摘要
森林树高的反演,一直是极化干涉SAR(Polari metric SAR Interferometry)研究领域中的热点。介绍了近年来分别基于相干优化算法和高分辨率波达算法(ESPRIT)的树高反演方法,并尝试了ESPRIT算法和三阶段反演算法的结合。最后利用德国宇航局(DLR)的全极化干涉L波段ESAR数据及其相应的林地平均树高实测数据,对所述方法进行了验证与分析。结果表明:三阶段算法和三阶段Sinc算法相对于其它算法而言具有较好的结果,但同其它基于散射模型的树高反演算法一样,存在着一定程度的高估;而ESPRIT相位差值算法则严重地低估了平均树高;尽管ES-PRIT算法与三阶段算法、三阶段Sinc算法相结合的尝试可在一定程度上提高计算效率,但其树高反演结果并没有得到明显改善。
The objective of this paper is to analyze the forest height retrieval methods from polarimetric SAR interferometry and their validation against ground truth.Both of the methods based on coherence optimization process and that on super-resolution method such as ESPRIT are generalized,analyzed,and compared.Also in this paper hybrid methods of ESPRIT and three-stages/three-stage Sinc are proposed and validated.The DLR quad-polarimetric interferometry ESAR data is used for forest height generation.The validation with ground truth shows that the three-stage and three-stage Sinc method have a relative better performance than all other methods despite with a overestimation of height but with a very good correlation with ground truth.In contrast,as in expectation,the direct ESPRIT phase differencing methods,due to their virtual height results,severally underestimates the forest heights but also with a poor correlation with ground truth.The hybrid methods of ESPRIT and three-stages/three-stage Sinc method,even a fast processing gained,lead to no improved results.
出处
《遥感技术与应用》
CSCD
北大核心
2009年第5期611-616,共6页
Remote Sensing Technology and Application
基金
国家自然科学基金重点课题“多维度微波成像的陆地遥感应用研究”(60890074)项目资助
