2X CVBS-Video + Thermal FPV Edge AI Boxboard-Videoanalyse

2X CVBS-Video + Thermal FPV Edge AI Boxboard-Videoanalyse

Beschreibung

This is a compact dual-CVBS edge AI video tracking module powered by the RK3588S processor. Designed for FPV drones and UAV platforms, it supports real-time object detection, Erkennung, and tracking with visible and thermal video inputs, delivering high accuracy, geringe Wartezeit, und flexible Anpassung.

---

Überblick

Das 2237 Dual Analog Edge AI Video Tracking Module ist eine Hochleistungsmaschine, ultra-compact embedded vision system designed for real-time intelligent video analytics in size- and weight-onstrained environments such as FPV drones and small UAV platforms.

Built around a domestic high-performance AI chipset (RK3588S) and integrated with proprietary intelligent image processing algorithms, das 2237 enables advanced capabilities including target detection, Erkennung, locking, and tracking based on real-time video streams from visible-light or infrared (Thermal-) Sensoren.

Das Modul unterstützt dual CVBS analog video inputs, along with an optional MIPI interface, allowing flexible integration with a wide range of imaging sensors. By leveraging deep learning-based computer vision algorithms, the system can automatically detect and identify targets such as humans and vehicles, and perform both automatic and manual target locking and tracking.

A key feature of the 2237 is its ability to output real-time target deviation data (miss distance / offset from the center of the field of view), enabling precise situational awareness and control feedback for downstream systems such as flight controllers or targeting mechanisms.

Designed specifically for small aerial platforms, the module offers a compact footprint (40 × 55 × 10.67 Millimeter) und ein leichtes Design (~24.5g), minimizing payload impact while maintaining powerful onboard processing capabilities. Its low power consumption (<6W) ensures efficient operation in battery-powered systems.

The platform is fully open for secondary development, providing SDKs and flexible interfaces that allow customers to quickly integrate, anpassen, and deploy their own AI models and application logic.

---

Eigenschaften

1. Dual Analog Video Input with Multi-Sensor Fusion

• Unterstützt 2-channel CVBS analog video input (PAL format)
• Optional 1-channel MIPI input (anpassbare)
• Enables integration of:
  • Visible-light cameras
  • Infrared/thermal cameras
• Simultaneous multi-stream processing for enhanced detection accuracy

2. Advanced AI-Based Detection and Tracking

• Echtzeit AI object detection and recognition
• Supports target categories such as:
  • Humans
  • Fahrzeuge
• Automatic or manual target locking and tracking
• Ausgänge target offset (miss distance) relative to frame center

3. Multiple Tracking Modes

• Centroid (center-of-mass) Verfolgung
• Template matching tracking
• Verfolgung mehrerer Ziele (bis zu 64 targets)
• Intelligent hybrid tracking

Flexible tracking initialization methods:

• Target ID selection
• Mouse/click selection
• Bounding box selection

4. Robust Tracking Performance

• Adaptive tracking window (gate) steuern
• Automatic re-detection after target loss
• Strong resistance to:
  • Background interference
  • Occlusion
• Minimum detectable target size:
  • 2×2 pixels (centroid tracking)
  • 8×8 pixels (correlation tracking)

5. High-Performance AI Hardware Platform

• Angetrieben von RK3588S AI processor
• 8nm LP process technology
• Octa-core 64-bit CPU (up to 2.4GHz)
• Integrated NPU with up to 6 TOPS AI performance
• Supports mainstream deep learning frameworks

6. Real-Time High-Speed Processing

• Tracking frame rate: bis zu 60Hz (up to 100Hz depending on input)
• Recognition frame rate: 25Hz
• Tracking accuracy: ≤1 pixel
• Tracking speed: ≥64 pixels/frame

7. Rich Customizable Control Functions

• Manual/automatic adjustment of tracking window size and position
• Adjustable target extraction thresholds
• Black/white target selection (centroid mode)
• Crosshair overlay with position adjustment
• Digital zoom:
  • 1× to 4×
  • Step size: 0.1
• Picture-in-picture (PiP) display control
• OSD -Overlay:
  • Target info
  • Tracking status
  • Gate display
• Real-time fine adjustment of tracking target position and size
• Built-in self-test functionality

8. Flexible Communication & Integration

• Supports multiple communication protocols:
  • RS422
  • RS232
  • UART
  • TCP / UDP
• Standard Ethernet interface (100Mbps)
• Default RTSP video stream output
• Custom protocol support for integration

9. Kompakt, Leicht, and Energy Efficient

• Maße: 40 × 55 × 10.67 Millimeter
• Gewicht: ~24.5g
• Energieverbrauch: <6W
• Betriebsspannung: 12V DC ±2V

---

Technische Daten

KI & Verarbeitung

• Prozessor: RK3588S
• Zentralprozessor: Octa-core 64-bit (up to 2.4GHz)
• NPU: 6 SPITZEN
• AI Framework Support: TensorFlow, PyTorch, ONNX

Video Eingang

• CVBS-Eingang: 2 Kanäle (PAL format)
• Lösung: 720 × 576
• Bildrate: 25Hz
• MIPI Input: 1 Kanal (anpassbare)

