面向对象的库区高分一号影像土地利用信息提取

高分一号（GF-1）是我国自主研发的第一颗高分辨率遥感卫星，其包含地物信息较为丰富，已应用于土地利用信息提取，但在水利工程库区土地利用调查方面研究较少。本文以峡江水利枢纽工程库区为例，首先对库区影像进行了基于RPC模型的正射校正、几何精纠正等预处理；然后针对GF-1的传感器响应特性，采用基于多元线性波段拟合的方法对多光谱与全色影像进行融合，该方法相对于传统分量替换法具有更好的融合性能；最后综合利用影像的光谱、纹理及形状等特征，采用面向对象的方法对融合后的库区影像进行了地类信息提取与分类精度评价。试验结果表明，融合影像可以有效提取水利工程库区的土地利用信息，总体分类精度达到87.9%，Kappa系数为0.836，能够满足库区土地利用调查和变化监测的要求。

Land-use information extraction of GF-1 remote sensing image of reservoir area by using object-oriented method

GF-1 remote sensing satellite is the first domestic high-resolution satellite. Due to abundant information included in the images, more and more GF-1 data have been used to extract the information of land-use. However, few application is found in the land-use investigation of reservoir area of hydraulic project. Firstly, take reservoir area of Xiajiang hydraulic project in Jiangxi Province as an example, the images are preprocessed including ortho rectification using RPC model, geometric precision correction and so on. Then, band simulation based pansharpening algorithm by multiple linear regression is used to fuse 8 m multispectral and 2 m panchromatic images, the performance of this method is better than the conventional fused method. Finally, land use information of reservoir area from 2 m fused image by using object-oriented method based on comprehensive utilization of spectral information, texture, shape characteristics and so on. The results show that land-use information could be extracted from GF-1 images of reservoir area effectively; the overall classification accuracy reaches 87.9%, the Kappa index is 0.836, and can meet the requirement of land-use investigation and change motoring in reservoir area.