Thermal data produced from remotely sensed data are significant parameters for investigating biophysical phenomena. However, the spatial resolution of TIR sensors is consequently constrained by the trade-off between spatial and spectral resolutions in Thermal InfraRed (TIR) remote sensing systems. Notable applications of thermal data include wildfires, volcanic activity, and land cover classification, but usage of remotely sensed data are limited to medium-resolution sensors such as Landsat-8 or Advanced Spaceborne Thermal Emission and reflection Radiometer (ASTER) products.In order to address this problem, various thermal sharpening methods of TIR data based on the Vegetation Index (VI), such as Normalized Different Vegetation Index (NDVI) and Fractional Vegetation Cover (FVC), have been developed to sharpen the coarser spatial resolution of TIR data. Although these methods exhibited a sufficient level of effectiveness, preservation of spatial details in the original TIR data still proved to be difficult, especially in urban areas due to the presence of heterogeneous land cover patterns.This study has improved the conventional thermal sharpening algorithm by modifying the input index and sharpening model. First, a novel index referred to as Fractional Urban Cover (FUC) is proposed for the thermal sharpening algorithm for Landsat-8 Thermal Infrared Sensor (TIRS). The FUC index is based on Visible Infrared Imaging Radiometer Suite (VIIRS) Day/Night Band (DNB) Night-Time Light (NTL) imagery. Second, to minimize the distortion of spectral features, a Modulation and Regression based Hybrid Thermal Sharpening (MRHTS) model, combined Least Square Regression (LSR) analysis and High Pass Modulation (HPM) were applied for Landsat-8 TIRS imagery. The proposed image sharpening algorithm using the MRHTS framework with the FUC index as its input is abbreviated as MRHTS-FUC.The MRHTS-FUC algorithm was applied to Landsat-8 Operational Land Imager (OLI), TIRS and VIIRS NTL images in the urban area. The proposed algorithm was compared with algorithms based on the Regression-based Thermal Sharpening (RTS) model and input variables: RTS-NDVI (DisTrad), RTS-FVC (TsHARP), RTS-FUC, MRHTS-NDVI, and MRHTS-FVC. Quantitative evaluation of the resulting sharpened images was conducted in terms of the synthesis and consistency properties.Regarding the synthesis property, Structural SIMilarity (SSIM) and Universal Image Quality Index (UIQI) yielded the highest values for the proposed method with values of 0.8013 and 0.8356, respectively. These results indicate that the proposed MRHTS framework effectively reflected the spatial details of the thermal imagery from the FUC index in this study. Regarding the consistency property, the proposed MRHTS framework returned better results over the RTS framework with respect to all of the image quality indices. In addition, visual analysis of the results revealed that the proposed algorithm successfully extracted spatial detail not only in vegetation areas but also in urban land cover, while preserving the spectral information.The experimental results demonstrated that the proposed algorithm was successfully applied to the TIR data, particularly in urban areas through quantitative and visual assessments. As an extension to the results from this study, urban morphology analysis of biotope and settlement structures can be conducted by enhancing the 100-m spatial resolution of Landsat TIR images to match the finer 30-m resolution of multispectral sensors.
【 预 览 】
附件列表
Files
Size
Format
View
Modulation and Regression based Hybrid Thermal Sharpening of Landsat-8 TIRS Imagery Using Fractional Urban Cover
PDF
download
878987797
028-85220240
