基于ENVI4.5的土地利用覆被信息监督分类对比研究——以海南省屯昌县为例

Comparative Study on Supervised Classification of Land Use and Cover Information Based on ENVI4.5

选取海南省屯昌县作为试验区,以2005年土地利用现状图和2008年经校正的SPOT影像为数据,利用ENVI4.5中的监督分类提取了影像中的耕地、园地和林地信息,并对提取精度进行比较分析。结果表明,平行六面体法在提取耕地信息方面的精度最高,可以达到85.60%,最小距离法在提取园地信息时可以实现82%的精度,最大似然比分类法提取的林地信息精度在80.38%。"异物同谱"和"同物异谱"现象是也是影响分类精度的主要原因。

Taking Tunchang County of Hainan Province as test area, using present situation map of land use of 2005 and corrected SPOT image of 2008 as data, and using supervised classification in ENVI4.5 to extract cultivated land, garden and forest land information, and the extracted accuracy was compared and analyzed. The results showed that the extracted accuracy of parallel six surface method in cultivated land information was highest, and reached 85.60%, and the extracted accuracy of the minimum distance method in garden information reached 82%, the forest land information accuracy extracted by maximum likelihood classifier amounted to 80.38%. The phenomenon of ＂different objects same image＂ and ＂same object different images＂ are the main influencing factors of classification accuracy.