Research On Novel Model Of Optical Remote Sensing Bathymetry For Complex Situations

Optical remote sensing bathymetry can realize the macroscopic dynamic observation of shallow seawater in an economical and effective way,especially for some areas where difficult to reach or have politically sensitive issues.The currently problem is that the models have poor portabilities,and the detection accuracies can be affected by the water environments,especially in the 0-5 m,the inversion error is mostly above 50 % or even higher.Based on the above problems,the paper proposed a multifactor log-linear model,aiming to improve the bathymetric quantification accuracy.We selected the East Island and the Bengkalis Island with different turbidity to verify and evaluate the inversion ability of the multi-factor log-linear model.Considering the bottom spatial heterogeneity,this paper proposed a partitional optical bathymetry method based on the discrete control points' residuals spatial information,and we selected East Island as an example to verify the inversion capability of the method from the qualitative and quantitative aspects.Besides,for weak illumination,we used the water radiation transmission model to establish the water-light field and explore the applicability of the novel multi-factor log-linear model.The main conclusions are as follows:1)In East Island,the inversion accuracy of the multi-factor log-linear model is higher than that of the traditional model and the improved model,which MAE(Mean Absolutely Error)is 1.04 m and MRE(Mean Relatively Error)is 9.6 %,especially in 0-5 m,the MRE is 25.5 %,which is 41.9 % lower than the traditional model and 24.93 % lower than the improved model.In Bengkalis Island,the multi-factor log-linear model also has the highest inversion accuracy.It can clearly reflect underwater terrain changes,which the near-shore terrain is flat and the off-shore forms a series of 1.0-1.8 km sandstone ridges with northwest-southeast trend.Therefore,the novel model shows a relatively strong applicability.2)The partition remote sensing bathymetry method was established based on the discrete control points' residual spatial information and established a series of inversion models that are adaptive to each sub-region.Specific to each segment,the accuracy of the partition model is improved compared to the global model.Especially in 0-5 m,the MRE of region-A is 14.1 %,which is 19.4 % lower than the global model;besides,the error of region-B in 0-5 m is reduced by 5.8%,and the accuracies of other segments are improved compared with the global model.3)Based on the typical Arctic Ocean water column parameters,this paper simulated the water field in the Arctic waterway,and studied the applicability of the multi-factor log-linear model.Compared with the single-band models,the multi-factor log-linear model inverts the water depth more accurately,and the highest accuracy model is the green-red multi-factor log-linear model with the MAE is 1.39 m and MRE is 15.9 %.The next are the blue-green multi-factor log-linear model and the blue-red multi-factor log-linear model respectively.Therefore,the multi-factor log-linear model compensates for the instability of the inversion ability of different single-band models,and it is suitable for water depth inversion studies in weak illumination situations.