Ship Detection in Polarimetric SAR Imagery Based on Target Decomposition and Weighted SVM Classifier

Target decomposition is an important method for ship detection in polarimetric SAR imagery. Under the condition of relative high resolution and complex sea state, the contrast between ship and sea descends in polarimetric entropy space that deduced from coherence matrix eigenvalue decomposition. The analyses of the polarimetric target decomposition theory and target’s scattering mechanism illustrate that parameters come from target decomposition describe the difference between targets and background from different point of view. The combination use of them promotes the detection of target in SAR imagery. However, each parameter has its own diverse significance in the practical detection problem. Therefore, this paper proposes an SVM classification method to detect ships in PolSAR (Polarimetric SAR) imagery. Firstly, the method constructs a feature vector consists of several decomposition parameters; and then, different decomposition parameters are weighted according their essentiality in the SVM classifier; ships are classified from sea background and other false alarms by the classifier in the end. Experiment results illustrate that the method detects ship targets more precisely and reduces false alarms effectively.