基于遥感和陆表信息集成的广东省土地覆盖分类方法研究

Study on Land Cover/use Classification of Guang Dong Province by Integral Use of Remote Sensing Data and Field Survey

土地覆盖信息是估算地-气间的生物物理过程和能量交换的关键参数,也是区域和全球尺度气候和生态系统过程模型所需要的重要参量。如何高效地利用遥感数据提取土地覆盖信息是当前研究迫切需要解决的问题。面向对象的分类方法不但充分利用了遥感数据的光谱信息,同时也利用了影像的纹理结构信息和更多的地物分布信息关系,在遥感分类中具有较大的潜力。研究基于2010年多时相的环境卫星数据、TM数据以及DEM数据,并结合地表采集的4000多个样点数据,采用面向对象的分类方法对广东省土地覆盖进行分类。经采样验证,广东省土地覆盖平均精度为85%,分类结果精度远高于常规的分类算法,说明结合陆表信息的面向对象分类方法比常规的分类算法更具有优势,可以实现高精度的土地覆盖分类。

Land cover/use information is a key parameter for estimating biophysical process and energy exchange between land and atmosphere,and meanwhile it is also an important parameter for regional and global climate models and ecosystem process models.Remote Sensing is an important and effective tool for mapping land cover /use on a large scale.But the quality of remote sensing data and weather condition have great impacts on the accuracy of land cover /use information from remote sensing data.This research proposes a new methodology for mapping land cover /use of Guang Dong province using object-oriented classification with remote sensing data and field survey information as inputs.Object-oriented classification has been increasingly employed for mapping land cover/use because of its advantages of using multi information of remote sensed objects such as spatial distribution,shape,size,spectral and texture over some traditional classification methods which often only use spectral information of objects.The remote sensing data used in this research include multitemporal and multi spectral China HJ-1A/B and TM data.Firstly multi temporal remote sensing data were processed and geo-rectified.The classification rules were then established using object-oriented classification method with processed remote sensing data,DEM,history land cover/use data and field observation data as inputs.Finally a new classification method is developed by integrating multi information from remote sensing data and field observations into classification rules.In situ observation data is used to validate classification result.The result shows that the average accuracy of mapping land cover /use of Guangdong province is 85%,much higher than that of traditional classification method.The accuracy can be further improved by using more effective classification rules and more field survey in future.