SMART – System for MARitime Targets Recognition and Retrieval

We propose to research and design a System for MARitime Targets Recognition and Retrieval. SMART will significantly advance ISAR/Electric-Optics-based ship detection and identification. Unlike existing solutions which focus on the image processing techniques only, the proposed solution integrates image classification, data simulation, and data fusion to achieve reliable and accurate target recognition and retrieval. One key innovation of SMART is its digital vessel model constructor. The vessel model can be constructed off-line and updated when new information concerning the target is available, and when working along with an image simulator, can be used to simulate reference images, which will be compared with the acquired images. Based on the image similarity, and classification results of the acquired images, SMART achieves more reliable and accurate target identification than conventional solutions. The vessel model also bridges the gaps caused by disparate sources and other factors such as view angle and illumination. Images from disparate sources, which are difficult to be compared directly, can be matched with each other through the vessel model. In this fashion, SMART achieves flexible and effective target retrieval. SMART will provide a suite of methods for image retrieval and model-based retrieval.