Improved surface swimmer detection through multimodal data fusion

Abstract

Waterborne intruder detection includes many new challenges not seen in land environments. One area of these challenges is the detection of surface swimmers. Swimmers, whose bodies are partially in air and partially submerged, have significantly reduced target strength (TS) for radar and sonar systems compared to intruders fully in air or fully submerged. This reduced TS results in more frequent missed detections or, if detection threshold is widened, increased nuisance alarms. Depending on sea state, a swimmer is also able to blend in with wave noise, making detection even more difficult. We present a method for improved surface swimmer detection in marine environments by fusing data from several sensor systems in both air and water domains to isolate a swimmer's signature from uncorrelated events. This system, tested in Dec 2011 in St. Petersburg FL, produced data indicating significantly improved detection over using any single system. By widening detection threshold of each sensor's detection algorithm but fusing data of each system together, more potential targets can be processed without the risk of increasing nuisance alarms. This work holds the potential to improve the security of several types of water-dependent assets, like commercial harbors, Navy or Coast Guard bases, and nuclear and other water-cooled power plans, and offshore oil platforms.