LYNNTECH INC. — Department of Defense SBIR Phase I: A19-158

Description

The Army seeks to protect tactical and operational bases from adversarial agents intending to cause harm, gather intelligence, or pursue criminal activities. Existing perimeter or internal surveillance systems are based on collecting video feeds from EO/IR sensors. Personnel must be dedicated to monitoring such video feeds trying to detect rare anomalous events. Humans have been shown to be generally ill-suited to this type of task due to the onset of boredom or expectation-bias. The proliferation of sensor feeds and potential threats compounds this problem. Recent advances in Machine Learning and Artificial Intelligence have shown that computer vision detection, segmentation, and identification of threats through Predictive Analytics is feasible at a very competitive cost. However, such supervised learning systems require massive amounts of training data. Robustness is also required to limit the probability of false alarms. An additional challenge is the proliferation of methods that overcome automatic detection through adversarial examples. Lynntech’s iVIDGIL system will provide a reliable solution to overcome the limitations of surveillance systems by developing a novel robust solution in collaboration with world-class researchers. iVIDGIL will provide real-time detection and alerting of perimeter or internal threat and cue human operators’ attention where and when it matters the most.