Video-Ausgang & Codierung

• Encoding Formats: H.264 / H.265
• Streaming -Protokolle:
  • RTSP
  • UDP
  • RTMP
  • HLS
• Konfigurierbare Parameter:
  • bitrate
  • GOP (I-frame interval)
  • QP value

Tracking Performance

• Tracking Frame Rate: up to 60Hz (max 100Hz)
• Minimum Target Size:
  • 2×2 pixels (centroid)
  • 8×8 pixels (correlation)
• Tracking Accuracy: ≤1 pixel
• Maximum Targets: ≤64
• Minimum Contrast: ≥5%

Detection Performance

• Detection Frame Rate: 25Hz
• Minimum Detection Target: 10×10 pixels
• Detection Accuracy:
  • mAP@0.5 = 0.616 (VisDrone2019 benchmark)

Supported Algorithms

• Verfolgung:
  • Centroid tracking
  • Correlation tracking
  • Verfolgung mehrerer Ziele
  • Intelligent tracking
• Erkennung:
  • AI object detection
  • Bewegungserkennung
  • Template matching

Schnittstellen

• Ethernet: 100M
• Seriell:
  • 1× RS422
  • 1× UART
• Video Eingang:
  • 2× CVBS
  • 1× MIPI (optional)

Lager

• Standard: 32GB
• Expandable storage supported
• Video storage: ≥2 hours (beim 4 Mbps bitrate)

Leistung & Mechanisch

• Energieverbrauch: <6W
• Stromspannung: 12V DC ±2V
• Maße: 40 × 55 × 10.67 Millimeter
• Gewicht: ~24.5g

Umweltfreundlich

• Betriebstemperatur: -20° C bis +60 ° C.
• Lagertemperatur: -20°C bis +65°C

---

Anwendungen

1. FPV-Drohnen & UAV-Systeme

Designed specifically for small aerial platforms, Aktivieren:

• Real-time target tracking
• Autonomous navigation assistance
• Visual targeting systems

2. Überwachung & Sicherheit

• Border monitoring
• Perimeter security
• Mobile surveillance platforms

3. Suche und Rettung

• Human detection in low visibility
• Thermal-assisted rescue missions
• Disaster response

4. Industrielle Inspektion

• Überwachung der Infrastruktur
• Pipeline and power line inspection
• Equipment anomaly detection

5. Verteidigung & Tactical Systems

• Target tracking and locking
• Situational awareness
• Lightweight embedded vision for tactical drones

---

Anpassungsoptionen

Hardwareanpassung

• MIPI camera interface adaptation
• Connector and pin definition customization
• Power input configuration
• Storage expansion

Softwareanpassung

• Custom AI model deployment
• Detection category expansion
• Tracking algorithm tuning
• OSD interface customization

Funktionsanpassung

• Custom tracking logic
• Target classification rules
• Multi-target prioritization
• Communication protocol customization

Integration Support

• SDK/API access
• Integration with flight control systems
• Support for third-party platforms

---

FAQ

Q: Can this 2× CVBS + Thermal FPV Edge AI video tracking module be directly integrated with existing flight controllers (z.B., open-source autopilots)?

EIN: At the current stage, the module does not support direct integration with open-source flight control systems.

The system is designed to perform onboard video analytics and tracking, and it primarily outputs target deviation data—that is, the miss distance or offset of the tracked target relative to the image center (in pixel or angular coordinates). While this output can be used as an input for external control systems, there is no native protocol or direct interface available for plug-and-play connection with common open-source flight controllers (such as PX4 or ArduPilot).

1. What types of cameras are supported?

Das Modul unterstützt dual CVBS analog cameras and an optional MIPI-Kameraschnittstelle, enabling both visible and thermal imaging.

2. Kann es mehrere Ziele gleichzeitig verfolgen??

Ja, es unterstützt multi-target tracking of up to 64 Objekte.

3. What happens if the target is lost?

The system automatically performs re-detection and reacquisition, ensuring robust tracking continuity.

4. Is it suitable for small drones?

Ja. Es ist speziell dafür konzipiert lightweight UAV and FPV platforms, with minimal size and power requirements.

5. Can I customize AI models?

Ja. The platform supports secondary development and custom AI model deployment.

6. What tracking methods are available?

• Centroid tracking
• Template matching
• Verfolgung mehrerer Ziele
• AI-based tracking

7. Does it support real-time streaming?

Ja. Es unterstützt RTSP, RTMP, UDP, and HLS streaming protocols.

8. Wie hoch ist der Strombedarf??

The module operates at 12V DC (±2V) with power consumption below 6W.

9. Can the tracking parameters be adjusted?

Ja. Users can adjust:

• Tracking window size
• Thresholds
• Zoomen
• Target parameters

---

10. Is the product ready for integration?

Ja. Es bietet SDKs, APIs, and flexible interfaces, enabling fast integration into existing systems.

---

Abschluss

Das 2237 Dual Analog Edge AI Video Tracking Module delivers a powerful combination of high-performance AI processing, dual video input capability, advanced tracking algorithms, and ultra-compact design.

With its ability to perform real-time detection, Erkennung, and tracking, along with flexible customization and robust environmental performance, it is an ideal solution for next-generation UAV systems and embedded vision applications.