航空影像农田类型分类在地理国情监测中的应用研究

航空影像的植被信息提取是遥感影像分类中的难点,仅利用光谱信息难以提取农田类型。本文以江苏农田为主要覆盖的典型区域为研究对象,选择航空影像利用随机森林算法提取不同的农田信息。本研究采用多尺度的分割方法,面向对象实现特征信息提取。根据光谱、纹理以及几何形状特性筛选出较为合适的特征作为参数,利用随机森林算法实现植被二级分类,分类精度达到84.60%,KAPPA系数为0.753,可为地理国情生产提供一定的参考。

Application of Aerial Image Classification in the Geographical Condition Monitoring

The extraction of vegetation information from aerial imagery is the difficulty in the classification of remote sensing images, using only spectral information is difficult to extract the type of farmland.In this paper, taking Jiangsu as a typical area of farmland covers the major as the research object, select the image using the random forest algorithm to extract the information of different farm-land.This study adopts multi -scale segmentation method, the feature information extraction of object oriented.According to the spectral, texture and shape features selected features more suitable as a parameter, the realization of the two class classification of veg-etation using random forest algorithm, the classification accuracy of 84.60%, KAPPA coefficient is 0.753, which can provide some reference for the geographical conditions of production.