Object Detection and Classification for Small Objects in/on Water

“Technologies with object detection capabilities sensitive enough to identify fast moving objects floating on the surface of water (or just below the surface) can offer important capabilities to understand the environment, upstream activities, and add to competencies for future military readiness and the warfighter.” To address this need, UHV and UTARI proposes a multi-pronged approach; (i) use of a Long Wave Infrared Imaging Polarimeter (LWIRP) for imaging objects on water, (ii) test other sensors such as radar and sonar (iii) leverage UHV’s previous success with object detection and classification neural networks for military applications and (iv) investigate a machine learning / artificial intelligence (ML/AI) based sensor fusion approach to detect threats quickly. UHV has contacted Polaris, manufacturer of the LWIRP camera, and they have expressed a high interest in working with UHV to develop object detection and classification neural networks for use with their existing lines of Pyxis LWIRP cameras. This strategy maximizes the chances of success by (i) using a commercially available product for the LWIRP, the Pyxis model, which has already been demonstrated for improved imaging on water surfaces (ii) using a large array of state-of-the-art image segmentation methods to determine architectures that are most suited to perform this task (iii) mitigating the risk associated with image segmentation methods on highly successful neural networks which UHV has demonstrated in real world conditions (iv) proposing a sensor suite enabling detection under different weather conditions  (v) proposes a novel neural network architecture for integrating datasets from multiple sensors in real time.