基于环境与灾害监测预报小卫星的树种识别

摘要

The experiment was conducted to classify typical forests including Mongolian oak, birch and larch in Wang Qing forest area of Jilin Province with the hyperspectral remote sensing image of small environment and disaster monitoring satellite. With the gray value difference among the forest types in the same band, three band sections are good separability for three forest types from 115 bands of HJ-1A remote sensing image.The classification rules to classify preliminarily were built based onvegetation gray value knowledge, and thenreclassified aspect data and DEM (Digital Elevation Model) data com-bined with terrain factors.The classification accuracy of preliminary classification was68 .33%, and the classification ac-curacy of aspect data and DEM data combinedwere 81.67%and 80.00%, respectively.Inpreliminary classification com-bined with aspect data and DEM data, the classification accuracy was 88.33%.%应用环境与灾害监测预报小卫星的高光谱遥感影像，对吉林省汪清林业局施业区内的典型树种蒙古栎、白桦和落叶松进行分类。依据各树种在相同波段灰度值的差异性，从HJ-1 A遥感影像的115个波段中提取3个树种可分性好的波段区域，建立基于植被灰度值的分类规则进行预分类，再结合地形因子的坡向数据和DEM数据等地形因子进行再分类。预分类的总体分类精度为68．33％，分别结合坡向数据和高程数据的分类精度为81．67％和80．00％；在预分类中，结合坡向和高程数据的总体分类精度为88．33％。