Research On Ship Detection Technology For Spaceborne SAR Image

Synthetic aperture radar(SAR)is a kind of advanced active microwave imaging sensor,which can realize the all-day,all-weather,long-term and large-scale earth observation,and has great significance to our country's marine monitoring industry.Marine monitoring involves multiple steps of SAR image preprocessing,maritime target detection,identification and classification.Ship detection is the basis of subsequent research,which directly affects the final monitoring results.In this thesis,speckle suppression and constant false alarm rate ship detection in SAR images are studied to improve the accuracy of ship detection.Due to the inherent coherence imaging system of SAR,speckle noise on SAR image seriously affects the subsequent processing of SAR image.Therefore,in this thesis,SAR image is preprocessed before ship detection.In the aspect of SAR image preprocessing,this thesis applies the block matching 3D method,which is excellent in natural image denoising,to SAR image speckle suppression.The algorithm includes three steps: similar block matching,transform domain speckle suppression and aggregation after inverse transform.Firstly,the grey theory(GT)is used to measure similarity of similar blocks to improve the robustness of block matching.In the speckle suppression step,the similar block groups are transformed into the non-subsampled Shearlet transform(NSST)domain with optimal sparseliness,and the speckle suppression is achieved by contracting the NSST coefficients.Experimental results show that,compared with other classical coherent speckle suppression algorithms,this algorithm can suppress coherent speckles while retaining the detailed information of the ground targets.In terms of ship detection,after obtaining high quality SAR images through speckle suppression,this thesis proposes a two-stage constant false alarm rate ship detection algorithm in intensity-space(IS)domain based on the two-stage constant false alarm rate detection method.The algorithm simultaneously uses the intensity and spatial information of the SAR image to convert the original SAR image into a joint SAR image in the IS domain,and performs a two-stage constant false alarm rate detection on the joint SAR image.In the global detection,Gaussian distribution is used for background clutter modeling to calculate the global threshold;in the local detection,clutter statistical model is selected according to K-S test,and then accurately detecte in the potential target points detected in the first stage.Through the comparison of simulation experiments,the proposed two-stage constant false alarm rate algorithm in IS domain has better detection effect than the original ship detection algorithm.