典型丘陵山区遥感影像分类方法研究  

Remote Sensing Image Classification Method Based on Spatial Information Inference and Multi-scale Statistics

作　　者：徐钊 温小荣[2] 佘光辉[2] 

机构地区：[1]江苏省森林资源监测中心,南京210000 [2]南京林业大学森林经理系,南京210000

出　　处：《西南大学学报（自然科学版）》2013年第10期125-132,共8页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金资助项目(30571491);南京林业大学科技创新基金资助项目(CX2011-24);江苏省林业三项工程项目(lysx(2009)46);国家林业局公益性行业科研专项(200804006/rhh-11)

摘　　要：以Landsat ETM+为数据来源,基于ENVI平台与ID语言构建的条件分类模块为基础,面对林业应用中时常出现的地类复杂、分布交错、高程变化大的丘陵山地区域,在多尺度的光谱、纹理特征统计的基础上,通过对影像的充分认知,结合地形特征、知识经验,引入多种归一化指数,根据包容性、区分性原则建立阈值规则,得到一次分类结果后,应用优先级理论修正未划分区域,并根据面向地类间的空间关系推理,结合经验常识构建两种搜索模型用于叠合图斑修正.经过多次调整,参数修正,易混区划的生产精度、用户精度分别为85.57%,84.85%.Using landsat ETM＋ as the data sources and based on the condition classification module constructed with the ENVI Platform ID language, this paper introduces a remote sensing classification method on the basis of spatial information inference and multi-scale statistics in detail when we face the mountain ous and hilly regions which have complex land types and similar spectral information. Through reasonable and logistic inference, the wrong classification of land types can be recognized from the results on the way of multi-scale statistics. The inferring model is established with the influencing factors, such as priori knowledge of image, spectrum characteristics texture, land type mutual relations combined with INDWI and NDVI index. The inference of land types can be made by using this model through two times of searches. The result shows the confusion-prone regionalization of production precision is 85.57%, and the user＇s accuracy is 84.85%. In fact, this inferring in application. The parameters of this model may accuracy of the classification. model has been expressed by Interactive Data Language be continuously revised, which can further improve the accuracy of the classification.

关 键 词：遥感 影像认知 空间推理 分类模型 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]