Edge-Based Facial Recognition

Edge Analytics with Limited Bandwidth Connection & Multi-Tier Processing
US Army Logo - Reversed
Army personel in front of computer analyzing data


To optimize the utilization of security personnel, the U.S. Army deploys remote surveillance systems. Challenged by limited bandwidth, the Army sought a system to process data at the edge to identify potential bad actors. With HarperDB, the Army could extract and transfer images and short video clips instead of a continuous video stream, reducing overall bandwidth requirements while ensuring efficient data distribution to mobile and central command systems for final identification.

HarperDB Edge

HarperDB has partnered with Command Cyber Solutions to provide the Army with an innovative solution that will enhance the Army’s tactical edge and data orchestration capabilities. Command Cyber will leverage HarperDB to develop an edge-based analytics platform that can capture facial recognition data and distribute it to other HarperDB nodes for real-time alerting and detection.

HarperDB will be deployed on the extreme edge, a mobile command center, and a server to not only collect and orchestrate data throughout the nodes, but also propagate rules engines and AI models from the command center to the edge to update logic and thresholds on the fly. This will allow the Army to adjust the priority level of the faces detected from the command center in case a high threat (person of interest) is found. In addition, priority levels of faces detected will be automatically upgraded when they are captured in high sensitivity areas.

The Army is looking for a system to increase situational awareness through computing and decision support on the extreme edge. HarperDB enables this by providing a single holistic solution that makes data sync and management easy. HarperDB’s bidirectional data movement enables the collection and movement of data and logic in real time, shifting decision making throughout the network as needed.

In addition, HarperDB’s Custom Functions allow for user-authored code to run within HarperDB, with direct access to core HarperDB methods. Comparable to AWS Lambdas, or old-school stored procedures, Custom Functions are simple to create, easy to maintain, and can be deployed across the entire network with a single click.

Visual showing three-tiered architecture that spans from the battlefield to the command network.
Conceptual System Architecture of Solution

Why the Army Chose HarperDB

- Bi-Directional Movement of AI Models & Rules Engines
- Custom Functions Deployed from Edge to Cloud
- Edge Data Filtering
- Enhanced Security
- Collapsed Data Stack
- Cloud Agnostic
- Single User Interface to Manage All Instances

3-Tier Approach

Tier 1 - Jetson Nano (Extreme Edge)

Jetson Nano with IMX219 Camera. OpenCV/GPU-based deep-learning script collects facial recognition data and inserts it into HarperDB, which replicates that data to the mobile command and desktop in real time. Simultaneously, a rules engine allows for editing and applying logic to collected data, adding context or value. These rules are down-propagated from the upper tiers, and can be hot loaded without restarting the unit. Faces found are displayed with a color-coded priority system based on logic applied by the rules engine.

Tier 2 - Laptop (Mobile command)

Provides real-time view into data synchronized by HarperDB. Details can be added to, or modified for, individual "faces" that may be of interest. Those updates are sent to the edge to be applied on future
detection events.

Tier 3 - Desktop/Server (Base Level)

Has all capabilities of the Mobile Command tier, along with processing power to double check the results or run further models against shared images. HarperDB Studio can easily run historical analysis on the data to improve operational efficiency.

Command Cyber Solutions is an IT consulting company specializing in Cyber Security Services and IT Support.
Image collage of person talking on the phone
Image collage of person talking on the phone

Request a Call

Consult a solutions architect for tailored recommendations and insights.
Request received! We'll reach out shortly. In the meantime, check out our Dev Center for helpful development resources.
Go to Dev Center
Uh-oh! It seems your submission failed to submit. Please try again.