The Study Of High-temperature Targets Temperature Retrieval Model In Shortwave Infrared Remote Sensing

It is evidence that high temperature targets’ temperature is higher than normaltemperature objects. High temperature targets such as coal seam spontaneouscombustion, grassland fire, oil well torch, forest fire, volcanic eruption etc. have greatsignificance in theories and have important pratical value to resources investigation,environmental monitoring and disaster warning, etc.Remote sensing temperature inversion mainly uses mid-infrared and thermalinfrared data. Mid-infrared and thermal infrared data are greatly influenced by theatmospheric background radiation and spatial resolution. This article based on theprinciple of energy conservation establishes short-wave radiation model and analysesparameters sensitivity. At last, it achieves temperature inversion. The mainconclusions are as follows:1. High temperature target’s mixed pixel consists of the emission of hightemperature target, the reflection of high temperature target, the reflection of normaltemperature background. The atmospheric transmittance Tθ, the reflectivity of normaltemperature background ρ, the emissivity of high temperature target ε, the percentageof high temperature target S are the key parameters of the model. Atmospherictransmittance Tθcan calculate through MODTRAN. The reflectivity of normaltemperature background ρ can be valued through background pixels. The emissivityof high temperature target ε can be replaced by the conversion of reflectivity in theheap coking and in other situation it can be seen as blackbody.2. According to the characteristics of Plank formula, the higher T is, the greaterblackbody radiation flux is. For high temperature mixed pixels at SWIR band, whenthe high temperature emission radiation energy equals to reflected energy or higher,high temperature targets can be recognized. That is to say, when the energy of hightemperature pixels are twice as normal temperature pixels, high temperature targetscan be significantly distinguished. So it is feasible to SWIR band to identify hightemperature targets and inverse temperature. In this case, the lowest temperature is525K and the smallest percentage of area is0.0003. It also can be indentified in asmall area if it has a highest temperature. 3. High temperature targets have significant spectra differences between hightemperature targets and normal temperature background, so from the perspective ofmultivariate analysis background-anomaly separation method for identification can beadopted. Mahalanobis distance can eliminate the interference of the correlationbetween variables and identify outliers. R-mode factor analysis can sum the multiplevariables samples up to a few factors via performing correlation analysis to them. Itcan extract information with the least amount of information loss and it can identifyhigh temperature targets. The identification precision of mahalanobis distancemultiple truncation method, mahalanobis distance multiple discriminant analysismethod and factor analysis method, respectively, are82%,79%and85%.4. The results of high temperature targets’ short-wave infrared temperatureinversion of the study area are conformity with the actual temperature(500K±). Theresults show that values of normal temperature background and high temperaturetargets are relatively close, which make it not easy to distinguish the targets frombackground. That is to say, thermal infrared remote sensing data fails to reflect thefeatures of high temperature targets properly.