Automated Swimmer Recongition Algorithms for the HarborGuard Intrusion Detection System

The remote identification of surfaced swimmers in a harbor environment is a critical requirement for a comprehensive waterside security system. This requirement can be clearly divided into detection and classification objectives. Detection in this context means identifying the presence of a target on the surface of the water. Classification means discriminating between different types of targets - specifically, discriminating between swimmers on the surface and non-swimmer targets. In Phase I, Barron Associates, Inc. collaborated with L-3 Communications Klein Associates to develop highly-effective baseline automatic swimmer recognition (ASR) algorithms. The baseline ASR algorithms were demonstrated using a diverse set of radar data that included images of swimmers and other small targets, such as birds, buoys, and floats. Barron Associates extracted features from the radar images and classification algorithms to identify swimmers. The ASR algorithm achieved perfect classification of the radar targets in closed-set tests and, importantly, achieved excellent classification performance in open-set tests. The objectives of this Phase II SBIR effort are to: (1) implement the algorithms and supporting software for online, near real-time operation; (2) evaluate the algorithms in adverse environmental conditions; and (3)enhance the algorithms to incorporate additional sensor data, such as multiple radars, video/thermal images, and sonar signals.