基于BP神经网络和MODIS-EVI时间序列的土地覆被分类

以matlab为平台设计了BP神经网络用于自动提取土地覆被信息,将覆盖内蒙古的2006年全年的MO- DIS数据处理成EVI植被指数时间序列集输入神经网络进行土地覆被分类,精度检验结果显示:总体分类精度(o- verall-accuracy)为78.32%,kappa系数为0.73。

BP neural network usedin classfication of land cover

This article has regarded matlab as the platform and designed BP neural network procedure,thus used for drawing surface feature,information automatically from the remote sensing image,that this kind of method not only can be applied to the high-resolution re- mote sensing image of the micro area,can also apply to the low resolution ratio remote sensing image of the macrnscopical area has been proved in the experiment.This article carry on the test for MODIS images with 1000 meters resolution and ETM images with 15 meter resolu- tion,the inspection result of the precision shows,the overall-accuracy is higher than 70%,kappa coefficient is higher than 70%.While classifying MODIS image,have dealt with MODIS image in the whole year of 2006 into the array collection of exponential time from EVI veg- etation index,uesing the exponential time on land use classify is the front where recent land utilizes the categorised method to study.