Extracting Urban Green Space Information Based On Hyperspectral Hyperion Remote Sensing Data

To obtain the urban compositional information quickly is of great theoretical and practical significance , it can increase the public environment protection consciousness and perfect the urban land price system, furthermore, it is of great importance for the regulation of green space management. Remote sensing technology offers an alternative to traditional ground-based survey of these green spaces. Traditional methods for the mapping of green space from remote sensing data such as classification techniques and vegetation indices were found to be inaccurate.In this study we apply a linear spectral unmixing approach to hyperspectral data to map urban green space. Spectral mixing analysis(SMA) is an model based on the linear mixing of two or more pure spectral endmembers , it allows for variability in composition and illumination within an image. SMA gives the opportunity to categorise the scene into various sub-areas and guilds the endmember selection for a better adjustment of the model to different surface types.EO-1 Hyperion data was used for this study, the Hyperion gathers near-continuous data in 242 discrete narrow bands along the 400–2500 nm spectral range at a 30 m spatial resolution and in 16 bits, and allows the relative contributions of different materials to the spectrally heterogeneous radiance field to be determined and their abundance to be mapped. Data processing occurred in several steps. The first of these was to select the useful bands, of the original 242 Hyperion bands, 176 bands were unique and calibrated, other bands were therefore dropped.Then, the original Digital Number (DN) of Hyperion from the 176 bands were converted to radiance , using the DN-to-radiance conversion factors that accompanied the Hyperion data. The next step of preprocessing was the atmospheric correction to convert the radiance to surface reflectance. We processed the data using the above water reflectance spectra with an atmospherically corrected image. Images were corrected using the ENVI FLAASH (Fast Line-of-Sight Atmospheric Analysis of Spectral Hypercubes) atmospheric correction software package. The FLAASH module incorporates MODTRAN 4 radiation transfer code with all MODTRAN atmosphere and aerosol types to calculate a unique solution for each image.Finally, we specifically used a linear SMA (LSMA) model and performed a minimum noise fraction (MNF) transformation and pixel purity index (PPI) on EO-1 Hyperion image to derive the proportion of ground cover (green space,buildsings ) and water within a pixel.With the endmember spectra, a fully constrained linear mixture analysis was performed to generate fraction images for each endmember. The resulting green space fraction map showed the distribution and relative abundance of its endmember in the field.The overall root mean square (RMS) errors was 0.237%, unmixing errors occurred mainly due to multiple scattering as well as close endmember spectral correlation. The result indicates that spectral unmixing applied to hyperspectral imagery can be a useful tool for mapping green space in the urban area. Compared with other information extraction methods, we found that LSMM got better green space information results than ISODATA method,Maximum Likelihood Classification (MLC) method and NDVI density slice method, thus indicate that LSMM is a better information extraction method for extracting green space in the research area.Future work will incorporate the imaging spectrometry to provide more accurate endmembers of the urban area as a reference, additionally, to detect the green space or other physical compositions, a time series of Hyperion data will be analyzed with the application of LSMM.