Study On The Methods Of Ship Detection And Identification In SAR Images

The Synthetic aperture radar is an active high-resolution imaging microwave sensor. It has significant performance on penetrating cloud, rain and smoke with the ability of operating all day and in all weather. Besides, its advantages over the visible light sensor and the infrared sensor enable its availability to multi-angle and high-resolution sea imaging. The ship target detection based on SAR image is developed in this context. This paper discusses several following key problems of ship target detection system based on SAR images.1. SAR image preprocessing is studied. SAR image preprocessing consists of two aspects, speckle noise suppression and land mask. With regards to speckle suppression, this paper introduces the conventional spatial domain filtering method, the filtering method based on statistical models and the filtering method based on wavelet transform. The conclusion that the wavelet transform algorithm performs better is attained by simulation experiment and comparison of characteristics with the other methods. In terms of land mask, this paper starts with two current methods, one of which utilizes coastline information available in GIS(geographic information system) while the other one uses SAR image processing to mask the land automatically without priori knowledge. The latter one falls into two types, based on texture segmentation or coastline detection. This paper studies the land mask method based on texture segmentation which proves effective through following experiments.2. As the core part of the ship detection system based on SAR images, ship target detection is discussed. Commencing with the introduction of the statistical modeling approach of sea clutter in SAR images, then the existing clutter statistical models are summed up. This paper also clarifies the mathematical derivation and parameter estimation of each model. Then performances of different models to fit sea clutter are compared by statistical modeling of measured SAR images. Next, this paper focuses on ship target detection based on CFAR. On basis of the discussion of advantages and drawbacks of global and local CFAR detection algorithms, the combined-CFAR detection algorithm is further studied. An advanced combined-CFAR detection algorithm is developed to modify remaining shortcomings of the former one.3. Target identification is studied. Considering the differences between the target and the false alarm in the geometry, contrast and texture feature, these features of each potential target are extracted. Then the distinction of each feature is calculated to select the best several ones as target identification operators to distinguish targets.4. A feasible scheme of target detection system is proposed. The filtering method based on wavelet transform is used for speckle suppression. This scheme applies texture segmentation for land mask. The improved CFAR method is utilized to detect target and finally the best-distinction features are used to eliminate false alarms. In order to further illustrate the feasibility of the scheme, it is verified by computer simulations and real-data processing. Besides, this paper discusses the influence of different SNRs, ship types and attitudes on detection performance. Experiments show that the proposed scheme has a great performance for SAR images with a certain SNR. In addition, it is commonly suitable for different ship targets.