Real-time Evaluation, Compression, and Organization of Recorded Data using Extensible Rules (RECORDER)

Navy ships have many cameras for situational awareness, but they produce more video than humans can effectively review, and the gap is widening. Currently, all video is recorded, but the sparse relevant data (e.g., showing interesting targets or conditions) is practically impossible to find among the rest. This lost data compromises decision-making and delays machine learning and autonomy efforts. Charles River is facing similar challenges reviewing video from our maritime perception product, AwarionÖ. Thus, we propose Real-time Evaluation, Compression, and Organization of Recorded Data using Extensible Rules (RECORDER). RECORDER is an intelligent video recording system that ingests video and associated metadata in a variety of formats, encodes the metadata in an ontology, evaluates the metadata against a collection of user-defined rules (encoded with the same ontology) to select only the data of interest, and stores that data and its metadata at an appropriate level of fidelity in an easily searchable database for subsequent use by human analysts or machine learning algorithms. Users can add rules to expand what is recorded. RECORDER also provides a graphical user interface to review recent data and manually initiate recording, which it uses to suggest new rules to capture similar data in the future.