益阳地区遥感土地利用分类的波段组合效应分析

波段组合对监督分类的运算效率和分类结果有着显著影响。本文以湖南益阳TM影像为实验数据，通过波段光谱特征统计分析、叠合光谱图分析，对该地区土地利用分类波段进行了初步选取；按照3波段组合方式进行分类，并对分类效应进行了比较分析。结果表明，耕地和草地难区分、未利用土地易错分为建设用地是共同问题；TM2、3、4组合能够较好地区分耕地和草地，同时对林地和草地的区分效果也最好，分类总精度最高；TM2、3、5组合对未利用土地的识别能力相对最高；TM2、3、7组合，TM2、4、7组合以及TM2、5、7组合对水体的分类效果最好。本研究对数据降维、资料不足时最大限度发挥资料可用性、针对特别应用目的进行传感器波段设计具有一定的参考价值。

Band Combination and Effect Analysis of RS-Based Supervised Classification of Land Use of Yiyang

The band combination has a significant effect on the operation efficiency and the classification results of supervised classification. In this study,with the TM data in Yiyang, Hunan, the classification bands are selected initially by the statistical analysis of band spectral characteristics and the analysis of coincident spectral plots. Then arbitrary three-bands are selected for classification and ten results of classifications are obtained and analyzed comparatively. The results show that, without exception, it is difficult to distinguish between cultivated land and grassland and unused land is tend to be identified as construction land in a high probability in all kinds of band combinations. However, with the combination of TM2, TM3 and TM4, it is easier to distinguish between cultivated land and grassland and between woodland and grassland relatively. That classification has the highest overall accuracy. The combination of TM2, TM3, and TM5 has a relatively strongest ability of identifying unused land. It is easier to identify water with the combination of TM2, TM3, and TM7, the combination ofTM2 ,TM4, and TMT, as well as the combination of TM2, TM5, and TM7. This study has reference significance on data dimensions reduction, mining data availability without sufficient data, and the bands setting of sensors with special application purposes.