Research On Application Of Data Fusion Method For City Monitoring By Remote Sensing Technology

Data fusion is processing the remote sensing image by fusion methods to improve the spatial and temporal resolution of study area.The application of data fusion methods to urban remote sensing monitoring can improve the accuracy of urban information extraction and is of great significance to the rapid acquisition of urban information.In this thesis,Landsat 8 OLI-TIRS image data is used as the data source to extract the urban information in the study area.This study chose the southern part of urban area from Hohhot,Inner Mongolia Autonomous Region as study area which is located in from latitude of 40°45?33?N to 40°48?12?N and longitude of 111°36?47?E to 111°44?08?E and is close to the rural area.The images were transformed into HSV,Brovey,Gram-Schmidt,Principal Component,NNDiffuse Pan Sharpening after data preprocessing.Then extracting urban information by support vector machine based on five images and extracting urban information by NDISI based on Gram-Schmidt,Principal Component and NNDiffuse Pan Sharpening images.In order to determination of best approach for spatial urban information extraction which including both of the data fusion and information extraction,this research implemented statistical analysis of different data fusion images for mean,standard deviation,entropy of information,average gradient and correlation coefficient,and accuracy assessment of extraction results which are consist of support vector machine based results and NDISI based results.The results show that NNDiffuse Pan Sharpening is the best method by analyzing the parameters comprehensively,because it has the highest fidelity,clarity and much information.Classification result by support vector machine based on NNDiffuse pan sharpening images is better than PC images,and the worst are classification results of HSV and Brovey.NDISI image of study area based on PC images is the best result,which is better than NNDiffuse pan sharpening.NDISI image of study area based on Gram-Schmidt images is the worst.The NDISI image of study area based on PC images was found to be the most accurate by comparing with others.