研究目的

Investigating the development of a robust and effective stereo underwater visual odometry (VO) system to accurately localize autonomous underwater vehicles (AUVs) despite poor imaging conditions and inconsistent motion caused by water flow.

研究成果

The proposed stereo underwater VO system successfully overcomes the challenges of poor imaging conditions and inconsistent motion, providing accurate localization of AUVs. It outperforms existing VO systems in underwater environments and achieves comparable performance on the KITTI benchmark dataset. Future work includes improving underwater image quality for better feature extraction and reducing computational costs for real-time applications.

研究不足

The system's performance is affected by poor underwater imaging conditions and inconsistent motion. The computational cost is high, requiring around 2.8 seconds per frame on an i7 quad-core 2.5Ghz laptop, which may not be suitable for real-time applications without GPU acceleration.