Charles River Analytics Advances Intelligent Video Data Management

Charles River Analytics, a GRVTY company, is developing advanced tools to improve the efficiency and usefulness of large-scale video data collection. Its Real-Time Evaluation, Compression, and Organization of Recorded Data using Extensible Rules (RECORDER) system is designed to help organizations extract greater value from high-volume video streams. The company has received a contract from the Naval Sea Systems Command valued at up to $1.75 million to further develop and expand this prototype capability.

Modern camera systems generate substantial quantities of footage, much of which contains limited operational value. This challenge is particularly pronounced for the U.S. Navy, where numerous onboard cameras rapidly consume storage capacity. As a result, critical information may be overlooked or lost among less relevant recordings. RECORDER addresses this issue by introducing a more adaptive and selective approach to video capture and storage, improving both efficiency and data relevance.

Conventional motion-detection approaches, similar to those used in residential security systems, are not suitable in maritime environments due to continuous vessel and ocean movement. Similarly, reducing recording fidelity to conserve storage can compromise the usefulness of captured imagery. Instead, RECORDER approaches the challenge as an optimization problem, balancing what data is recorded, how it is prioritized, and how storage resources are utilized. Ross Eaton, Director of Marine Systems at Charles River Analytics and Principal Investigator on the RECORDER program, explains, “What we really have is an optimization problem, determining what is getting recorded and saved, how to sort the value of the data that’s saved while minimizing the size.”

The system is built around three core objectives. As Eaton states, “to create a way for users to specify the recordings that might be of interest, to share the reason why these things are interesting so we can understand what kinds of compression we can apply, and to use that information to maximize our usage of the available disk space.” These principles guide RECORDER’s ability to tailor data capture according to mission requirements.

By identifying what content is important and why, RECORDER can determine the appropriate level of detail and recording parameters. For instance, when monitoring distant vessels near the horizon for classification purposes, full-motion, high-frame-rate recording may not be necessary. Adjusting these parameters allows for more efficient use of storage without compromising mission objectives.

Operators can define specific ontologies, or categories of interest, such as vessels within a defined proximity or particular marine life behaviors like whale tail flukes. Additional constraints, including time-of-day parameters, can further refine recording criteria. This targeted approach ensures that high-value data is prioritized while reducing the capture of extraneous information.

RECORDER employs multiple data compression strategies, including spatial and temporal methods, along with bit rate reduction tailored to video streams. These techniques enable the system to retain meaningful content while significantly increasing available storage capacity. Eaton notes, “We’re not trying to eliminate the data. We’re increasing the amount of data that is interesting and increasing the available disk space by 90%.”

The system is also being developed for integration with future iterations of the Awarion® platform, enhancing its applicability across both government and commercial sectors. Potential use cases extend to industries such as offshore wind, where continuous visual monitoring generates large datasets requiring efficient management.

The underlying framework of RECORDER supports extensible ontologies, allowing the system to be adapted for a wide range of domains beyond maritime operations. Eaton explains, “We’re putting together a system where we can swap out the maritime ontology for a sports ontology or a speech ontology. Whatever you envision it to be, so you can build a system that is relevant in a variety of domains.”

Currently in Phase II development, RECORDER is being scaled to manage increasingly large volumes of video data. Eaton highlights the broader significance of the effort, stating, “Scaling a prototype to include a large amount of video is an exciting opportunity to address a growing problem that many organizations are eager to solve”. He adds, “We’ve got a really interesting problem and a solution that we feel is going to be good. We’re introducing the concept of compressing videos into a format that’s easy to access and engage with, rather than letting them sit unused, metaphorically collecting dust on a disk.”

Charles River Analytics continues to expand its capabilities in computer vision, perception, robotics, and autonomy, with RECORDER representing a focused effort to improve how organizations capture, manage, and utilize video data.

Find out more on the Charles River Analytics website.