Detection Methods Based On Hough Transforms Of Sea-surface Small Targets

Our country has a vast ocean area and the coastline of length more than 32000 kilometers.With the development of the oceanic strategy of our country and needs of maritime trade and deep-sea fishing,subtle information on various sea-surface targets is required.Maritime radars are one of the main sensors to detect various sea-surface targets and their detection performance is severely influenced by sea clutter,the backscattering signals from the sea surface.In order to master the statistical characteristics of sea clutter,besides the out-field measurement,the simulation of the radar returns from the sea surface is also an important approach to obtain data.As for sea-surface target detection in high-resolution maritime radars,sea clutter and sea spikes are always major factors to affect the detection performance and particularly sea spikes are the main reason to result in high false alarms in highresolution maritime radars.In this thesis,we investigate the simulation method of largescene sea clutter data by combining the statistical models of sea clutter with the sea and wave generation in the oceanography and carry out the simulation software system.Moreover,the multi-scan joint sea-surface small target detection methods are also investigated that using the Hough transformation and its modified versions and the proposed methods are verified by using simulation experiments.The main content of this thesis is as follows.Chapter 2 reviews and summarizes the physical scattering mechanism of sea clutter,amplitude distribution models of sea clutter,and the Doppler spectral characteristics of sea clutter,as well as the other relevant knowledge on sea clutter,which form the foundation for us to carry out the simulation of large-scene sea clutter data.Chapter 3 investigates the simulation method of large-scene sea clutter data and carries out the corresponding simulation software system.Sea clutter is substantially the results of complicated interaction between the electromagnetic waves transmitted by radar and dynamically evolutive sea surface and thus sea clutter simulation only based on statistical models suffers from limitations.The basic idea is to combine the simulation of large-scene dynamically evolutive sea surface with statistical simulation.The sea spectrum model and the sea-surface electromagnetic scattering model are introduced and the sea spectrum consists of waves with different wavenumber,directions,and amplitudes,which are the foundation to simulate large-scene dynamically evolutive sea surface.The empirical formulae between the reflectance coefficients of the sea surface and sea state,the viewing geometry of the radar,and radar parameters is the bridge between the physical parameters of the sea surface and the statistics of sea clutter.Besides,the compound-Gaussian models of sea clutter,the KA model of sea clutter plus sea spikes form the foundation in statistics.The fusion of the three factors supports the simulation method and software proposed in this thesis.Chapter 4 investigate the multi-scan joint detection method of sea-surface small targets based on the Hough transformation and proposes a robust ridge integration detection method to compensate some defects of the traditional method based on the Hough transformation.The multi-scan joint detection framework of sea-surface small targets often consists of the main blocks of the intrascan detection and interscan detection.The intrascan detection transfers radar returns at individual spatial resolution cells into binary plot data and the interscan detection integrates the plot data in terms of the moving model of targets so as to lower false alarm rate and exclude sea spikes.The scheme in this thesis adopts the block whitening ratio median CFAR detection and the research focuses on the intrascan integration methods.The motion model of sea-surface small targets often use the constant velocity(CV)model or the constant acceleration(CA)model.The traditional Hough transform is an effective method to detect traces of lines or approximate lines.However,experiments show that the detection performance of the Hough transformation based method is sensitive to errors in plots and random change of target radial velocity.In order to overcome this defect,this thesis proposes a robust ridge integration detection method,which exploits a single parametric ridge function to overcome this defect of the Hough transformation based method.By a mass of simulation experiments,the robustness of the proposed ridge integration method is verified and the influence of the target moving models and the parameter of the ridge function on the detection performance is analyzed.Finally,the works in the thesis are summarized and the future works in this field are discussed.