Abstract:

Forest canopy height is an important parameter.In this paper,the phase-coherence amplitude method as the forest height inversion model based on airborne multi-baseline PolInSAR data acquired in the Gabonese Republic on the west coast of Central Africa,the RVoG three-stage method is used to calculate the forest canopy height in the phase difference part,and compared the differences of forest canopy height inversion results under four baseline selection methods of PROD,LINE,ECC and VAR.The results show that among the inversion results with four different baseline selections,the best inversion results are obtained using the LINE method for baseline selection with R² of 0.746,RMSE with 7.715m,and BIAS with -3.412m,which are more satisfactory.The inversion accuracy of baseline selection using the PROD and ECC methods is slightly lower than that of the LINE method,and the difference between them is not significant.The R² of the baseline selection inversion results using the ECC method is 0.732,RMSE is 7.918m,and BIAS is 0.139m,and the underestimation and overestimation in the inversion results are more obvious.The inversion results of baseline selection using the PROD method have an R² of 0.720,RMSE of 8.099m,and BIAS of -4.056m,showing a slight overestimation overall.The inversion results of baseline selection by VAR method had the lowest accuracy among all methods,with R² of -0.245,RMSE of 17.079m,and BIAS of 12.945m,and the overall underestimation was more serious,and serious underestimation occurred after the forest canopy height was greater than 10m.The findings suggest that multi-baseline PolInSAR can accurately and efficiently invert forest canopy height,but the baseline selection method has a significant effect on the results.

Key words: Phase-coherence amplitude method, RVoG three-stages method, PolInSAR, multiple baselines, baseline selection