Target Recognition of Ships in Harbour Based on Simulated SAR Images Produced with MOCEM Software

A model-based approach to recognition of ship targets in SAR images has been investigated in a French-Norwegian cooperation project. Around 800 SAR images of four vessels have been collected from helicopter with the SAR system PicoSAR. A simulated reference set has been created with the SAR simulation software MOCEM, based on 3D models which are made from photographs and drawings. The Structural SIMilarity index (SSIM) is used as a comparison metrics for the real and simulated images. Classification results are promising, indicating a true potential for this target recognition approach, particularly if SAR images with multiple orientations can be collected.