多光谱数据波段选择方法试验研究——以湖北神农架林区为例

本文以神农架林区植被信息提取为例，从统计特征的角度出发，采用最佳指数因子、联合熵与类间、类内可分性判别准则三种波段选择方法。在对三种波段选择方法计算结果综合分析的基础上，结合试验区地物光谱特征和TM传感器不同波段功能，采用逐步逼近的思路，从候选波段组合中确定了最佳波段组合。试验得出TM传感器453波段组合为神农架林区植被信息提取的最佳波段组合。

Study on Band Selection Algorithm of Multi-spectral Remote Sensing Data:Taking Hubei Shennongjia as an Example

Rapid and accurate selection and combination of multi spectral remote sensing data play an essential role in the multi spectral data processing and information extraction. In this paper, a case study of extracting Shennongjia vegetation information is exhibited. From the view of statistical characteristics, three kinds of optimal feature selection are tested, namely the optimum index factor, the joint entropy and category separability criterion. Based on a comprehensive analysis of the above results, combined with spectral features of the sample area and functions of different bands of TM sensor, a step by step approach is applied for extracting the optimum band combination from the candidates. For extracting Shennongjia vegetation information, the trial shows that TM bands 453 is the optimum band combination.