Research On Ship Targets Polarimetric Interpretation And Detection In Imaging Radar

Polarimetric scattering interpretation is a key procedure of Pol SAR information processing,which has vital influence on the applications of Pol SAR,such as ship detection.This paper focuses on the ship targets polarimetric interpretation and detection in imaging radar,aiming at mining polarimetric information and extracting polarimetric feature.Pol SAR ship detection methods based on the derived polarimetric saliency feature and rotation domain features are proposed in this paper,the effectiveness of which is validated with Radarsat-2 sparse area and dense area,and Gao Fen-3 data I and data II.The main works and innovations of this paper are as follows:（1）A novel polarimetric saliency feature and saliency detection method for Pol SAR ship detection is proposed.Firstly,the polarimetric saliency feature SPN（Similar Pixel Number）is derived using similarity test of polarimetric covariance matrix.Secondly,fast algorithms of similarity test and SPN feature extraction are proposed in terms of code optimization.The time consumption is reduced to 4.12%of the traditional algorithm for 1000×1000 size data.Lastly,saliency detection method for ship target is proposed based on the derived polarimetric saliency feature,which mainly contains three steps:polarimetric saliency feature extraction,thresholding procedure,and morphological filtering.Experimental studies validate the advantages of the propsed method.（2）Polarimetric correlation pattern visualization tool is developed and ship detection methods based on polarimetric rotation domain features are proposed.Firstly,the polarimetric correlation pattern is proposed based on the polarimetric rotation domain theory.A set of new polarization features are derived.Secondly,polarimetric feature with high TCR（Target to Clutter Ratio）are selected.Finally,ship detection methods based on the selected polarimetric features are proposed,which mainly contains three steps:polarimetric saliency feature extraction,thresholding procedure,and morphological filtering.Compared with other methods and other polarimetric rotation domain features,polarimetric feature（?）γ_{（HH-VV）-（HV）}（θ）_organd（?）γ_HH-HV（θ）_orgwith the top TCR have the highest comprehensive performance.（3）Ship detection methods combining polarimetric features and superpixel technique are proposed.On the one hand,target-level ship detection method based on single polarimetric feature and superpixel technique is proposed.In detail,superpixel segmentation is conducted on the original polarimetric rotation domain features.Then,thresholding procedure is applied and detection results are obtained.Experimental studies prove that the target-level detection method outperforms pixel-level detection method in terms of less false alarms and high detection probability.On the other hand,the ship detection method based on multi-feature and superpixel technique is proposed.Firstly,the polarimetric feature combination is selected from the detection results of single polarimetric features.Secondly,the selected feature combination makes up a3-channel RGB image,and the selected features are fused through the superpixel segmentation of RGB image.Finally,thresholding procedure is applied on the fused feature and the detection results are obtained.Experimental studies validate the advantages of the detection method based on the fused feature than that based on single feature.Specially,the fused feature based on（?）γ_HH-HV（θ）_{o rg},（?）γ_{（HH+VV）-（HH-VV）}（θ）_minand（?）γ_{（HH-VV）-（HV）}（θ）_minhas the best performance.Above all,polarimetric saliency feature and rotation domain feature are developed through polarimetric interpretation,which perform well in ship detection in imaging radar.This paper further promotes the development of Pol SAR scattering interpretation and improvement of ship detection performance.