Object Detection Using a Background Anomaly Approach for Electro-Optic Identification Sensors.

摘要

Electro-optic identification (EOID) sensors are transitioning to the fleet and will be used as a short-range identification tool for mine-like contacts from long-range sensors. The present operation of the EOID sensors uses an operator for identification. Whereas the human operator is unparalleled in detecting and recognizing objects of interest, there are still some limitations which may be needed to distinguish between mine types, such as differentiating a 68 inch object from a 72 inch object in a still image or moving waterfall display. To help overcome some of these weaknesses and improve the mine identification process, computer aided identification (CAI) and automatic target recognition (ATR) algorithms are being developed. In addition to building a foundation towards the long-term goal of fully autonomous operation, these algorithms can be used to queue operators of potential mine- like objects within the data as well as to segment and compute vital geometric information Eon manually flagged objects of interest. The operator can then use this supplementary information for a more accurate identification. The near- term objective is to develop and implement these CAI/ATR algorithms into a real- time console and/or a post mission analysis (PMA) tool that can be used in the FY05 Organic Mine Warfare future naval capability (FNC) demonstration.