利用logistic模型对MODIS数据进行湿地提取  

Method of abstraction wetland using modis data based on logistic model

作　　者：刘宏娟[1] 胡远满[1] 布仁仓[1] 周昊[2] 冷文芳[1] 吕久俊[1] 

机构地区：[1]中国科学院沈阳应用生态研究所 [2]沈阳农业大学林学院,沈阳110161

出　　处：《辽宁工程技术大学学报（自然科学版）》2007年第5期791-794,共4页Journal of Liaoning Technical University (Natural Science)

基　　金：中国科学院创新基金资助项目(KSCX2-SW-133);国家自然科学基金资助项目(40671013);国家自然科学基金重点基金资助项目(40331008)

摘　　要：MODIS是近年来较常用于宏观土地利用/土地覆盖的一种遥感数据来源。为了克服MODIS空间分辨率的限制,在建立提取模型的时候使用了多期遥感数据,并且加入了地形辅助数据。Logistic模型是一种非线性的回归分析模型,它通常是用于预测和判定未知单元的类别属性。先划出训练样区,计算出湿地分布与各因子之间的关系式,进而得到湿地分布概率图,选取一定的阈值(0.5),最终提取出湿地。研究结果显示,Logistic模型用于遥感分类,分类精度高于最大似然法。另外,加入了地形因子(dem,slope,tpi)后,分类精度得到明显的提高。MODIS is one source of remote sensing images which is often used in the study of regional land cover recently. In order to overcome the restrictions of limited spatial resolution, the abstracting models make use of multi-temporary data and topography data. Logistic model is a nonlinear regression analysis model. It is usually used to forecast and classify the unknown units. After lining out training areas and calculating the relationships between wetland distribution and all the factors, the wetland probability of each cell is gained. When the cut value is 0.5, the wetland distribution is abstracted. The results show that, logistic model is suit for remote sensing classification, especially after adding topographic factors （dem, slope, tpi）, and the classification accuracy is higher than that of MLC.

关 键 词：LOGISTIC模型 MODIS 湿地 遥感分类 地形因子 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]