Traditional bathymetry methods have the drawbacks of small coverage, long duration and high energy consumption, while the bathymetry inversion around islands by integrated active and passive remote sensing can not only overcome the weakness of traditional methods, but also provide fundamental data for shipping safety, marine disaster reduction and marine eco-environmental protection. Taking the Oahu Island of Hawaii as an example, this paper studies the impacts of different point-density LiDAR data and different gridding approaches on the inversion accuracy, as well as analyzes the ability of large-scale water depth inversion using a few LiDAR controlled blocks based on the multi-spectral images of Landsat-8 and LiDAR bathymetry data. The following results have been obtained. (1) The inversion accuracy is not severely affected by the change of point-density of LiDAR data, with the difference of mean relative error lower than 0.3% and mean absolute error less than 0.03 m. (2) The bathymetry inversion accuracy with the equalization method is slightly higher than that with the median method, which is proved by the fact that the mean absolute error of the equalized value decreases by 0.04-0.05 m as against that of the median value, with the mean relative error lowering by 1%-10%. Besides, the residual distribution of inversion results shows that the equalization approach has a more concentrated residual and its mean value is close to 0 m in the depth ranges of 0-2 m and 20-25 m, while both approaches show a basically same distribution trend in other depth ranges. (3) Bathymetry inversion based on a few LiDAR controlled blocks achieves relatively ideal results. The R 2, mean absolute error and mean relative error of 2 check blocks are 0.877, 1.66 m, 23.5% and 0.941, 1.62 m, 28.4%, respectively. Analysis indicates that the inversion precisions in different depth ranges are satisfactory in general. Except the 20-25 m range, almost all the mean absolute errors are below 2.5 m; and only the ranges of 0-2 m and 2-5 m have the mean absolute error of beyond 25%.%主被动遥感结合反演远海岛礁周边水深信息,不仅可以有效弥补传统测深方法覆盖范围小且费时费力的不足,也可为航运安全、海洋减灾、生态环境保护等领域提供基础资料。以夏威夷瓦胡岛周边水深反演为例,应用Landsat-8多光谱遥感数据和机载LiDAR测深数据,开展了不同密度LiDAR测深数据对水深多光谱遥感反演精度的影响分析、不同水深网格化处理方法对水深遥感反演结果的影响分析和基于少量LiDAR控制区块的大区域水深反演能力分析三方面的研究工作。结果表明：(1) LiDAR测深数据密度的改变对水深反演结果的影响不大,变化后的水深反演结果与原始的水深反演结果相比,平均相对误差变化在0.3%以内,平均绝对误差变化在0.03 m以内；(2)采用均值格网处理方法的多光谱遥感水深反演精度要略高于采用中值格网处理方法的水深反演精度,具体体现在均值的平均绝对误差要比中值的低0.04~0.05 m,平均相对误差低1%~10%,反演结果的残差分布显示在0~2 m和20~25 m的水深段内均值统计法的残差分布更集中且其平均值接近于0 m,而在其它水深段二者的残差分布基本相同；(3)基于少量LiDAR控制区块的大区域遥感水深反演结果较为理想,两个检查区块的水深反演结果R2、平均绝对误差和平均相对误差分别为：0.877,1.66 m,3.5%和0.941,1.62 m,28.4%。反演结果分段分析表明各水深段内反演的精度都比较理想,平均绝对误差除20~25 m水深段外,均低于2.5 m,平均相对误差除0~2 m,2~5 m外,均低于25%。
展开▼
