海底地形的SAR影象仿真与反演

IOCAS-IR > 海洋环流与波动重点实验室

	海底地形的SAR影象仿真与反演
	金梅兵
学位类型	博士
	1998
学位授予单位	中国科学院海洋研究所
学位授予地点	中国科学院海洋研究所
学位专业	物理海洋学
摘要	本文旨在建立SAR水下地形探测的理论和应用的基本框架。研究内容包括：1. 水下地形的SAR成象机制采用Bragg散射理论与海面微尺度波高频谱的饱和谱形式相结合，建立了后向散射截面与海面微尺度波流调制作用间的函数关系，该函数为雷达相关项与大尺度海流对海面微尺度波的调制作用项之积，该调制项包含海底地形的信息。2. 建立水下地形的SAR影象仿真模型，定性和定量地解释SAR影象的类地形条纹以及SAR在不同的水文条件下对水下地形的可视性以二维深度平均的海流模式或三维σ坐标的海流模式计算流场，结合风波即可计算出大尺度海流对海面微尺度波的调制作用项，文中记为函数G，并称计算所得的G值灰度影象为SAR的仿真影象。通过仿真模型可以研究不同微尺度波波向、流向及地形梯度方向情况下SAR影象对海底地形变化的可视性，结果表明，当海流流向、地形梯度方向和微尺度波波向三者平等时，SAR影象中的类地形牲的可视性最好。随着这三个方向夹角逐渐增大，可视性逐渐降低，当其中两个方向垂直时，类地形特征消失或变得很弱。较大的流速可以增强类地形影象的对比度，这在SAR影象中存在背景噪声时是重要的。3. SAR水深反演的反问题的提法与解法以及反演模型的反演精度分析通过SAR探测海底地形可以归为已知雷达后向散射截面或G值求水深的数学物理反问题，本文的反演模型由G 值函数表达式和二维深度平均的海流运动方程组以及适当的初边值条件构成。由于该反问题的边界条件的提法为超定的，所以其解设计为求目标泛涵的最小偏差解，反演算法为最速下降法。该算法不受反演大小的限制，具有较好的收剑性和解的存在性，但解依赖于初猜值的选择。当反演区域内部分点的水深已知时，这些数据可以很容易地融合在最小偏差解的算法中，从面加快收敛速度，提高计算精度。利用仿真模式所给出的SAR影象证实了该反问题的求解方法是可行的，且具有较高的计算精度。通过反演模型进行的数值试验表明： 1）微尺度波波向、流向和地形梯度方向对反演精度有明显影响。三个方向平等时反演精度最高，当其中任两向垂直时，都会成为反演的盲点。这是由反演模式本身的奇性所决定的。虽然三向夹角不同时，反演精度有变化，但在大多数情况下，反演值与真值地形波动的位置，形状和幅度等都具有很好的一致性。只有当其中两向接近垂直时，才出现较大的差异或反演失效；2）虽然一定流速的存在是反演的基础，但流速的大小对反演精度影响并不大；3）本反演系统对G值误差和反演未知量的初猜值的误差以及波向角选取误差都有较强的容差能力；但对水深边值误差较敏感，要求其精度较高；4）由于反演模式需要水深的初猜值，所以需要知道反演区域的平均水深，一个好的初猜值可以加速算法的收敛，而一个较差的初猜值则可能引起较大的误差，一船说来，平均水深就已经可以获得较高精度的结果。4. 仿真和反演模型的一个实例研究本文以北海（the North Sea)的一张 Seasat SAR 影象和该区域的海图水深为例，采用所建立的仿真模型和反演模型对这张SAR影象上的类地形明暗条纹进行了成功的仿真，也通过SAR影象数据经过数据处理和反演，求出了影象区域的水深，反演水深数据和实际海底沙脊吻合较好。研究结果初步验证了海底地形结构的SAR影象仿真与反演的可行性。本文的研究表明，只要SAR影象在低海况下拍摄，且有精确的定位，再加上对SAR影象所覆盖海区的粗精度的水深和边界水深的已知条件，就可以在对该海区进行潮流模拟的基础上用反演模式来获取高空间分辨率的水深资料。
其他摘要	This paper intends to establish a basic structure of the theory and application of detecting water depth by SAR. The contents covers: 1. Mechanism of SAR imaging under water topography A function formula between the radar backscatter cross section and sea surface short gravity wave-current interaction is established on the basis of the combination of Bragg backscatter theory and the equilibrium spectrum formula of high frequency sea surface short gravity waves. The function is related to radar parameters and the modulation of large-scale current on sea surface short gravity waves which includes the sea bottom topography information. 2. Simulation model of under water topography on SAR image, qualitative and quantitative explanation of the topography-like features on SAR image and visibility of topography on SAR image under different hydrometeorological conditions The tide current field is modeled by the two-dimensional depth-averaged shallow water current model or three-dimensional σ-coordinate current model. The modulation of large scale current on sea waves is calculated on the basis of the current and wind wave parameters. This modulation is defined as function G and the gray tone level of calculated G is called as SAR simulation image in the paper. The visibility of sea bottom topography on SAR image is studied under different conditions of sea surface short gravity wave direction, current direction and topography gradient direction using the simulation model. The results show that the visibility of topography-like features on SAR image is the best when the three directions of current, topography gradient and sea surface short gravity wave are parallel. The visibility decreases when the angles between the three directions increase and disappears or becomes very weak when two of which are perpendicular. Larger current velocity can increase the contrast of the topography-like features which is important while background noise exists. 