植被指数在城市绿地信息提取中的比较研究

利用植被指数从TM 影像中提取植被, 从技术与经济成本方面综合考虑, 是一个比较好的手段。但在城市绿地信息提取中, 由于城市下垫面的特殊性和植被指数的繁多, 究竟哪些植被指数最适合于城市绿地, 还仍然是一个急待解决的难点问题。通过以上海中心城区为研究靶区, 利用单因子方差分析与多重比较对植被指数在城市绿地信息提取中的优劣进行比较研究, 得到如下结论: ①TM 影像经过植被指数计算处理后, 植被信息确实得到了增强, 但不同的植被指数也有所差别。如果以区分植被与非植被之间差异程度做标准, 那么植被指数提取植被由优到劣则依次是GEMI、RDVI、NDVI、GNDVI、RVI、TNDVI、DVI、EVI 和TGDVI。②植被指数基本能从TM 影像提取植被, 但把植被再细分的效果不是太好。总体来看, 除EVI 和TGDVI 以外, 植被指数能较好的区分草地与农田; 而树林与农田及草地与树林的区分则因不同的植被指数有所差异。区分草地与树林较好的是EVI, 区分草地与农田较好的是GEMI, 区分树林与农田较好的是TNDVI。③植被指数不但细分植被的效果不是太理想, 而且也不能很好的细分非植被地物。总体来说, 所有的植被指数都很难把建筑物与道路区别开, 尤其TGDVI、DVI 和EVI 更是如此。不过NDVI、GNDVI、TNDVI 和GEMI 能很好地把水体从TM 影像中提取出来, 其余的植被指数则只能区分植被与非植被, 不能再进一步的区分非植被地物。

A Comparative Study of Extracting Urban Vegetation Information by Vegetation Indices from ThematicMapper Images

As to technology and economic cost synthetically, extracting urban vegetation features by vegetation indices from Thematic Mapper (TM ) images is a comparatively good way. Because of the particularity of the urban vegetation and varieties of vegetation indeices (VIs) , Which vegetation index is suitable to discriminate urban vegetation features still remains a p roblem to be solved urgently. This study area is located in Shanghai. We utilize a combination of one-way ANOVA and multiple comparisons approaches to demonstrate the differences and similarities in sensitivity to vegetation conditions of the nine
indices including NDVI, DVI, EVI, GEMI, GNDVI, RDVI, RVI, TGDVI and TNDVI. The analysis results indicate: ①After TM images being managed by VIs, the vegetation information in TM images has really been enhanced. But their potential for discriminating vegetation of different VIs is not the same.Our result shows the performance and validity of discriminating vegetation of VIs is evaluated as follows:GEMI, RDVI, NDVI, GNDVI, RVI, TNDVI, DVI, EVI and TGDVI from excellent to bad. ②VIs can extract vegetation features from TM images basically, however, can’t do well in discriminate the vegetation in detail. In all, VIs can identify grass and farmland except EVI and TGDVI. EVI does better in discriminating grass and woods while GEMI is more suitable to distinguish grass and farmland. On discriminating woods and farmland TNDVI is a better choice. ③VIs can identify neither vegetation nor non-vegetation further. On the whole, it is very difficult for all the vegetation indices to discriminate buildings and roads, specially for TGDVI, DVI and EVI. NDVI, GNDVI, TNDVI and GEMI can extract the water information from TM images well, while other vegetation indices can only separate vegetation from non-vegetation and are not suited to discriminate non-vegetation further.