基于GF-2多光谱数据的水深多模型探测比较研究---以香港平洲岛为例

水深是浅海重要的地形要素，利用遥感手段探测水深具有经济、高效等优势。本文利用高分二号多光谱遥感数据，采用不同波段组合的对数线性模型、Stumpf对数转换比值模型和改进的对数转换比值模型，以香港平洲岛为研究区域进行水深反演，并开展精度评价。结果表明：对数线性模型、Stumpf对数转换比值模型和改进的对数转换比值模型中，B1、B2、B3和B4（蓝、绿、红和近红外）4波段组合的对数线性模型水深反演精度最高，其检查点平均绝对误差MAE为1.63m，平均相对误差MRE为12.67%，决定系数R2达0.80；0-5m、5-10m、10-15m、15-20m分水深段分析发现，3种模型在10-15m水深段的水深反演效果均较好，最小的平均绝对误差MAE和平均相对误差MRE分别为1.09米和8.99%，10m以浅和15m以深的反演误差较10-15m的较大，显示上述模型更适合于中等浅水区域的应用。

英文摘要:

Water depth is an important topographic factor in shallow water. There have advantages such as economy and efficiency to detect water depth by using remote sensing. In this paper, GF-2 multi-spectral remote sensing data are used, and log-linear model with different band combinations, Stumpf log-transport ratio model and improved log-transform ratio model are applied to detect the water depth around Pingzhou Island in Hong Kong. The results show that the logarithmic linear model of B1, B2, B3 and B4 (blue, green, red and near-infrared) achieves the highest inversion accuracy. The mean absolute error (MAE) of the checkpoint is 1.63m, the mean relative error (MRE) is 12.67%, and the coefficient of determination R2 is 0.80. The regression fitting effect of 4 bands combination is better than that of the others, and the fitting goodness is higher. The three models is used to analyze the inversion accuracy in different water depths range of 0-5m, 5-10m, 10-15m and 15-20m, and it is found that the water depth inversion result in 10-15m water depth is the best with the mean absolute error of 1.09 meters and the mean relative error of 8.99%. The inversion error of 10m in shallow and 15m deep is worse than 10-15m.It indicating that the above model is more suitable for medium shallow water applications.

查看/发表评论下载PDF阅读器