The authors have proposed an innovative method to navigate an autonomous underwater vehicle (AUV) for visual mapping of seafloor with high positioning accuracy without using any vision-based matching. The proposed method was implemented in the AUV Tri-Dog 1 and sea experiments were carried out at Tagiri vent field, Kagoshima bay in Japan (Maki et al., 2008).
Based on the success of the experiments, a series of dives was carried out at the same place. The AUV Tri-Dog 1 succeeded in 12 fully autonomous dives with a total duration of 29 hours. The vehicle took 9,288 pictures of the seafloor, keeping the altitude of 1.2 m with a surge speed of 0.08 m/s. A photomosaic of the seafloor was created by mapping 7,289 pictures based on the real-time estimates of the AUV state, without any pictorial correlation. The distributions of detailed features such as tube-worm colonies and bacteria mats are clearly shown. The photomosaic covers around 3,000 square meters. To the knowledge of the authors, this is one of the largest underwater photomosaic ever reported. The mapping accuracy was estimated to be 0.3 to 0.8 m based on the comparison of the photomosaic between dives.
