Helmet Mounted LWIR 3D Perception System with Machine Learning

We propose a robust head mounted low SWaP-C LWIR 3D perception system to capture a depth map of the environment and provide 3D data that can be successfully shared over the network to increase the group situational awareness. This subsystem will be fused with and enhance existing AR technology, such as ARC4. The system will allow dismounted Battlefield Airmen to operate in hostile environments that are completely devoid of ambient illumination, where use of light sources (including LIDAR) would result in detection by enemy combatants. Airmen would be able to detect information about the surrounding complex environments, without needing to divert attention from their mission to interpreting and collating 2D images. Elphel technology utilizes larger numbers of individually small sensors and deep subpixel resolution critical for accurate ranging without increase of SWaP-C. These innovations enable low weight and high performance. Frequency-domain image rectification, aberration correction and phase correlation combined with the trained deep neural network yield disparity measurements with 0.05 pixel accuracy. Developed for the head mounted operation, the proposed system could be modified for adversarial environment navigation by UGV and for driver assistance in manned vehicles.