基于地相位优化估计的RVoG三阶段森林冠层高度反演
作者:
作者单位:

作者简介:

通讯作者:

中图分类号:

基金项目:

国家自然科学基金项目(42061072)、云南省科技厅重大科技专项(202002AA100007-015)和云南省教育厅科学研究基金项目(2022Y579)


Forest Canopy Height Inversion in RVoG Three-stage Based on Optimal Estimation of Ground Phase
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    极化干涉合成孔径雷达(PolInSAR)估测森林结构参数中,数据受基线长度、信噪比、环境地形以及雷达波长的影响,尤其在复杂森林环境条件下,会导致观测到的复相干存在误差,从而影响最终的反演结果。为解决此问题,首先探讨了体相干选择对RVoG三阶段森林冠层高度反演的影响,以地相位为参考逐像素选择距离地相位最远的相干性作为体相干。其次改进了地相位估计方法,采用戴明回归(DMR)和正交回归(OGR)2种相干直线拟合方法来改进地相位的估计,并在DMR拟合方法中设置了不同的误差比(0.3和0.6)来比较地相位估计方法对RVoG三阶段森林冠层高度反演的影响。研究结果表明:以地相位为参考逐像素选择体相干的反演结果相较于直接使用HV极化通道的复相干γHV为体相干的反演精度有明显提升,决定系数(R2)由0.349增加到0.383,均方误差由7.097m2降低到5.755m2。在体相干优化选择的基础上,采用了戴明回归和正交回归对地相位估计方法进行了改进。表明基于最小二乘回归(LSR)地相位估计的RVoG三阶段反演精度最低,采用DMR和OGR进行相干线拟合的反演精度相较于LSR均有一定提升,所有反演结果的决定系数(R2)均在0.440左右,均方误差(MSE)均降低了2m2左右。研究结果说明采用RVoG三阶段方法反演森林冠层高度时,在复相干存在误差的情况下,用传统最小二乘回归(LSR)估计地相位进行高度反演会对结果带来一定误差,通过其他相干直线拟合方法来克服复相干误差的影响能改善最终的森林冠层高度反演结果,以地相位为参考选择体相干的反演方法也更为合理。

    Abstract:

    In polarimetric interferometric synthetic aperture ray (PolInSAR) forest structure parameter estimation, the data are affected by the baseline length size, signal-to-noise ratio, environmental topography, and radar wavelength, especially under complex forest environmental conditions, which can lead to errors in the observed complex coherence and thus affect the final inversion results. Firstly the effect of volume coherence selection on the RVoG three-stage forest canopy height inversion was explored, and the coherence farthest from the ground phase was selected as the volume coherence with the ground phase as the reference pixel by pixel. Secondly, the ground phase estimation method was improved by using two coherence linear fitting methods, Deming regression (DMR) and orthogonal regression (OGR), to improve the estimation of the ground phase, and different error ratios (0.3 and 0.6) were set in the DMR fitting method to compare the effects of the ground phase estimation method on the RVoG three-stage forest canopy height inversion. The results showed that the inversion accuracy of the inversion of volume coherence with ground phase as the reference pixel-by-pixel selection was improved compared with that of the complex coherence with the HV polarization channel directly. The coefficient of determination(R2)was increased from 0.349 to 0.383, and the mean square error(MSE) was decreased from 7.097m2 to 5.755m2. Based on the optimal selection of the volume coherence, the ground phase estimation method was improved by using Deming regression and orthogonal regression. It was shown that the least squares regression (LSR)-based ground phase estimation had the lowest accuracy of RVoG three-stage inversion, using DMR and OGR for coherence line fitting had a certain improvement in inversion accuracy compared with LSR, and the coefficient of determination (R2) of all inversion results was around 0.440, and all MSE was reduced by about 2m2. The conclusions indicated that the forest canopy height inversion using the RVoG three-phase method introduced some errors in the height inversion by using the traditional LSR estimation of the ground phase in the presence of errors in the complex coherence. Using other coherence linear fitting methods to overcome the influence of the complex coherence error can improve the final forest canopy height inversion results, and it was also more reasonable to choose volume coherence inversion method with the ground phase as the reference.

    参考文献
    相似文献
    引证文献
引用本文

罗洪斌,朱泊东,岳彩荣,杨文俊,龙飞,徐婉婷.基于地相位优化估计的RVoG三阶段森林冠层高度反演[J].农业机械学报,2022,53(7):301-307. LUO Hongbin, ZHU Bodong, YUE Cairong, YANG Wenjun, LONG Fei, XU Wanting. Forest Canopy Height Inversion in RVoG Three-stage Based on Optimal Estimation of Ground Phase[J]. Transactions of the Chinese Society for Agricultural Machinery,2022,53(7):301-307.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-12-08
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2022-07-10
  • 出版日期: