One-dimensional Convolution Neural Network for Spectral Classification of Hyperspectral Data Collected from UAS Platforms

Physical Science Inc. (PSI) will adapt an existing, one-dimensional convolutional neural network (1D-CNN), for real-time automated identification of targets during SUAS-borne intelligence, surveillance, and reconnaissance (ISR) in contested environments based on data collected with long wave infrared (LWIR) hyperspectral imaging (HSI) sensors. Air Force stakeholders that stand to benefit from the proposed technology include the Tactical Off Board Sensing (TOBS) And Predator/Reaper Off-Board Sensing & Integrated Targeting (PROSIT) programs. The proposed technology is directly related to the C-SWaP EO/IR ISR Sensor Technology 20.3 Missions Focus Area. The Phase I program will demonstrate feasibility of the approach.