大光斑波形数据在土地覆盖分类的适用性分析

Land cover classification of the loess plateau based on GLAS full waveform data

针对GLAS地学激光测高系统是冰、云和陆地高程卫星(ICESat)的唯一监测工具,能够记录地表光斑内的地物信息,是否能应用于黄土高原土地覆盖分类的问题进行了研究。利用粒子群和最小二乘法相结合的方法对GLAS波形数据进行高斯分解,获取高斯波个数、波形总能量、波形信号起始和信号结束位置4个波形参数;基于波形自动分类方法对黄土高原水体、森林、城市用地、其他地类(裸地、低矮植被等)进行分类。通过基于覆盖相同研究区域的30 m地表覆盖数据(Globe Land30),验证分类的准确性。结果表明,GLAS大光斑波形数据对黄土高原的4种地类能够很好地进行区分,总分类精度高达87.68%,Kappa系数为65.79%。研究表明,GLAS波形数据可以作为获取土地覆盖信息的有效数据源,为研究黄土高原土地覆盖变化提供更丰富的数据支持。

The GLAS geoscience laser altimeter system is the only monitoring tool for ice,cloud and terrestrial elevation satellites(ICESat).ICESat/GLAS large spot radar waveform data can record the ground object information in the surface.Therefore,the study the possibility of land cover classification using GLAS data in the Loess Plateau.Firstly,Gaussian decomposition of GLAS waveform data is performed by combining particle swarm optimization and least squares method.And,four waveform parameters of Gaussian wave number,total waveform energy,waveform signal start and signal end position are obtained.Secondly,the water bodies,forests,urban land and other land types(naked land,low vegetation,etc.)of the Loess Plateau were classified based on waveform automatic classification.Finally,the accuracy of the classification is verified by calculating the confusion matrix based on the 30 m surface coverage data(Globe Land30).The results of this study show that the GLAS waveform data can distinguish the four types of land in the Loess Plateau,with a total classification accuracy of 87.68% and a Kappa coefficient of 65.79%.This study shows that GLAS waveform data can be used as an effective data source for obtaining land cover information,providing more abundant data support for studying land cover change in the Loess Plateau.