Research On Method Of Detecting Ship And Wake Based On SAR Image

As an important part of marine scientific research,the ship detection on sea surface plays a key role in the safeguard of maritime rights and interests and protection of marine ecology.The detectors for sea surface targets are generally required to have the capabilities of remote real-time monitoring and weather interference resistance.As an active microwave imaging radar,synthetic aperture radar(SAR)is more suitable for marine target detection than other sensors.Therefore,SAR images are widely used in ship detection research.According to the overseas and domestic research status in recent years,ship detection in SAR images can be divided into two main categories,i.e.direct detection and indirect detection.Direct detection refers to using the characteristics difference between ship and sea clutter in SAR images to detect the ship target,while indirect detection utilizes auxiliary features such as wake to judge whether the ship exists.In this thesis,the detection methods for ships and ship wake are studied respectively.Then the corresponding improved algorithms are proposed.With respect to the direct detection,according to the difference between the statistical characteristic of clutter background and that of ship target in SAR images,constant false alarm rate(CFAR)detection algorithm based on clutter statistical modeling are firstly discussed in terms of the principle and workflow.Then several common statistical distributions are introduced for clutter modeling in SAR images.To overcome the problem of inaccurate clutter modeling in multi-target environment,two schemes of clutter censoring and clutter truncation are compared and analyzed.Besides,to improve the detection efficiency,a superpixel-level CFAR detection algorithm is proposed based on the truncated Gamma distribution and the preprocessing of superpixel segmentation.The comparative experiment shows that the truncated Gamma distribution has the best clutter fitting performance for pure clutter samples.Thus,the proposed method achieves the highest detection rate among all comparative algorithms.In addition,the comparison of time consumption also verifies that the proposed method has high detection efficiency.With respect to the indirect detection,the component,spatial and frequency domain characteristics are analyzed for wakes of ships and underwater targets.Then the electromagnetic scattering modeling and modulation mechanism imaging are carried out based on hydrodynamic model and sea surface physical model.With this basis,a local CFAR wake detection algorithm is proposed based on normalized grayscale Hough transform.The contributions of the proposed method lie in two aspects.First,the detection performance is improved due to the enhanced contrast of wake features in the Hough transform domain.Second,the preprocessing of ship masking is unnecessary to eliminate the influence of target on wake detection.It is validated by experimental results with simulated and measured SAR images that the proposed algorithm can improve the wake detection performance of ships and underwater targets under certain conditions.Finally,the proposed algorithm is integrated into the client software by hybrid programming with MATLAB and MFC,where the SAR image ship and wake detection system is developed to facilitate the research work.The realization process is introduced for each function module in the software design and development.