DOI number:10.3390/f13071031
Journal:Forests
Key Words:random volume over ground (RVoG); complex unitary circle (CUC); coherence region; estimation bias; finite multilooking effect
Abstract:In polarimetric interferometric SAR (PolInSAR) technology, the random volume over ground (RVoG) model establishes the mapping relationship between polarimetric complex coherence and forest biophysical parameters (e.g., forest height). However, due to speckle noise and the finite multilooking effect, the real observed coherence region in the complex unitary circle (CUC) is an ellipse, which is biased by the ideal noise-free coherence region represented as a straight line by the RVoG model. Multilooking processing can reduce speckle noise at the cost of resolution loss. Therefore, this paper analyzes the influence of different multilooking sizes on forest height inversion. Experimental results show that the accuracy of forest height inversion first increases and then decreases with the increase in multilooking size, which means there exists an optimal size for PolInSAR forest estimation. From statistical analysis of the forest height estimation error, inversion accuracy mainly depends on estimation bias rather than estimation variance. This is mainly because, in a homogeneous forest area, a large multilooking size helps to reduce the statistical bias effect; in the textured area, the inversion accuracy benefits from a small multilooking size for avoiding the mixing of multiple types of ground targets.
Co-author:Hu Chihao
Indexed by:Journal paper
Correspondence Author:Shen Peng
Discipline:Engineering
Document Type:J
Volume:13
Issue:7
Page Number:1031
Translation or Not:no
Date of Publication:2022-06-30
Included Journals:SCI
Links to published journals:https://www.mdpi.com/1999-4907/13/7/1031
Date of Publication:2022-06-30