3. Formulation and solution of the inverse problem of detecting water depth by SAR image and analysis of the resolution of the inversion model The detection of sea bottom topography by SAR can be described as a math-physical inverse problem of detecting water depth with known radar backscatter cross section or G. The inversion model constitutes of G formula, two dimensional depth-averaged shallow water equations and suitable initial and boundary conditions. Because the boundary condition of the inverse problem is improper, the solution of the inverse problem is designed to solve the minimum deviation solution of an objective functional and the algorithm is 'fast falling method'. This algorithm is free of limitation of inversion area and has good convergence and existence of solution, but the solution depends on the initially guessed value of the variable. The known data of water depth within the inversion area can be easily combined in the algorithm which can accelerate the convergence and resolution of the solution. This method of inversion is verified to be feasible and has high resolution by using the SAR image from the simulation model. The numerical test of the inversion resolution using the inversion model shows that : 1) the inversion resolution is affected by the directions of sea surface short gravity wave, current and topography gradient. The inversion resolution reaches the best when the three directions are parallel and goes to 'blind' when two of the directions are near perpendicular as determined by the characteristics of the inverse problem itself. Although the inversion resolution changed with the angles between the three directions, the position, shape and amplitude of the topography wave of the real and inversed water depth show good agreement under most conditions except the condition when two of the three direction are near perpendicular; 2) although the existence of certain current velocity is necessary for the detection, the magnitude of the current velocity is not important for the inversion resolution; 3) the inversion system is insensitive to the error of G, initially guessed value of unknown variables, and the angle of sea surface short gravity wave direction, but sensitive to the value of the water depth on the boundary, so water depth on boundary is required to have high resolution; 4) A initially guessed value of water depth is needed in the inversion model, so the mean water depth of the inversion area should be known previously. A good initially guessed value can accelerate the convergence of the algorithm and a too bad initially guessed value may arose deviations of the solution, but generally, using mean depth as initially guessed value can obtain high resolution results. 4. An example application of the simulation and inversion model Using the simulation and inversion models, the topography-like features on a Seasat SAR image of the North Sea is successfully simulated according to the sea chart water depth, and also the water depth of the image area is calculated on the basis of the digitally processed SAR image. The inversed water depth shows a good match with the sea chart depth. This result demonstrates the feasibility of the simulation of SAR image and the extraction of water depth by SAR. The research indicates that it is possible to extract water depth with high spatial resolution by SAR image on the basis of tide current modeling and the knowledge of low resolution water depth and boundary water depth of the SAR image area if the SAR image is obtained under low sea condition and with accurate positioning.
语种	中文
文献类型	学位论文
条目标识符	http://ir.qdio.ac.cn/handle/337002/385
专题	海洋环流与波动重点实验室
推荐引用方式 GB/T 7714	金梅兵. 海底地形的SAR影象仿真与反演[D]. 中国科学院海洋研究所. 中国科学院海洋研究所,1998.

条目包含的文件
文件名称/大小	文献类型	版本类型	开放类型	使用许可
N97624.pdf（2659KB）			限制开放	--	浏览