格陵兰以北海域卫星激光测高的回波模型与地物分类

The accurate changing of mean sea level and sea ice is the foundation of developing marine science research in the polar regions, and precise classification of water and sea ice data is essential precondition of monitoring. In recent years, owing to many advantages such as smaller footprint, space-borne laser altimeter has acquired a great of achievements in the field of polar science, and has a promising application prospect. This project proposes that through establishing the laser pulse echo theoretical models including water, sea ice, and snow-covered ice, for different surface categories the theoretical ranges of waveform parameters, e.g. width, kurtosis, etc, are calculated as the primary basis of classification. The statistical characters of GLAS waveform parameters, which are extracted by the designed waveform processing algorithm, are calculated as the auxiliary basis of classification. Using airborne Lidar data and NCEP meteorological data as prior information of earth surface, a new supervised surface category method will be designed and trained to complete the three kinds of surface type classification in the north waters of Greenland. The sea ice freeboard will be retrieved on the north of Greenland using the classified GLAS elevation data, which will be contrasted with the results calculated by airborne Lidar data at the same time and region, in order to verify the precision of classification. This research will form a new method of surface category based on analytical model and semi quantitative, and compared to the current method based on statistical character, the precision of classification will be significantly improved. Besides, the established echo theory and waveform simulator are meaningful for system parameter design of space-borne laser altimeter used for polar region monitoring.

准确监测极地平均海面与海冰变化，是极地海洋科学研究的基础，而正确分类海水与海冰，又是监测工作的前提。近年来，具有光斑直径小等优点的卫星激光测高在极地研究中效果显著，具有良好的应用前景。项目拟通过建立极地海域海水、海冰、覆盖积雪的海冰三种地物的激光脉冲回波模型，计算不同地物的回波宽度、峰度等参数的理论范围作为首要分类依据；设计波形处理算法提取GLAS实测回波参数，以其统计特性作为辅助分类依据；使用机载Lidar和NCEP气象数据作为地表先验信息，以格陵兰以北海域为例，训练监督分类方法实现三种地物分类。通过分类后的GLAS数据计算该区域海冰干舷，与同时同地的机载Lidar计算结果进行对比，验证分类的正确性。项目将分类特征参数模型化和半定量化，其对极地海域地物分类精度，比目前基于统计特性的激光测高地物分类方法将有明显改善；所建立的回波模型对我国用于极地测量星载激光测高仪的系统参数设计有参考意义。

准确监测极地平均海面与海冰变化，是极地海洋科学研究的基础，而正确分类海水与海冰，又是监测工作的前提。近年来，具有光斑直径小等优点的卫星激光测高在极地研究中效果显著，具有良好的应用前景。在现有的激光测高数据分类方法研究中，传统方法是利用激光回波的波形参数统计信息（例如使用目标高度、反射率、波形宽度、振幅、峰度和斜度等）进行机器学习或者设置阈值的分类方式。项目研究过程建立了星载激光测高仪在海冰、海水目标的回波模型和波形仿真器，对波形采用逐点时间差距离加权计算理论海水波形和实测回波的总振幅差异值，建立一种半解析型的海水、海冰分类方法；同时，与传统方法类似，使用基于支持向量机的分类方法，输入回波波形的多种特征参数进行监督分类，最终将半解析分类方法和SVM方法进行取长补短，形成一种新的分类方法。通过机载Lidar在格陵兰北部海冰区的实测点云作为地面类型先验真值，对美国GLAS激光测高仪在该区域实测波形进行基于该方法的分类准确性验证，半解析型分类OA精度95.62%，优于传统的支持向量机分类器的90.44%，两种分类器融合后OA精度98.21%，Kappa系统0.96。项目使得星载激光测高仪回波的地物类型分类方法由目前的基于数值分析为依据向理论解析模型为依据的分类方向延伸，其对极地海域地物分类精度，比目前单纯基于统计特性的激光测高地物分类方法将有了明显改善，且具有更好的物理意义；研究过程中所建立的回波模型，对我国用于极地测量星载激光测高仪的系统参数设计也有参考意义。