Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.18381038 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>Maritime autonomous systems require robust perception to address the high rate of human-error-caused accidents in the maritime domain. We present a hybrid detection framework combining YOLO11-based neural network detection on bird's-eye-view LiDAR projections with deterministic algorithms for identifying isolated floating objects and coast-anchored vessels. Our approach achieves significantly improved recall on real marina data compared to neural network detection alone. We further integrate AIS data using Kalman filtering and path matching, demonstrating high matching accuracy under realistic noise conditions. The system operates at real-time rates on modest GPU hardware, making it suitable for autonomous navigation applications. </p>