Underwater Robot Distance Measurement Using Deep Learning
Author Information
Author(s): Hu Qing, Wang Kekuan, Ren Fushen, Wang Zhongyang
Primary Institution: Sanya Offshore Oil and Gas Research Institute, Northeast Petroleum University
Hypothesis
Can a new underwater vehicle distance measurement system based on semantic segmentation and binocular vision improve measurement speed and accuracy?
Conclusion
The proposed system increases measurement speed by 30% while maintaining accuracy within 5% error.
Supporting Evidence
- The measurement speed increased by 30% while maintaining accuracy.
- The system effectively reduces the calculation burden by focusing on the object region.
- Underwater refraction and scattering can affect measurement accuracy.
Takeaway
This study shows that using smart technology can help underwater robots measure distances faster and just as accurately.
Methodology
The study used Deeplabv3+ for semantic segmentation and SGBM for stereo matching to measure distances.
Limitations
The accuracy decreases with longer distances due to environmental factors affecting image quality.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website