Perceive
The eyes. Real-time detection and tracking at the edge — any camera, any platform.
Detect, classify, and track what matters, in real time, even in chaotic scenes.
Overview
Raven is an AI vision system for real-time object detection, classification, tracking, and situational awareness. Deploy it on drones, robots, fixed and PTZ cameras, vehicles, and edge devices. Raven holds onto relevant objects in dynamic, cluttered environments — supporting aerial surveillance, perimeter security, drone detection, vehicle tracking, search and rescue, and autonomous platform guidance.
Under the hood, Raven runs modern detector models (YOLO-family and TensorFlow Lite) with low-latency inference accelerated on NVIDIA Jetson, Hailo, and Coral edge hardware, as well as ARM-based platforms. It ingests USB, CSI, RTSP/IP, and PTZ camera feeds, and offers a selectable suite of tracking algorithms that trade raw performance against precision.
When a target is occluded or lost, configurable recovery logic re-acquires it; a motion-scan mode layers additional detections on top of the AI pipeline; and a command-and-telemetry interface — over UART or ROS 2 — lets Raven plug directly into autonomous platforms and the ODIN Command Center.
Mounted on a drone, Raven closes the loop from detection to action. As a full guidance, navigation, and control (GNC) system with ArduPilot integration, it locks onto a designated target and guides the platform to track, follow, or intercept it. Target features are stored at launch and automatically re-identified after stabilization, so lock survives hard manoeuvres. A line-of-sight (LOS) intercept solver continuously computes the intercept point and time-to-intercept from the platform's own velocity and the target's position, speed, and heading — the line of sight being the direct line from platform to target, which the solver uses to steer toward where a moving target will be, not where it was.
Mounted on a pan-tilt-zoom platform, Raven becomes a persistent sentry. It scans the sky or a designated area, automatically slews the camera to follow detected targets, and overlays compass bearing on the feed. A map-based target manager lets operators designate targets, assign target velocity, and see the live LOS intercept solution — target track, intercept point, and distance — on the map.
Under sustained load, Raven exceeds 100 frames per second at 640×640 on a low-power edge accelerator (Raspberry Pi 5 with Hailo-8), fusing video with onboard sensors through optimized C++ post-processing — with tunable profiles from indoor inspection to outdoor navigation.
Military & security use cases
- Real-time detection of drones, vehicles, aircraft, vessels, and personnel
- Drone-based target follow and intercept guidance
- PTZ sky-watch: persistent air surveillance from fixed positions
- PTZ-camera surveillance with automatic target tracking
- Drone-mounted ISR and reconnaissance
- Robotic security patrols
- Threat detection around bases, borders, and critical infrastructure
- Wide-area motion scanning and search and rescue in difficult terrain
What it delivers
- Edge AI processing for low-latency detection on-platform
- One vision stack across many cameras, accelerators, and platforms
- Resilient tracking with automatic recovery on occlusion
- Standards-based streaming to any client or network
- Direct integration with ODIN Command Center, ROS 2, and autopilots
- Stronger situational awareness with less manual monitoring
Key capabilities
Inside Raven.
- Real-time detection with YOLO-family and TensorFlow Lite models
- Multiple selectable tracking algorithms (performance vs. precision)
- Configurable lock-recovery on occlusion or target loss
- Edge acceleration on NVIDIA Jetson, Hailo, and Coral; ARM platforms supported
- Over 100 FPS at 640×640 on a Raspberry Pi 5 with a Hailo-8 accelerator
- Tunable configuration profiles per mission type
- Drone guidance via ArduPilot: track, follow, or intercept designated targets
- Line-of-sight (LOS) intercept solver — intercept point and time from own and target motion
- Target re-identification after launch: lock stored on command, re-acquired after stabilization
- PTZ sentry mode: automatic sky and area scanning with pan-tilt-zoom target following
- Map-based target management with target velocity and live LOS intercept display
- Inputs: USB, CSI, RTSP/IP, and PTZ cameras (ONVIF) with compass and OSD overlay
- Motion-scan detection layered on AI detections
- Low-latency streaming over RTSP, WebRTC, MJPEG, and UDP/RTP
- ROS 2 and autopilot integration; UART command & telemetry interface
- Model pipeline from training to deployment (YOLO → ONNX → Hailo / TensorRT)
Field view
Product FAQ
Raven — questions & answers
What can Raven detect and track?
Raven performs real-time detection, classification, and tracking of objects such as drones, vehicles, aircraft, vessels, and personnel — holding onto relevant objects in dynamic, cluttered scenes.
What platforms can Raven run on?
Drones, robots, fixed and PTZ cameras, vehicles, and edge devices. Raven uses edge AI processing for low-latency detection on-platform.
Can Raven guide a drone to follow or intercept a target?
Yes. Raven is a full guidance, navigation, and control (GNC) system with ArduPilot integration. It locks onto a designated target and steers the platform to track, follow, or intercept it. Target features are stored at launch and automatically re-identified after stabilization, so lock survives hard manoeuvres and temporary loss of view.
What is LOS intercept guidance?
LOS — line of sight — is the direct line between your platform and the target. Raven's intercept solver combines the platform's own velocity with the target's position, speed, and heading to continuously compute where the two paths meet: the intercept point and time-to-intercept. Instead of chasing where a moving target was, the platform steers toward where it will be. The solution is displayed live on the operator map, with target track, intercept point, and distance.
Can Raven run on a fixed PTZ camera instead of a drone?
Yes. Mounted on a pan-tilt-zoom platform, Raven works as a persistent sentry: it scans the sky or a designated area, automatically slews the camera to follow detected targets, and overlays compass bearing on the feed — suited to air surveillance and perimeter watch from fixed positions.
Does Raven integrate with command software?
Yes — Raven integrates directly with ODIN Command Center to feed detections into a unified operational picture and support decision workflows.
Rest of the system
Put Raven to work.
Tell us about your mission and we'll show you the system in context.