effect of such problems in the retrieval of LST for UAV remote sensing.To address these issues,the main works of this thesis are as follows:（1）A digital number histogram function fitting and radiative transfer simulationbased method（DRBM）was proposed to remove temperature-drift of the thermal infrared images,that is,the temperature levels of the images were normalized to the level without temperature-drift at a specific time.The results show that the standard deviation（STD）of the mean value of the normalized bright temperature image sequence is reduced by39.4%–84.1% with an average of 60% compared with that before normalization.The visual effect of the normalized mosaics is significantly more uniform and more consistent with the temperature distribution pattern of the ground objects in the study area than that before normalization.（2）An LST retrieval method for UAV broadband thermal imager data was established.This method is based on a simplified look-up table algorithm to construct a mapping relationship between the LST and the surface emitted radiance,thus avoiding the large uncertainty caused when the equivalent wavelength of the thermal imager is directly substituted into the Planck formula to calculate the LST.In addition,an estimation model of atmospheric parameters is established in this method for atmospheric correction of thermal infrared data.The validation results in the study area show that the retrieved LST agrees well with the in-situ LST,with a coefficient of determination of 0.96,a root mean square error of 1.80 K,and a mean bias error of 0.56 K.In this thesis,we explored targeted solutions for the main problems faced by LST retrieval from UAV remote sensing.Considering the similar structure and working principle of UAV thermal imagers,the findings of this research are also expected to be extended to other models of thermal imagers,so as to further improve the application prospect of UAV thermal infrared data.