Vision-primary autonomous drone systems — no radar, no GPS dependency, no RF signals to jam. Starting with counter-drone defense, expanding to every domain where autonomous aerial intelligence creates value.
Existing autonomous systems rely on GPS, radar, or RF signals — all jammable, all infrastructure-dependent. Anor builds vision-primary from the ground up. Cameras see. Edge AI decides. The system acts. No external dependencies. This is how the aerial internet gets built.
No RF signal to defeat. No GPS to spoof. Vision systems operate on light. You cannot jam light.
Radar needs towers. GPS needs satellites. Vision needs a camera and power. Deploy anywhere in minutes.
Hardware, CV software, flight control, and C2 interface — all designed in-house. One product. No integration risk.
One CV stack powers Kestrel drones and ground vehicles. Every advancement benefits both lines simultaneously.
Pipelines. Power lines. Bridges. All inspected manually today — slowly, expensively, dangerously. Kestrel Scout changes that.
Explore Kestrel →Every existing interceptor was built radar-first. Kestrel Interceptor was built vision-first. No radar. No RF. It sees the threat and closes.
Explore Kestrel →Named after David Marr — the MIT neuroscientist who defined the computational basis of biological vision. The proprietary CV stack powering every Anor system. See everything. Miss nothing.
Real-time video. No radar. No lidar. Vision only. The raw material of autonomous intelligence.
30fps detection at >95% mAP. Runs on NVIDIA Jetson Orin at 10W. Classify targets, threats, and obstacles in a single forward pass.
Multi-frame target locking with Kalman filter prediction. Deep appearance descriptors maintain identity through occlusion and speed changes.
Vision converts to flight commands. Pixhawk executes in milliseconds. LLM-powered C2 interface for operator control.
Multiple units share target data. Distributed intelligence. No single point of failure. Swarm-native from day one.
"Vision is the process of discovering from images what is present in the world, and where it is."
David Marr · Vision, 1982 · MIT AI LabTarget ID, depth, velocity, trajectory. Define what the system needs to extract from every scene to accomplish the mission.
YOLOv8 + DeepSORT + LLM translation. Fast enough for real-time flight decision-making at the edge.
NVIDIA Jetson Orin, Sony IMX477, Pixhawk. Hardware chosen for field conditions, not lab conditions.
Two variants. One CV stack. From scanning a pipeline to intercepting a drone swarm — Kestrel sees, decides, and acts without a human in the loop.
Pipelines. Power lines. Bridges. All inspected manually today — slowly, expensively, dangerously. Kestrel Scout changes that. Vision-primary autonomous inspection that finds problems before they become failures.
Every existing interceptor was built radar-first. Kestrel Interceptor was built vision-first. No radar infrastructure. No RF signal to jam. It sees the threat, locks on, and closes — autonomously.