HADES

Real-Time Aerial Object Detection & Tracking

HADES is a real-time aerial object detection and tracking platform that uses AI-powered computer vision to detect and track moving aerial vehicles. Processing video at up to 30 FPS, HADES delivers high-confidence detections and actionable metadata to operator consoles via secure interfaces. Designed for situational awareness and both military and civilian applications, the system enforces human-in-the-loop decisioning, full audit trails, and strong security controls.

Highlights

  • • Real-time AI detection and multi-target tracking (up to 30 FPS)
  • • Low-latency pipeline (typical < 100 ms) with secure APIs
  • • Human-in-the-loop decisioning with full audit trail
  • • EO/IR support, PTZ control, and radar/ADS-B integration
  • • Scalable tracking capacity with modular deployment
  • • Robust security (TLS, JWT, RBAC) and GDPR-aware processing

Project Info

  • Category: AI, Computer Vision, Multi-Object Tracking
  • Scope: Real-time aerial object detection, multi-target tracking, and secure metadata delivery
  • Processing speed: Up to 30 FPS (real-time video pipeline)
  • Core algorithms: CNN-based detection; Kalman/DeepSORT (or equivalent) multi-object tracking
  • Typical accuracy: Targeted mAP ≥ 0.8 after calibration (field-dependent)
  • Latency: Image → detection → metadata: typical < 100 ms (configuration dependent)
  • Supported sensors: HD EO, IR/night mode, PTZ; adapters for radar, ADS-B, other telemetry
  • Communication protocols: HTTPS/REST, MQTT, TCP/UDP; optional serial (RS-232/RS-422) with encryption
  • Output format: JSON per detection, RTSP/HLS for video, H.264/H.265 recording
  • Tracking capacity: Scales with compute/network — simultaneous N targets (configurable)
  • Storage & archiving: Local/remote archiving, event DB, exportable logs
  • Security: TLS, JWT auth, RBAC, tamper-evident audit logs
  • Environmental rating: Typical hardware ranges: −20°C to +55°C (enclosure dependent)
  • Compliance & privacy: Configurable retention, GDPR-aware processing, audit trails

Technical Specs

  • • Processing speed: up to 30 FPS
  • • Latency: typical < 100 ms (pipeline dependent)
  • • Accuracy: targeted mAP ≥ 0.8 after calibration
  • • Algorithms: CNN detection + Kalman/DeepSORT tracking
  • • Outputs: JSON/RTSP/HLS; recording H.264/H.265
  • • Sensors: EO/IR/PTZ; radar & ADS-B adapters

Architecture

Ingest → AI detection → multi-target tracking → metadata packaging → secure delivery to operator consoles. Enforces human-in-the-loop workflows, role-based access, and tamper-evident audit logs.

  • • Modular video ingest with EO/IR and PTZ control
  • • CNN inference and online tracker fusion
  • • Secure APIs (REST/MQTT) and stream outputs (RTSP/HLS)
  • • Event database, storage/archiving, and audit trails