| Land Surface Temperature is not only an important indicator to characterize the surface energy balance, but also a key parameter to study the surface bio-physical and chemical processes. Remote sensing is only the best means to fast access the regional or global surface temperature. The precise regional surface temperature is an urgent need for the research of many surface processes and thermal infrared remote sensing, such as the gas to water heat exchange, numerical weather prediction, global ocean circulation, climate change, ecological environment monitoring. In view of the thermal infrared band energy is lower than the visible band, thermal infrared image resolution is also not higher, how to get high-resolution thermal infrared band image is of great significance to the research of the geology regularity.The objective of this paper is, based on the visible band data of30m spatial resolution from landsat TM data, to get the thermal infrared band data of30m spatial resolution with the pixel decomposition to thermal infrared band data of120m spatial resolution from its dataset and to retrieve the Land Surface Temperature (LST) of30m spatial resolution with a mono-window algorithm.In the view of the conservation of energy, an attempt has been made in the paper to develop an overlay algorithm in the basis of the surface type for decomposing the thermal infrared band pixel. Firstly, based on the object-oriented thinking, each thermal infrared band pixel has been seen as independent and rules segmented。 Secondly, according to each of the segmented object corresponding to the type of surface types, for which remote sensing index has the highest correlation with the type of surface can been identified, thus, through the relationship between remote sensing index and LST, each of the divided sub-pixel weights in the mother pixel can been determined. Thirdly, a practical approach of decomposing the thermal infrared band of120m spatial resolution is been made.In the research, we chose the Chinese capital of Beijing as our study area, the Landsat TM data is the main data source, and the normalized vegetation index (NDVI)、impermeable layer coverage(ISP)and the modified Water Index (MNDWI) are the main remote sensing Index. Firstly, we get the land use and land cover of the study area through the object-oriented approach. Secondly, we retrieve LST of120m spatial resolution with the mono-window algorithm. Then linear statistical relationship between the remote sensing index corresponding to different surface types and LST in the appropriate sampling area was added up to calculate the weight of the Overlay Method. Finally, LST of30m spatial resolution in the study area was retrieved with the Overlay Method.In order to prove the feasibility of the method, the result is verified by two methods. The first one is self-examination. We scale the thermal infrared band of120m spatial resolution up to480m spatial resolution, then downscaled to120m spatial resolution. Contrast to the original TM6band data, the mean and the Variance of the error of two images DN value are0.0051,3.3. More than95%of the pixel error is within the range of±5. And, More than95%of the LST is within the range of±2℃with the Overlay Method. So it has verified the feasibility of the method. The other one is the comparison analysis of cubic convolution resampling method, Distrad methods and neural network fusion method. The result shows that the overlay method not only can effectively reflect the spatial distribution of surface temperature, but also can more effectively maintain the decomposition of the overall image and local energy conservation, and the error is small. So it is suitable to downscale the thermal infrared band image data in the complex surface coverage area.The innovation point of this paper is to decompose the thermal infrared band pixel with the conservation of energy and analyze the error. And discuss the feasibility in the complex surface coverage area, which can get a better decomposition results. the error of the method of analysis. |
PDF Full Text Request