Inversion Algorithm Of Ocean Waves Parameters From X Band Radar Based On Intelligent Learning

21 century is a new age to exploit ocean resources,which is essential for a further economic development.Has been suffering from provocation,ocean is important for national security.Meanwhile,ocean pollution has heavily impacted ocean environment.Thus,ocean study and researches are always popular topics in kinds of disciplines and fields.Among the study,physical environment monitoring is drawn special focus due to its importance for social economic development.Only with better study and knowledge about ocean,can we utilize more of the unlimited resources in ocean.X-band nautical radar has a high resolution in both time and space domains.It has the advantage of low cost,stability,and installation facility.X-band nautical radar has been applied to ocean detection and early warning for a long time.Because of the ability of automation and of 24-hour working in all-weather conditions,X-band nautical radar is also convenient to be used in wave and steam supervisions.The thesis first reviews several mature methods for ocean wave parameter supervisions,and current situations for x-radar based ocean wave parameter supervisions.After an introduction to the theory of radars in ocean imaging and modulation principles,the thesis presents concepts and current study situations about artificial neutral networks.Aiming at the inversion problem of ocean surface parameters,the thesis adopts Least-Squares Method,Iterative LSM,Normalized Scalar Product(NSP),respectively.In practice,ocean waves do not present an ideal relationship of linear super positions.Especially in complex scenes,linear relationship based inversion algorithms cannot produce an accurate relationship between radar image and ocean wave parameters.The inaccuracy is caused by the adopted linear relationship based transfer functions in calculations.Therefore,the thesis proposed a new inversion method based on generalized artificial neutral networks,in order to fit ocean spectrum and invert ocean parameters.By the method,the thesis contructed the mapping relationship between one-dimensional spectrum and buoys spectrum in radar images,and thus replaces the traditional modulation relationship with the proposed relationship,to calculate effective wave height.The numerical simulation of ocean radar images can help us understand the mechanism of backscatter signal generations,and the calculation of simulated images can be used to compare differences of three algorithms.After describing wave spectrum expressions and fitting radar spectrum and simulated spectrum,the thesis compared different algorithms in inversion simulation of surface stream and wave parameters,and proved the feasibility of proposed method.By huge amount of ocean radar image data and corresponding buoys data,which was acquired by actual measurement,the thesis inverted radar data parameters,and compared the results with actual corresponding buoys data.For the inversion problem of effective wave height,the thesis adopted and compared traditional linear fitting parameters based algorithms,and generalized neutral networks based algorithms.Generally,all algorithms achieved good results on corresponding surface wave parameters.However,in terms of effective wave height,the neutral network based algorithm has a better result.As for buoys parameters,the coefficient of calculation results are all higher than 75%.