High-temperature Target Identification And Temperature Retrieval Using Shortwave Infrared Remote Sensing Data

The surface temperature is mainly retrieved by using thermal infrared data. Ma-jor research methods include Mono-window Algorithm and band combination algo-rithms. However, the results of these methods are affected by the atmosphere, thebackground environment and the spatial resolution. Therefore, the aforementionedmethods are not effective in the identification and retrieval of small hot target. On theother hand, in some high temperature condition, the high-temperature surface hasspecial characteristics in the shortwave infrared band (1.3~3.0μm) with respect to thenormal-temperature surface, which can be used for high-temperature target recogni-tion and temperature retrieval.In this article, we analyze the feasibility of the shortwave infrared identificationand temperature retrieval. Based on the research about the mechanism ofhigh-temperature target remote sensing characteristics, we propose a single bandthreshold algorithm and ratio operation to identify the high-temperature target usingsensitive spectral range of high temperature target. Finally, we design a NormalizedDifferential Fire Index (NDFI) method for Landsat/ETM+data.Based on the high temperature target recognition, we establish a physical modelof short-wave infrared radiation on top of the blackbody radiation law, and then ana-lyze the sensitivity of parameters of physical model that then determines the range ofthe physical model parameters. In the end, we achieve the high-temperature targetidentification and temperature retrieval.The article made the following conclusions:1High-temperature target recognition method based on multi-spectral data per-forms well. The recognition accuracy of single band threshold algorithm, ratio opera-tion and Normalized Differential Fire Index (NDFI) method are81%,73%and94%.The results show that NDFI method has the best recognition results, and it ought beused for high-temperature target recognition.2Short-wave infrared radiation physics model results show that the radiationenergy of the mixed pixel consisting of the room temperature background surface feature and high-temperature target contains three components: the reflected radiationof the room temperature background surface, the reflected radiation of high tempera-ture target and the emitted radiation of high temperature target. The atmospherictransmittance, the reflected radiation of the room temperature background surfacefeatures and the emitted radiation of high temperature target are the key parameters oftemperature retrieval.3The physical model parameter sensitivity analysis results show that: when at-mospheric transmittance changes from0.8to0.9, the temperature increases5de-grees; when the emitted radiation of high temperature target changes from0.8to1.0,the temperature drops4degrees; when the reflected radiation of the room temperaturebackground surface features ρ changes from0.1to0.3, the temperature drops31de-grees; and when the high temperature target area ratio changes s from0.1to0.3, thetemperature drops54degrees. Therefore, atmospheric transmittance and the emittedradiation of high temperature target have little effect on the temperature T. In con-trast, the reflected radiation of the room temperature background surface featuresρhasa great effect on the temperature T, and the high temperature target area ratio changess has the largest effect on the temperature T.4The result of high-temperature target temperature retrieval using shortwaveinfrared remote sensing ranges from454.809K to608.228K. The high-temperaturetargets, including indigenous coking plants, magnesium smelter, thermal power plant,have temperature retrieval results coinciding with the real data. However, in the samecondition, the temperature retrieval result of thermal infrared data is294.2K to306.5K, which did not show obvious thermal anomalies.