Research Of Ocean Wind Fields Retrieval Algorithm Base On Marine Radar Images

Overall sea dynamic environment monitoring in real time has significant meaning tounderstand the ocean and exploit its resources. Marine radar can obtain sea clutter imageswith high resolution in space and time. By analyzing the spatial and temporal radar imagesequences with related algorithm, the ocean surface dynamic parameters can be extracted, andthen the sea state is known. Compared with other methods, marine radar has an incomparableadvantage and arouses wide interest of scholars all over the world. In this thesis, based on aresearch project of laboratory, ocean wind field retrieval algorithm is investigated, whichestablishes a foundation for further application and popularization of marine radar in seadynamic environment monitoring area.By summarizing the outcome of marine radar on ocean remote sensing in domestic andinternational, this dissertation makes a thorough study of wave noise suppression, oceansurface wind vector calculation. And a relatively complete technical scheme on ocean surfacewind field retrieval is formed, which was validated to be feasible by experiments usingsimulated and real image sequences.In order to remove the noise disturbance, a dispersion relationship band filter based onmain wave number is designed according to the dispersion relationship of gravity waves,named LDRBF. The bandwidth of LDRBF is decided by the main wave number andmaximum relative speed. Compared with traditional dispersion relationship band filter,LDRBF could work independently without real-time ocean current. In contrast to ordinarydigital image filter, LDRBF doesn’t destroy the original sea clutter structure by working onthe basis of sea clutter feature. Radar images recorded during two high sea state processes areapplied for experiment. The error of measured significant wave height derived from LDRBFis below0.26m compared with reference significant wave height, which shows the highperformance and feasibility of LDRBF.The traditional dispersion relationship band filter can be used for removing wave noisedisturbance of wind field retrieval. Ocean current is the key parameter of the traditional filter,while the current error is large calculated by traditional ocean current algorithm. In order toovercome the problem, a new method for retrieving ocean surface current is presented, namedWS-LSM. Compared with traditional algorithm, WS-LSM algorithm is significantlyimproved on noise suppression and weight selection of the least squared method. Long timeexperiment has been carried out. The comparison of ocean current speed resulted in an accuracy of0.11m/s with an increase of64.5%, and ocean current direction resulted in anaccuracy of16.33°with an increase of57.87%. It can be seen that the WS-LSM algorithm hashigh performance and feasibility.A new ocean surface wind field retrieval algorithm based on brightness and gradientconstancy model was presented aiming at the nautical radar image features that is, the imagehas much noise, strong nonlinearity and long interval. One is wind direction calculation.Firstly, with the constancy assumptions of brightness, gradient and considering thesmoothness effect, the energy function is established. By using the regularization method tosolve the energy function, the nonlinear Euler-Lagrange equations are deduced. Finally, oceansurface wind direction can be achieved by use of iterative method to solve Euler-Lagrangeequations. Another is wind speed calculation. With the radar cross section, the measured winddirection and the wave signal and noise ratio as the input, the structure of BP neutral networkis designed for calculating wind speed when radar works alone. When the parameters ofmarine and air boundary layer are known, another structure of BP neutral network is designedfor calculating wind speed with the parameters as additional input.In order to study the correlation between wind vector retrieval results and environmentalfactors as well as the feasibility of ocean surface wind retrieval technical scheme, whichprovide a basis for marine radar remote sensing ocean surface wind, ocean wind retrievalexperiments are carried out by simulated image sequences. Firstly, based on the basicprinciples of wave spectrum and radar geometric imaging, radar images are simulated. Aproblem worthy to be pointed out is that the effect of antenna rotation is taken into account inthe simulation process. Secondly, ocean surface wind field retrieval technical scheme isdesigned in detail. And then the effect of wind speed, measured area and antenna rotation onwind vector retrieval is analyzed by experiments under ideal condition. Finally experiment iscarried out with simulated images, and the results show that the wind direction resulted in anerror of2.76°, wind speed error of0.26m/s, which verify the feasibility of wind vectorretrieval technical scheme.Remote sensing experiment with shore based marine radar was carried out in2010. Agreat number of radar images were recorded simultaneously with observations of wind vectorby in-situ sensor. Marine radar images are analyzed by wind vector retrieval technical scemedescribed above. The comparison of wind direction results in an error of11.3o. Thecomparison of wind speed results in an error of0.50m/s when considering radar cross section,wind direction and wave signal noise ratio. While considering marine and air boundary layerparameters, the wind speed error is0.26m/s. The feasibility of ocean surface wind field retrieval technical scheme is further verified by the experimental results.