Sensor-based High Resolution and Long distance ML-driven object recognition

Visual sensors currently employ methods like convolutional neural networks to detect objects by analyzing images or video frames, identifying patterns and features. Unlike active sensors emitting signals, passive sensors capture existing stimuli like visible light or thermal radiation. Implementing passive sensors for object detection offers benefits such as lower energy usage, reduced interference, and enhanced stealth capabilities, particularly valuable for discreet and resource-efficient applications. In collaboration with the University of Missouri-Kansas City Data Science and Analytics Innovation Center (dSAIC), ProAILabDefence LLC, and Photon-X, Overmatch proposes a Phase II SHELL effort aimed at developing an innovative passive object recognition pipeline with extended range and higher resolution. To achieve this, the project team will execute on the following technical objectives: (1) Developing a Unity simulation environment for synthetic data generation; (2) Establishing a Unity-to-PXI pipeline; (3) Training AI/ML models on object data; and, (4) Conducting a Final Pipeline System Demonstration, achieving a confidence value of 75% or higher for real-world objects (Coffee Mug, Small Drone, Car). Additionally, a parallel focus will be on enabling augmented reality-based additive manufacturing in contested logistic environments, executing the following technical objectives: (1) Creating training content for 3D Metal Additive Printers operation; (2) Developing maintenance content for 3D Metal Additive Printers. Currently there is no modular, reconfigurable training and maintenance augmented reality tool that can be fielded to various environments, including remote/austere environments such as shipboard, that can be used to support training and maintenance for an array of printers, both traditional 2D and 3D printers.