Object Detection and Recognition in Degraded Visual Environment (DAR-DVE)

Phelps2020 will adapt its USSOCOM developed SAM real-time Artificial Intelligence (AI) solution that performs accurate Electro-Optical persistent surveillance and automatic detection of threats (people and vehicles) in the presence of jitter and atmosphere turbulence. Our solution performs automatic and concurrent mitigation of several environmental degradations, including low light, fog, haze, blur, high dynamic range, backlighting, low resolution, and atmospheric distortions, which were shown to highly improve the accuracy of threat detection and reduce false alarms. Convolutional Neural networks-based Object detection (OD), the second component of DAR-DVE, has been traditionally used with EO sensors for lack of data in the IR. Phelps’ base product Crystal already extends EO-based OD to the majority of low light situations (dusk, dawn, and many nighttime situations). We will extend AI to the infrared bands for moonless nights, and retrain our existing AI OD algorithms for Geospatial AI to detect new classes, potentially including weapons, surface to air missile systems or other peer Integrated Air Defense Systems (IADS) in the visible and in the infrared, while associating the same objects in both bands for increased accuracy and improved intelligence.