基于地表生物物理参数的土地利用/覆盖遥感分类方法研究——以长沙市以例

土地利用/覆盖分类通常是利用地物的波谱反射特征进行监督或非监督分类,分类结果由于"同物异谱、异物同谱"现象的存在,往往分类精度不高。而植被指数和地表温度作为表征地表覆盖状况的生物物理参数,已成功用于宏观尺度的土地利用/覆盖分类,使得分类结果有所提高,而对于区域尺度的土地利用/覆盖分类却少见报道。本文充分利用TM数据的多光谱特征,从中提取了植被指数NDVI、地表温度Ts、温度植被角度TVA和温度植被距离TVD这四种分类特征进行监督分类,通过对7种组合方案(反射率波段组合、NDVI与反射率波段组合、Ts与反射率波段组合、NDVI与Ts和反射率波段组合、TVA与反射率波段组合、TVD与反射率波段组合、TVA与TVD和反射率波段组合)的分类结果进行比较,得出以下结论:①NDVI、Ts、NDVI和Ts、TVD作为分类特征参与到多波段地表反射率影像分类中,能够提高分类精度,而TVA、TVA和TVD的加入却没有改善分类结果;②总体分类精度受到训练样本与检验样本比例的影响。

Research on land use/cover remote sensing classification using surface biophysical parameters——A case study of Changsha city

Spectral reflection images are used for supervised or unsupervised classification.Because of the phenomena of "the same kinds of targets with different spectral or the different kinds of targets with the same spectral",the classification accuracy is low.Vegetation Index and surface temperature are two Biophysical parameters that represent the condition of the land cover,and have successfully applied into the large scale land use/cover,but it is seldom reported that they have applied into regional land use/cover.Based on the advantage of TM multi-spectrum data,this paper extract four classification features including Vegetation Index NDVI,land surface temperature Ts,Temperature-Vegetation Angel TVA and Temperature-Vegetation Distance TVD for supervised classification.Compared seven image processing routines(incorporation of vegetation index,surface temperature,vegetation index and surface temperature,temperature-vegetation angel,temperature-vegetation distance,temperature-vegetation angel and temperature-vegetation distance),the results indicate that the proportion of training samples and test samples influences the overall classification accuracy.It hasa higher classification accuracy using NDVI,Ts,NDVI and Ts,TVD,whereas,TVA 、TVA and TVD almost have no effect on the classification accuracy improvement.