Research On The Forest Height Retrieval Based On Polarimetric And Interferometric Synthetic Aperture Radar

Polarimetric Interferometric Synthetic Aperture Radar(PolInSAR),a combination of the PolSAR and InSAR techniques,has the characteristics of InSAR being more sensitive to the vertical information of ground objects and the attributes of PolSAR being sensitive to the geometric shape and dielectric constant of ground objects.It is an essential technique for the inversion of forest parameters.The tree height estimation algorithm and model based on the PolInSAR technology have been studied in this study.First,we study the algorithm based on the dual-pol SAR data.The basic principles of the coherent phase difference method and the coherent phase-amplitude joint algorithm are analyzed.A new vegetation model was then established based on the ALOS/PALSAR datasets and derived a tree height using the two algorithms with the dual-pol SAR data input.The average of the inverted values and the mean of the measured 37 values at sampling points were calculated.The average tree heights were 9.7 m and 20.9 m for the phase difference and coherent phase-amplitude joint algorithms compared to the mean of 24.4 m from the measured data.The latter algorithm should potentially be one possible approach to estimate tree heights at local,regional,and global scales because of the improved performance.The volume of the dual-pol interferometric SAR data is currently large.However,more and more quad-pol interferometric data will be available in the future with the continuous development of SAR technology.The quad-pol data obtain additional ground feature information.Therefore,the tree height retrieval algorithm based on the quad-pol data is also the focus of the research.The volume scattering and double scattering phase centers are derived with the polarimetric interferometric matrix and Freeman-Durden decomposition algorithm.The former is selected to represent the canopy phase center,and the latter is the phase center at the trunk-ground intersection.The tree height is then estimated based on the phase difference between the two centers and effective two-dimensional vertical wavenumber.Unfortunately,the algorithm underestimated the tree height,as shown later.Then,the three-stage tree height inversion algorithm based on the Random Volume over Ground Model(RVoG)was considered.The algorithm overestimated the tree height,though.Thus,the optimal coherence coefficients and phase diversity coherence coefficients were introduced to improve the fitting process of the straight line of the three-stage algorithm.With a simulated mean height of 14 m for conifer trees,errors of the average values of the three algorithms were –38.6%,12.1%,and –2.1%,respectively.A simulation for deciduous trees at a mean of 14 m was carried out as well.Errors from the three algorithms were–41.4%,10%,and –7.1% when the mean values were compared.For both coniferous and deciduous trees,the optimized three-stage algorithm has the highest inversion accuracy.