A Universal Automatic Bottom Tracking Method of Side Scan Sonar Data Based on Semantic Segmentation

Determining the altitude of side-scan sonar (SSS) above seabed is critical to correct geometric distortions in images. Usually, a technology named bottom tracking applied estimate distance between and seafloor. However, traditional methods for often require pre-defined thresholds complex optimization processes, which make it difficult achieve ideal results underwater environments without manual intervention. In this paper, universal automatic method proposed based on semantic segmentation. First, waterfall images generated from SSS backscatter sequences are labeled as water column (WC) parts, then split into specific patches build training dataset. Second, symmetrical information synthesis module (SISM) designed added DeepLabv3+, not only weakens strong echoes WC area, but also gives network capability considering symmetry characteristic lines, most importantly, independent can be easily combined with any other neural networks. Then, integrated trained established Third, coarse-to-fine segmentation strategy well-trained model segment quickly accurately. Besides, fast line search algorithm further reduce time consumption tracking. Finally, validated by data measured several commonly used SSSs various environments. The show that accuracy 1.1 pixels mean error 1.26 standard deviation at speed 2128 ping/s, robust interference factors.