Extraction of the urban green space based, on the high resolution remote sensing image

摘要

High resolution image can be used to distinguish the small difference of the ground things. Texture information canavoid the matter of same spectral from different objective and different spectral with same objective which must be facedwhen making classification with only spectral information. The main objective of this research 'was to determine thecapacity of high spatial resolution satellite image data to discriminate vegetation in urban area. A high spatial resolutionIKONOS image, coincident field data covering the urban area of linping scenic region in Yuhang town, Zhejiangprovince in china, was used in this analysis. The vegetation of test region was classified as tea garden, masson pine, fir,broadleaves, and shrub/herb based on the field data Semi-variograms were calculated to differentiate vegetation classesand assess which window sizes were most appropriate for calculation of grey-level co-occurrence texture measures. Thetexture analysis showed that co-occurrence mean, variance, contrast, and correlation texture measures provided the mostsignificant statistical differentiation between vegetation classes. Subsequently, a decision tree classification was appliedto spectral and textural transformations of the IKONOS image data to classify the vegetation. Using both spectral andtextural image bands yielded the good classification accuracy (overall accuracy=81.72%). The results showed that it hasthe higher accuracy to extract the urban green space from IKONOS imagery with the spectral and texture information, aswell as the vegetation index.