Research On Target Detection And Segmentation Method Of Spaceborne SAR Images In Complex Sea Scene

Spaceborne Synthetic Aperture Radar can realize long-term,high-quality and widerange high-resolution observation and imaging of the sea,and it is an important tool for managing marine areas and maintaining marine safety.Target detection and segmentation method in complex sea scene of spaceborne SAR images refers to discovering,locating and separating interested targets from complex sea scenes based on spaceborne SAR images.which can provide valuable intelligence information for maritime traffic planning,coastal defense monitoring and early warning.Because the complex sea scene of spaceborne SAR images has the characteristics of changeable clutter background and diverse target types and scales,target detection and segmentation face many challenges.Focusing on the theme of target detection and segmentation in complex sea scenes of spaceborne SAR images,this dissertation conducts research on offshore ship target detection,harbor target detection,inshore ship target detection,and ship target segmentation.The main contributions are as follows:1.A robust detection method of ship target in offshore area is proposed.By designing scale adaptive and context aware modules,the robustness of the method is enhanced to the variable offshore scenes such as target scale change and clutter environment change.Through the ship density map,the shape characteristics of the ship target can be better fitted,thus improving the generalization ability of the detection method in the changeable offshore scene.2.An efficient harbor target detection method is proposed.Through multidirectional one-dimensional scanning,the invariant features of complex and diverse harbor targets are extracted,so the detection of harbor targets in two-dimensional images is transformed into the identification problem of one-dimensional harbor feature vectors,which improves the efficiency and accuracy of harbor target detection.3.An inshore ship targets detection method is proposed.By analyzing the global semantic information,the candidate ship targets with global stability characteristics are obtained.Through local context information mining,the regional information of the targets’ local background is extracted,which effectively suppresses the false alarm in the complex land background and realizes the accurate detection of the inshore ship target.4.An fine segmentation method of ship target is proposed.The strategy of cluster center transformation is proposed to capture ship pixels more accurately.And the adaptive similarity criterion is proposed to better cluster similar pixels.Effectively ensuring the fitting accuracy of ship target edges and obtaining pixel level ship target segmentation results.The effectiveness of the above work has been verified by the experiments.The results show that the proposed methods can effectively solve the main problems in target detection and segmentation in complex sea scenes of spaceborne SAR images,and improve the target interpretation ability of complex sea scenes in spaceborne SAR images.