Research on an Improved YOLO+SLAM+EKF Algorithm Framework for Underwater Obstacle Recognition, Localisation and Path Planning

DOI: https://doi.org/10.62517/jike.202604209

Author(s)

Yanjie Li

Affiliation(s)

School of Electronic Information and Control Engineering, Guangzhou University of Software, Guangzhou, China

Abstract

Addressing challenges in underwater environments-including insufficient target recognition accuracy due to light attenuation and suspended particle interference, cumulative drift in single-SLAM positioning, and poor obstacle avoidance robustness from path planning lacking semantic support-this study proposes an integrated algorithmic framework based on enhanced YOLOv8 combined with semantically augmented ORB-SLAM3+EKF fusion positioning.This framework enhances detection accuracy for underwater obstacles (debris, organisms, equipment) by introducing underwater-specific attention mechanisms and data augmentation strategies to YOLOv8. Detection results are embedded into ORB-SLAM3 for semantically enriched mapping. Combined with an EKF algorithm for fusing SLAM poses, precise positioning is achieved. Finally, a globally optimal path is generated based on the semantic grid map.Experimental results on the Trash-ICRA19 dataset demonstrate that the framework achieves an 81.2% target detectionmAP@0.5, with positioning RMSE controlled within 1145.25cm and positioning drift rate at 22.93%. This provides effective technical support for autonomous underwater vehicle operations in marine environments.

Keywords

YOLOv8n; Object Detection; Path Planning; SLAM; Kalman Filter

References

[1] United Nations Environment Programme (UNEP). Marine Plastic Pollution: A Global Review of Sources, Impacts and Policy Responses [R]. Nairobi: UNEP, 2023. [Report No.: UNEP/DEPI/WEH/2023/1; Online access: https://www.unep.org/resources/report/marine-plastic-pollution-global-review-sources-impacts-and-policy-responses] [2] Cong, Y., Gu, C., Zhang, T., & Gao, Y. (2021) Underwater robot sensing technology: A survey. Fundamental Research, 1: 337–345. [3] Zhou, H., Kong, M., Yuan, H., Pan, Y., Wang, X., Chen, R., Lu, W., Wang, R., & Yang, Q. (2024) Real-time underwater object detection technology for complex underwater environments based on deep learning. Ecological Informatics, 82: 102680. [4] Merveille FFR, Jia B, Xu Z, Fred B. Enhancing Underwater SLAM Navigation and Perception: A Comprehensive Review of Deep Learning Integration. Sensors (Basel). 31 October 2024;24(21):7034. doi: 10.3390/s24217034. PMID: 39517928; PMCID: PMC11548088. [5] Li, G.; Wang, J.; Dai, J.; Zhao, T.; Chen, D.; Chen, C. Adaptive Path Planning for Autonomous Underwater Vehicle (AUV) Based on Spatio-Temporal Graph Neural Networks and Conditional Normalising Flow Probabilistic Reconstruction. Algorithms 2026, 19, 147. https://doi.org/10.3390/a19020147 [6] Fulton M S, Hong J, Sattar J. Trash-ICRA19: A Bounding Box Labeled Dataset of Underwater Trash [Database/Online]. Minneapolis: Data Repository for the University of Minnesota, 2020. https://doi.org/10.13020/x0qn-y082. [Data volume: 5,700 images; labelled categories: debris (plastic/metal/fishing nets etc.), biological objects, ROV equipment; data format: JPEG+XML (Pascal VOC annotations)] [7] Wang, Z., Yu, Z., & Zheng, B. (2025). YOLO-NeRFSLAM: Underwater object detection for the visual NeRF-SLAM. Frontiers in Marine Science, 12, 1582126. https://doi.org/10.3389/fmars.2025.1582126 [8] Li, C., Liu, W., et al. (2025). SU-YOLO: Spiking neural network for efficient underwater object detection. arXiv preprint arXiv:2503.24389. https://arxiv.org/abs/2503.24389 [9] Lin, Y., Zhang, Y., Yang, D. et al. DLFE-YOLO: an enhanced framework for underwater object detection based on YOLOv8. Intell. Mar. Technol. Syst. 3, 37 (2025). https://doi.org/10.1007/s44295-025-00088-x [10] Peng, Y., Hong, Y., Hong, Z., Chui, A. P. Y., & Wu, J. (2025). AquaticVision: Benchmarking visual SLAM in underwater environment with events and frames. arXiv preprint arXiv:2505.03448. https://arxiv.org/abs/2505.03448 [10] Glenn J, Ali G, Steve G. YOLOv8: Real-Time Object Detection at Scale [J]. arXiv preprint arXiv:2302.05790, 2023. [Online access: https://arxiv.org/abs/2302.05790; Open-source repository: https://github.com/ultralytics/ultralytics] [11] Yuan, X., Martínez-Ortega, J. F., Sánchez Fernández, J. A., & Eckert, M. (2017). AEKF-SLAM: A new algorithm for robotic underwater navigation. Sensors, 17(6), 1343. https://doi.org/10.3390/s17061343 [12] Heshmat, M., Saad Saoud, L., Abujabal, M., Sultan, A., Elmezain, M., Seneviratne, L., & Hussain, I. (2025). Underwater SLAM meets deep learning: Challenges, multi-sensor integration, and future directions. Sensors, 25(11), 3258. https://doi.org/10.3390/s25113258 [13] Fulton, M., et al. (2019). Trash-ICRA19: A dataset for underwater trash detection. In Proceedings of the IEEE International Conference on Robotics and Automation (ICRA) (pp. 5752–5758). IEEE. https://doi.org/10.1109/ICRA.2019.8794130