Name
Abstract | ||
|---|---|---|
Polarimetric SAR Interferometry (PolInSAR) has been widely applied in tree height retrieval. Many inversion methods using PolInSAR data was investigated. There are mainly three class of retrieval algorithms, which are respectively algorithms based on Polarimetric Interferometric Phase Coherence Optimization(1, 2), algorithms based on ESPRIT(3), and algorithms based on the Maximum Likelihood Estimation(4, 5). Among them, the three-stage inversion algorithm which is based on Polarimetric-Interferometric Phase Coherence Optimization proposed by S. R. Cloude and K. P. Papathanassiou is quite simple and most widely used. The method breaks the tree height inversion process into three steps. First, we make use of the estimated coherence value of many polarization channels to fit the coherence line. Second, vegetation bias is removed, and during this step we can get the estimation of the ground phase. Finally, tree height is estimated. Their method is based on the assumption that the ground-to -volume scattering ratio in one of the polarization channel is 0. In order to simplify the problem, we usually make an assumption that the ground-to-volume scattering ratio in the HV channel is 0. That is, the estimated volume coherence coefficient is obtained from the estimated coherence of HV channel. However, this is usually not true, and there is an ambiguity in the estimation of volume coherence. In consequence, the inversion results of the tree height contain some error. Here we propose a refined algorithm of three-stage inversion process to resolve the ambiguity problem. In the proposed algorithm, we focus on resolving the ambiguity estimation of the volume coherence and obtain a more accurate estimation of it. We make an assumption that volume coherence lies in the coherence region and the distance between the volume coherence point and the ground phase point is the longest among all the coherence points. According to the physical explanation of the scattering model of the forest area, we can conclude that this assumption is valid. Based on this assumption, we make use of the region extraction algorithm(6) to obtain the boundary point in the coherence line and choose the point which is the most distant from the ground phase point as the volume coherence point. Since the coherence points lie in a line, we can extract the boundary of the coherence region by searching in the whole polarization space for the extremum of the real part of the coherence. The two scattering vectors corresponding to the extremum will be the candidate volume scattering channel. An accurate estimation of volume coherence can be obtained through this process. In order to confirm the validity of the proposed algorithm, we use the simulated vector coherent SAR data for a random canopy simulated by the software PolSARpro provided by ESA. The data is free from temporal decorrelation, motion or co-registration errors, and SNR effects. We respectively use the traditional three-stage inversion algorithm and the refined algorithm on the simulated data to invert the tree heights. Fig. 1 shows the simulated SAR image of the canopy height of 18m. Fig. 2 shows the retrieval results of marked line in Fig. 1 using 3-stage method and the refined method respectively. Table 1 demonstrates the quantitative results. Experiment results show that the proposed algorithm can provide more accurate tree height estimation. This paper contains 5 parts. Section I is the introduction. The coherent scattering model is introduced in Section II, and in Section III we focus on deriving the modified algorithm. The experiment results obtained by retrieving the simulated data is listed in Section IV. The last part is the conclusion. |
Year | DOI | Venue |
|---|---|---|
2008 | 10.1109/IGARSS.2008.4780150 | IGARSS |
Keywords | Field | DocType |
estimation,feature extraction,inverse problems,polarimetry,interferometry,coherence,scattering,surfaces,correlation,shape,data mining,esa,maximum likelihood estimate,forest,decorrelation,scattering parameters,vegetation | Decorrelation,Computer science,Inversion (meteorology),Remote sensing,Interferometry,Coherence (physics),Feature extraction,Inverse problem,Polarimetric sar,Scattering | Conference |
Volume | Issue | ISSN |
5 | 1 | null |
ISBN | Citations | PageRank |
978-1-4244-2808-3 | 0 | 0.34 |
References | Authors | |
2 | 2 | |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Lulu Tan | 1 | 0 | 1.01 |
| Ruliang Yang | 2 | 12 | 4.62 |