Research On Large Ship Recognition Technology In Wide-swath SAR Image

Synthetic Aperture Radar(SAR)plays an important role in ocean monitoring because of its unique advantages of all-day,all-weather imaging.Compared with Strip Map and Spot Light SAR acquisition modes,the wide-swath SAR represented by Top SAR and Scan SAR imaging modes can observe a vast area of ocean scenes.However,wide-swath SAR achieves a wide swath while reducing the quality of imaging resolution,which brings a great challenge to the identification of large maritime ships.The recognition of large ships in wide-swath SAR images has problems such as low image resolution,difficulty in obtaining ship samples efficiently,and scarcity of specific ship samples at sea.To address the above issues,the main innovations and work of this thesis are as follows:(1)The download,storage,and processing of wide-swath SAR data are constrained by computer performance and human resources,which makes it difficult to obtain ship samples of wide-swath SAR images efficiently.Aiming at this problem,a ship detection method of wide-swath SAR images based on a remote sensing cloud platform is proposed.And the method is combined with Automatic Identification System(AIS)data to construct a wide-swath SAR ship sample set.First,a remote sensing cloud platform is used to process the SAR images in a large area of sea online to achieve real-time fast detection of wide-swath SAR image data and fast acquisition of ship information.The experimental results show that the method proposed takes a shorter time and faster speed in overall detection.Then,the ship information is combined with AIS data to construct a wide-swath SAR ship sample set,which realizes the efficient acquisition of wide-swath SAR civil ship samples.(2)The data of wide-swath SAR ship recognition is lacking,especially the large military ship samples at sea.Aiming at this problem,a wide-swath SAR large maritime moving ships dataset is constructed.First,the sample data of large military ships in the port area are obtained from prior knowledge.Second,the sample data of large civilian ships are obtained via the length screening of the Open SARShip dataset with attribute information.Finally,the imaging results of moving ships at sea are simulated by adding quadratic phase error in a range-Doppler domain.In the experiment,the wide-swath SAR large maritime moving ships dataset is analyzed in terms of basic information,clustering,and similarity.Then the depth learning method is used to analyze the recognition performance of the amplitude image,and the influence of motion simulation on the recognition performance is discussed.(3)The ships in wide-swath SAR images lack clear structural features,few features that can be used for target recognition,and only uses amplitude images with less information.Aiming at this problem,a recognition method of ships in wide-swath SAR complex images based on an improved Res Net network is proposed.First,the complex information representation of input data is implicitly provided by combining the real part,imaginary part,and amplitude information of the SAR image.Then the channel attention mechanism is introduced to the Res Net18 network structure so that the network can adaptively learn the complex information contained in the three channels of the real part,imaginary part,and amplitude.Finally,label smoothing regularization is introduced to alleviate the over-fitting phenomenon due to the lack of samples in complex data sets.The experimental is based on the Open SARShip dataset and wide-swath SAR maritime large moving ship dataset,which verifies that the proposed method can make good use of the complex information of SAR images and improve the accuracy of neural network-based ship recognition to a certain extent.