森林植被碳储量的空间分布格局及其动态变化是陆地生态系统碳收支核算的基础。作为森林地上生物量的重要指示因子，森林高度的精确估算是提高森林植被碳储量估算精度的关键。现有研究已证明，由专业星载摄影测量系统获取的立体观测数据可用于森林高度提取，但光学遥感数据最大的问题是受云雨等天气因素的影响严重。区域森林地上生物量产品的生产需要充分挖掘潜在数据源。国产高分二号卫星(GF-2)虽然不是为获取立体观测数据而设计的专业星载摄影测量系统，但其获取的图像空间分辨率可达0.8 m，且具备±35°的的侧摆能力，在重复观测区域可构成异轨立体观测。本文以分别获取于2015年6月20日和2016年7月19的GF-2数据作为立体像对，其标称轨道侧摆角分别为0.00118°和20.4984°，以激光雷达数据获取的林下地形(DEM)和森林高度(CHM)为参考，对利用GF-2立体观测数据进行森林高度提取进行了研究。通过对立体处理得到的摄影测量点云的栅格化得到DSM，以激光雷达数据提供的DEM作为林下地形，得到了GF-2的CHM。结果表明GF-2提取的CHM与激光雷达CHM空间分布格局较为一致，两者之间存在明显的相关性，像素对像素的线性相关性(R²)达到0.51，均方根误差(RMSE)为3.6 m。研究结果表明，在林下地形已知的情况下，GF-2立体观测数据可用于森林高度估算。

A map of carbon storage and its dynamics of forest vegetation comprise the basic dataset for the calculation of the carbon balance of terrestrial ecosystems. Accurate estimation of forest canopy height, which is an important indicator of forest aboveground biomass, is essential for the mapping of carbon storage of forest vegetation carbon. Existing research has indicated that stereo images acquired by professional spaceborne or airborne stereo systems can be used to extract forest canopy height. However, the acquisition of optical photogrammetric images is hindered by cloud coverage and rains. Regional mapping of forest aboveground biomass should be based on the synergy of multiple sources of optical stereo images. Although the Chinese GF-2 is not a satellite specially designed for professional spaceborne photogrammetric systems, GF-2 theoretically possesses a stereo mapping capability because of its high spatial resolution of 0.8 m and rolling capability of ±35°. This study aims to explore the practical feasibility of extracting forest canopy height by using GF-2 stereo imagery.
Two scenes of GF-2 images acquired on June 20, 2015, and July 19, 2016, are processed by a commercial stereoscopic processing software to generate the point cloud. A Digital Surface Model (DSM) is derived through the rasterization of the GF-2 point cloud. The canopy height model (CHM) is extracted from GF-2 DSM subtracted by the digital elevation model from LiDAR data.
GF-2 CHM and LiDAR CHM are compared. GF-2 CHM exhibits approximately the same spatial pattern as LiDAR CHM. Pixel-by-pixel linear regression shows that GF-2 CHM is highly correlated with LiDAR CHM, with R²=0.51 and root-mean-square error=3.6 m.
Stereo imagery acquired by the Chinese GF-2 can be used to describe the vertical structure of forest canopy top. Forest canopy height can be extracted from GF-2 stereo imagery if an understory elevation of the ground surface is available.