ZY-3卫星影像雪盖信息提取方法

摘要

针对国产资源三号(ZY-3)卫星多光谱影像空间分辨率高但波段范围较窄,在积雪提取方面研究缺乏这一现状,提出一种综合近红外(near infrared,NIR)波段和归一化植被指数(normalized difference vegetable index,NDVI)的决策树雪盖信息提取方法.以各地物光谱特性差异及NDVI特性差异为基础,方法充分利用积雪在NIR波段上反射率远高于其他地物,且其对应NDVI<0这两大特性,借助二次阈值化处理实现对积雪的判别划分.以最大似然监督分类提取结果为参考,对研究区雪盖信息整体提取结果、局部提取结果和提取精度进行统计和对比分析.实验结果表明,该方法具有受建筑物、阴影和裸土影响较小的优点,提取结果准确性、完整性较高.%Though ZY-3 satellite multispectral imagery have much better resolution, the snow information abstraction method based on it has not been studied yet because of its narrow wavelength range.Taking the actuality above into consideration, a decision tree snow-cover information extraction method which combines the Near Infrared (NIR) band and the Normalized Difference Vegetable Index (NDVI) using ZY-3 imagery was presented in this paper.Based on the difference of spectral characteristic and the difference of NDVI characteristic of land objects, by making the most use of two characteristics which the reflectance at NIR band of snow is much stronger than others and its NDVI is less than zero, the secondary threshold value method was used to achieve the identification of snow-cover.Taking the maximum likelihood classification result as the reference, the overall area of the snow-cover extraction results, the detail extraction and the confusion matrix were used to make statistic and comparative analysis.The experimental results show that the proposed method is less affected by buildings, shadow and uncovered soil and has higher precision and integrality in snow-cover identification.