摘要
专题指数对遥感影像自动解译至关重要,现有研究多针对单目标信息提取来筛选专题指数,无法得到适用于多目标遥感自动解译的最佳专题指数。以德州市城区及周边地区为例,采用Landsat 5TM影像提取了2个植被、3个水体和3个建筑用地专题指数,基于面向对象分类方法,分析了单个专题指数、指数组合、指数数量对同时提取植被、水体和不透水层信息的精度影响。结果表明:(1)3类地物的最小分类精度基本上随着专题指数增加而增大;(2)从单个专题指数来看,不透水层和植被提取的最佳指数分别是建筑物指数和土壤调整植被指数,而新型水体指数则能显著提高总体分类精度;(3)从专题指数的组合来看,植被分类精度随所用的植被指数数量增加而下降;建筑用地指数越多,不透水层和总体分类效果越好;随着水体指数数量增加,水体分类精度有所提高,而不透水层和总体分类精度则随之下降。
Thematic indices are very significant for automatic remote sensing interpretation.However, exist- ing research methods,which are applied to select thematic indices, mostly aim to the single-objective re- mote sensing interpretation,and cannot obtain the optimum multi-objective thematic indices.According to process the Landsat 5 TM data of Dezhou city, two vegetation indices, three water indices and three built- up indices are produced; Then, based on the object-oriented classification,it is analyzed that how single the- matic index, composite of indices and number of index affect the classification accuracy of vegetation, water and impervious features respectively by extracting these land cover types simultaneously. The following conclusions are drawn： （1） Minimum classification accuracy of three land cover types will increase as the increase of number of thematic index;（2） Considering the single index, the best one to extract imperious and vegetation is Built-up Index （BUD and Soil Adjusted Vegetation Index （SAVI） respectively. Mean- while,the optimal water index to interpret the overall image is New Water Index （NWI）;（3） With regard to composite of indices, the interpretation accuracy of vegetation becomes worse as the increase of number of vegetation index; And the more built-up indices are used into the classification, the more accurate results are produced for the impervious features and the whole study area.Moreover, with the increases of number of water index,water can be more precisely identified in a certain degree,but the classification accuracy of impervious features and the overall region are declined gradually.
出处
《遥感技术与应用》
CSCD
北大核心
2017年第3期564-574,共11页
Remote Sensing Technology and Application
基金
国家自然科学基金项目(41371186)
关键词
植被指数
水体指数
建筑用地指数
面向对象分类
指数数量
指数组合
Vegetation index
Water index
Built-up index
Object-oriented classification
Number of index
Composite of indices
