Unpixel Technique Of Remote Sensing Image And Its Applications In Land Cover And Ecological Environmental Changes

Medium-resolution remote sensing images (TM/ETM+, ASTER) are widely used in urban land use/ land cover due to their low-cost, long acquiring time and multi-spectral characteristics. But there exist a lot of mixed pixels of different covers in medium spatial resolution remote sensing images because of strong spatial heterogeneity of urban land. Traditional image processing methods of remote sensing (supervised and unsupervised classification) are based on the pixel level when one pixel corresponds to a type of object. The precision of results often does not meet practical requirements. To improve the accuracy of Medium Resolution Image in urban studies, we must solve the problem of decomposition for the mixed pixel.Taking Huadu, Guangzhou as an example, this study discussed V-I-S model for the problem of mixed pixels in medium resolution remote sensing images. The ETM+ and ASTER images of Huadu in 2003 were used for performing depixel process. Firstly, based on V-I-S model, four kinds of endmember including vegetation, soil, low reflectance objects and high reflectance objects were selected. Then, some transformations were carried out to optimize the selection of endmembers. Through MNF, the image dimensionality was reduced. PPI transformation was used to narrow the range of endmember. In the scatter plot of MNF, four kinds of endmember were selected, which were used as four kinds of extracted endmember including vegetation, soil, low reflectance object and high reflectance object. From the abundance images of low reflectance object and high reflectance object abundance images, the imperious surface abundance image can be calculated after excluding the water by masks. Here high-resolution QuickBird image was used to check up the precision of the imperious surface abundance image and the results show that both ETM+ and ASTER image can be used to extract imperious surface abundance image validly and the results from ETM+ is better. At last, the vegetation, imperious surface and soil abundancc from ETM+ and ASTER were analyzed according the actualities of Huadu.In this study, the changes of vegetation, imperious surface and soil abundance from 1998 to 2003 were also performed after depixel processing TM data in 1998. The results show that vegetation abundance of Huadu from 1998 to 2003 declined by 3.35%, imperious surface abundance increased by 3.86% and soil aboundance declined by 0.74%. After studing the changes of V-I-S of each towns of Huadu, it is found that ecological environment exhibie degradation trend with the development of the city, the vegetation abundance may drop, the imperious surface abundance will increase furtherly and the soil cover play a transition role. The resules demonstrate the application potential of depixel technique in land cover and ecological study.