基于WorldView-2数据和随机森林算法的遥感水深反演
|
摘要
鉴于传统水深反演线性回归模型易受水质和环境因素的影响,利用甘泉岛区域的高分辨率WorldView-2遥感影像,结合相应的机载LiDAR实测水深数据,使用随机森林算法构建了浅海水深反演非线性回归模型。以反演的水深值和实测水深值的相关系数(R~2)和均方根误差(RMSE)为指标,并同传统的水深反演单波段线性回归模型、双波段比值线性回归模型以及多波段组合线性回归模型进行比较。结果表明,随机森林水深反演非线性回归模型反演精度最优,R~2和RMSE分别为0.967和0.868m。
Traditional water depth inversion linear regression models are susceptible to water quality and environmental factors.This paper uses the high resolution WorldView2 remote sensing image in the Ganquan island region and the corresponding measured water depth data by airborne LiDAR.The random forest algorithm is used to construct the shallow regression model of shallow water depth.The random forest algorithm is compared with three classic water depth inversion models,namely single-band linear regression model,two-band ratio model and multi-band model.Correlation coefficient(R~2)and root mean square error(RMSE)are used to evaluate bathymetry accuracy.The results show that the inversion accuracy of the random forest regression model is optimal,with R~2 and RMSE are 0.967 and 0.868 m,respectively.
Traditional water depth inversion linear regression models are susceptible to water quality and environmental factors.This paper uses the high resolution WorldView2 remote sensing image in the Ganquan island region and the corresponding measured water depth data by airborne LiDAR.The random forest algorithm is used to construct the shallow regression model of shallow water depth.The random forest algorithm is compared with three classic water depth inversion models,namely single-band linear regression model,two-band ratio model and multi-band model.Correlation coefficient(R~2)and root mean square error(RMSE)are used to evaluate bathymetry accuracy.The results show that the inversion accuracy of the random forest regression model is optimal,with R~2 and RMSE are 0.967 and 0.868 m,respectively.
引文
[1]王晶晶,田庆久.基于TM遥感图像的近海岸带水深反演研究[J].遥感信息,2006,21(6):27-30.
[2]李秀瑞,朱金山,孙林.WorldView-2影像在南海岛礁浅海水深反演中的应用[J].遥感信息,2016,31(5):114-121.
[3]许海蓬,马毅,梁建,等.基于半经验模型的水深反演及不同水深范围的误差分析[J].海岸工程,2014,33(1):19-25.
[4]LYZENGA D R.Passive remote sensing techniques for mapping water depth and bottom features[J].Applied Optics,1978,17(3):379.
[5]STUMPF R P,HOLDERIED K,SINCLAIR M.Determination of water depth with high-resolution satellite imagery over variable bottom types[J].Limnology &Oceanography,2003,48(1):547-556.
[6]SU H,LIU H,WILLIAM D,et al.Automated derivation of bathymetric information from multi-spectral satellite imagery using a non-linear inversion model[J].Marine Geodesy,2008,31(4):281-298.
[7]BRAMANTE J,RAJU D K,SIN T M.Multispectral derivation of bathymetry in Singapore’s shallow,turbid waters[J].Machine Learning,2001,45(1):5-32.
[8]BREIMAN L.Using iterated bagging to debias regressions[J].Machine Learning,2001,45(3):261-277.
[9]李旭青,刘湘南,刘美玲,等.水稻冠层氮素含量光谱反演的随机森林算法及区域应用[J].遥感学报,2014,18(4):923-945.
[10]王丽爱,周旭东,朱新开,等.基于HJ-CCD数据和随机森林算法的小麦叶面积指数反演[J].农业工程学报,2016,32(3):149-154.
[11]刘艳丽.随机森林综述[D].天津:南开大学,2008.
[12]LYZENGA D R,MALINAS N P,TANIS F J.Multispectral bathymetry using a simple physically based algorithm[J].IEEE Transactions on Geoscience &Remote Sensing,2006,44(8):2251-2259.
[2]李秀瑞,朱金山,孙林.WorldView-2影像在南海岛礁浅海水深反演中的应用[J].遥感信息,2016,31(5):114-121.
[3]许海蓬,马毅,梁建,等.基于半经验模型的水深反演及不同水深范围的误差分析[J].海岸工程,2014,33(1):19-25.
[4]LYZENGA D R.Passive remote sensing techniques for mapping water depth and bottom features[J].Applied Optics,1978,17(3):379.
[5]STUMPF R P,HOLDERIED K,SINCLAIR M.Determination of water depth with high-resolution satellite imagery over variable bottom types[J].Limnology &Oceanography,2003,48(1):547-556.
[6]SU H,LIU H,WILLIAM D,et al.Automated derivation of bathymetric information from multi-spectral satellite imagery using a non-linear inversion model[J].Marine Geodesy,2008,31(4):281-298.
[7]BRAMANTE J,RAJU D K,SIN T M.Multispectral derivation of bathymetry in Singapore’s shallow,turbid waters[J].Machine Learning,2001,45(1):5-32.
[8]BREIMAN L.Using iterated bagging to debias regressions[J].Machine Learning,2001,45(3):261-277.
[9]李旭青,刘湘南,刘美玲,等.水稻冠层氮素含量光谱反演的随机森林算法及区域应用[J].遥感学报,2014,18(4):923-945.
[10]王丽爱,周旭东,朱新开,等.基于HJ-CCD数据和随机森林算法的小麦叶面积指数反演[J].农业工程学报,2016,32(3):149-154.
[11]刘艳丽.随机森林综述[D].天津:南开大学,2008.
[12]LYZENGA D R,MALINAS N P,TANIS F J.Multispectral bathymetry using a simple physically based algorithm[J].IEEE Transactions on Geoscience &Remote Sensing,2006,44(8):2251-2259.