AMMOD Detection and AITT: Hybrid AI for the TRIDENT ROV/AUV Platform

AMMOD Detection and AITT

Hybrid AI Detection and Target Tracking for TRIDENT ROV/AUV

Trident Subsea Systems is developing an AI architecture for the TRIDENT Mini ROV, combining autonomous object detection, operator-assisted target selection and persistent tracking.

Core Capabilities

• AMMOD — Automated Man-Made Object Detection

  • Detects potential artificial objects underwater.

  • Analyses shape, contrast, contours, motion and persistence.

  • Filters natural background clutter.

  • Can detect previously unseen objects without requiring semantic classification.

• AITT — AI Target Tracking

  • Maintains lock on a selected target.

  • Updates target position, centroid and confidence.

  • Supports centring, temporary-loss recovery and reacquisition.

AI Detection-to-Track Handover

• AMMOD detects and validates a persistent candidate.

• Target data is automatically transferred to AITT.

• The tracker receives:

  • bounding area;

  • centroid;

  • confidence;

  • persistence status;

  • visual target reference.

https://youtu.be/3lsEi22OoYQ

Workflow:Detection → Candidate Validation → Target Lock → Handover → Tracking → Reacquisition

Teach-to-Track

• The operator selects any detected object directly from the video feed.

• The system creates an immediate target reference.

• AITT starts tracking without model retraining or a predefined dataset.

• Useful for unknown, unclassified or mission-specific targets.

Workflow:Operator Selection → Target Reference → AITT Lock → Persistent Tracking

Distributed AI Architecture

• Onboard Coral TPU

  • real-time detection;

  • candidate generation;

  • persistence analysis;

  • autonomous target lock;

  • low-latency metadata processing.

• Surface AI Computer, aka Control Box (https://www.trident-subsea.com/post/introducing-the-trident-wifi-5g-control-box)

  • advanced multi-frame tracking;

  • higher-level scene analysis;

  • mission assistance;

  • additional processing when the tether connection is available.

Pool Validation

Testing has included:

• detection in cluttered underwater scenes;

• small and previously unseen objects;

• AMMOD-to-AITT automatic handover;

• Teach-to-Track selection;

• tracking under vehicle and target movement;

• temporary target loss;

• disturbance recovery and reacquisition.

  
  

Hybrid ROV/AUV Development

• Tether connected: conventional ROV operation with surface-assisted AI.

• Tether disconnected: autonomous AUV detection and tracking using onboard processing.

• Future sensor fusion will combine visual AI with sonar for detection beyond camera range.

Future detection chain:Long-Range Sonar Detection → Acoustic Approach → Visual Confirmation → AITT Tracking

The objective is a scalable system combining acoustic discovery, automated visual detection, operator-guided target selection and persistent AI tracking.