Research on bathymetry estimation by Worldview-2 based with the semi-analytical model

South Sea Islands of China are far away from the mainland, the reefs takes more than 95% of south sea, and most reefs scatter over interested dispute sensitive area. Thus, the methods of obtaining the reefs bathymetry accurately are urgent to be developed. Common used method, including sonar, airborne laser and remote sensing estimation, are limited by the long distance, large area and sensitive location. Remote sensing data provides an effective way for bathymetry estimation without touching over large area, by the relationship between spectrum information and bathymetry. Aimed at the water quality of the south sea of China, our paper develops a bathymetry estimation method without measured water depth. Firstly the semi-analytical optimization model of the theoretical interpretation models has been studied based on the genetic algorithm to optimize the model. Meanwhile, OpenMP parallel computing algorithm has been introduced to greatly increase the speed of the semi-analytical optimization model. One island of south sea in China is selected as our study area, the measured water depth are used to evaluate the accuracy of bathymetry estimation from Worldview-2 multispectral images. The results show that: the semi-analytical optimization model based on genetic algorithm has good results in our study area;the accuracy of estimated bathymetry in the 0-20 meters shallow water area is accepted.Semi-analytical optimization model based on genetic algorithm solves the problem of the bathymetry estimation without water depth measurement. Generally, our paper provides a new bathymetry estimation method for the sensitive reefs far away from mainland.


INTRODUCTION
Shallow bathymetry around the sea reefs or islands is important to the terrain elements, and also is the premise of marine work.Commonly used bathymetry estimation models are the combination of semi-empiricaland semi-theoretical model, depending on certain theoretical and empirical parameters.
However the application and accuracy of these models are limited by the usable and precision of measured depth data before model construction, which is difficult to obtainfor the islands or reefs far away from the mainland.Therefore the theoretical models based on the radiation field distribution of sea water are selected to estimate the bathymetry by remote sensing data in this paper.Currently, because of many uncertain parameters in theoretical models are complex and *Bai Jie. Email: 13811199372@126.comdifficult to be fixed, related researches are few.Based on double layer approximation model, Zyzenga (1978) derived the surface remote-sensing reflectance for the shallow waters,and then obtained the bathymetry, but the model is only a theoretical experiment.Based on the simplified model of the radiative transfer equation by Jupp (1988) and Bierwirth (1993)

Water Information Extraction
The land information in the image should be removed before bathymetry estimation model constructing, which is used to generate a new image only containing water area.In our study, the image of Near-IR1 band is applied in edge detection with Canny algorithm.After series of experiment, the value of the double threshold in the Canny algorithm has been determined, with the lower threshold is 20 and the higher one is 1000.The extracted edge line of our study area is showed in Figure1.In order to verify the effect of edge detection, the edge has been overlapped on the original image, which is also shown in Figure2.firstly determined based on simulated data (Lee, 1999).
Where 0.5 0.5 (0.084 0.170 ) 1.03(1 2.4 ) 1.04(1 5.4 ) Where w  is subsurface solar zenith angel;  is subsurface viewing angel;  is an attenuation coefficient; H is bathymetry; a is total absorption coefficient of water; b b is total backscattering coefficient;  is bottom albedo.But in order to prevent pathological inversion,the original model has been parameterized by bio-optical model (Lee, 1998).
Wherein the total water absorption coefficient ()

Algorithm
After parameterization, the parameters of , P G X B H , , , determine the remote-sensing reflectance in theory to inverse bathymetry, as long as there is no less than five channels of remote sensing data.But the radiative transfer equation is still complex to solve in the situation of no adequate data, and the speed of the calculation is slow.Thus, the genetic algorithm is used to solve the parameters under the condition of lack of data sources; moreover, OpenMP parallel processing model is applied to the genetic algorithm for saving the hours of solving process.

Bottom Type Determination
The bottom sediment types of sea water can be divided into two types: sand and algae.We use the empirical formulas to distinguish between different types of sediment.If the water-leaving reflectance of pixel satisfiesequation 13, the type of plant is judged; otherwise the type of bottom sediment is seemed as sand (Lee, 2002a).
(550) 0.01 (725) (660) 1.2 The water-leaving reflectance of red edge, red and green band is substituted into equation ( 13).The result shows that the experimental area of bottom    Table 2 shows that accuracy of semi-analytical model based on genetic algorithm of bathymetry estimation, the average error of the 0 to 30 meters is 2.99 m, the average relative error is 35.28%;error for 0 to 5 m water is 2.32 m, the relative error is 49.09%; the error for 5 to 10 m water is 1.54 m, the relative error is 17.56%, error of 10-20 m is 1.99 meters, the relative error is 14.38%; the average error of 20-30 m water is 4.12 m, the average relative error is 50.28%.

CONCLUSIONS
In this paper, the semi-analytical model of ocean biomass inversion is applied to bathymetry estimation with remote-sensing data.The genetic algorithm and OpenMPprogramming model for shared memory parallel processing is introduced.This method of bathymetry estimation solves the problem under the condition of lack prior knowledge, so it provides an effective way for bathymetry estimation for the sensitive islands far away from the mainland.
, Zhou Yan (2011) established the direct inverse radiative transfer model.But the value of the attenuation coefficient in this model was directly achieved by the quasi-analytical algorithm,resultingthe increment of waterreflectioninformation and decreasing the model accuracy.So it is necessary to study the theoretical bathymetry estimation model without measured water depths.In this paper, by referencing ocean biomass estimation in the ocean color remote sensing, the semi-analytical model is establishedto estimate the bathymetry.2. STUDY AREA AND DATA SOURCES East Island locates in the east of the Xuande atoll of Paracel Islands and is selected as our study area, which is 50 kilometers apart away from Yongxing Island.Worldview-2 remote sensing image is used for bathymetry estimation.The acquisition time is September 20, 2012; with 2 meters spatial resolution.Some details are shown in

Wavelength
Based on UVA DOMimage and header file from Worldview-2 image, geometric correction, radiation calibration and atmospheric correction has been carried out for original image.The values of DN are transformed to reflectance, the influence of atmospheric scattering and absorption are also corrected.

Figure 1 Figure 2
Figure 1 Result of extracting edge line by Canny algorithm taken from empirical parameters;is the spectrum slope in the range of 0.011-0.021nm-1;S value of 0.015 nm is used in the inversion process.The total coefficient of backscattering () of  is expressed by a 550-nm-normalized albedo shape.

The
Figure3 550-nm normalized bottom albedo spectrum

Figure 4
Figure 4 Parts of running results of genetic

Figure 5 Figure 6
Figure 5 Acceleration efficiency before using OpenMP parallel processing model

Figure7
Figure7 Bathymetry estimated from semi-analytical inversion model over East Island Reef Bathymetry within the scope of 5-20 mhas a high precision;relatively the inversion precision is lower in 0 to 5 m and 20 to 30 meters.Due to the inversion

Table
. The measured water depth data has been acquired between May and August, The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-7/W3, 2015 36th International Symposium on Remote Sensing of Environment, 11-15 May 2015, Berlin, Germany This contribution has been peer-reviewed.doi:10.5194/isprsarchives-XL-7-W3-1305-20151306