极化干涉SAR森林冠层高反演的地形坡度改正

Terrain Slope Correction on Pol In SAR Forest Canopy Height Inversion

极化干涉SAR森林冠层高反演是当前SAR领域研究的热点。经典的森林冠层高反演算法主要基于随机地表二层相干散射模型(Random Volume over Ground,RVo G),该模型在山区受到植被层下地表的地形坡度影响,反演精度存在较大误差。为了提高森林冠层高反演精度,采用地形坡度改正的S-RVo G(Sloped Random Volume over Ground)模型,结合三阶段算法,应用德国宇航局DLR提供的星载Tan DEM-X全极化干涉数据反演森林冠层高,并对结果进行验证。结果表明:坡度级为II、III级,RVo G模型反演效果接近于S-RVo G模型;坡度级为IV级,RVo G模型与二调平均树高的相关关系明显下降,加权相对误差和RMSE增大;S-RVo G模型与二调平均树高保持显著相关关系,反演误差同比小于RVo G模型。因此,S-RVo G模型一定程度上改正了地形坡度造成的误差,提高了森林冠层高反演精度,在坡度大的地区精度提升程度更为明显。

In order to correct terrain distortion and improve the accuracy of forest canopy height inversion, we used S-RVoG （Sloped Random Volume over Ground） model which takes terrain slope into consideration, and employs three-stage algorithm to acquire forest canopy height. The validation was verified by spaceborne TanDEM-X quad-polarimetric and interferometric data. RVoG model behaved similarly to S-RVoG model in slope level II and IlL However, in slope level IV, RVoG model had a lower correlation coefficient with data from forest resource inventory and grown weighted relative error and RMSE. S-RVoG model obviously performed better than RVoG model not only in correlation coefficient but also in weighted relative error and RMSE. S-RVoG model could correct errors caused by terrain slope to some extent and improve the accuracy of forest canopy height inversion. Slope correction and accuracy improvement are more obvious in the area with high slope level.