Automated River Substrate Mapping From Sonar Imagery With Machine Learning

• Published in: Journal of geophysical research. Machine learning and computation Vol. 1; no. 3
• Main Authors: Bodine, C. S., Buscombe, D., Hocking, T. D.
• Format: Journal Article
• Language: English
• Published: 01.09.2024
• Subjects: acoustic remote sensing; aquatic habitat; deep learning; semantic segmentation; side‐scan sonar; substrate mapping
• ISSN: 2993-5210; 2993-5210
• DOI: 10.1029/2024JH000135

Knowledge of the variation and distribution of substrates at large spatial extents in aquatic systems, particularly rivers, is severely lacking, impeding species conservation and ecosystem restoration efforts. Recreation‐grade side‐scan sonar (SSS) instruments have demonstrated their unparalleled value as a low‐cost scientific instrument capable of efficient and rapid imaging of the benthic environment. However, existing methods for generating georeferenced data sets from these instruments, especially substrate maps, remain a barrier of adoption for scientific inquiry due to the high degree of human intervention and required expertise. To address this shortcoming, we introduced PING‐Mapper, an open‐source and freely available Python‐based software for automatically generating geospatial benthic data sets from popular Humminbird® instruments reproducibly. The previously released Version 1.0 of the software provided automated workflows for exporting georeferenced sonar imagery. This study extends functionality with version 2.0 by incorporating semantic segmentation with deep neural networks to reproducibly map substrates at large spatial extents. We present a novel approach for generating label‐ready sonar data sets, creating label‐image training sets, and model training with transfer learning; all with open‐source tools. The six‐class substrate model achieves an overall accuracy of 78% and the best performing class achieves 91%. Grouping substrates into three classes further improves the overall accuracy (87%) and best performing class accuracy (94%). Additional workflows enable masking sonar shadows, calculating independent bed picks and correcting attenuation effects in the imagery to improve interpretability. This software provides an improved mechanism for automatically mapping substrate distribution from recreation‐grade SSS systems, thereby lowering the barrier for inclusion in wider aquatic research. Plain Language Summary Little is known about the location and amount of substrates such as sand, gravel, cobblestones, and bedrock on the bottom of rivers, lakes, and oceans. Without substrate maps, scientists and resource managers cannot fully understand, manage, and restore aquatic species and habitats. Recreation‐grade side‐scan sonar (SSS) devices, used by fishermen to locate fish or potential hazards, can be used to create grayscale images of the bottom of waterbodies. Substrate maps can be made through careful study of these images and manual drawing of boundaries. This skill takes time and expertise to develop, which is often not available to aquatic studies. We have developed new tools, utilizing machine learning (ML), to automatically map substrates from Humminbird® SSS data sets. The ML substrate models with six classes achieve 78% overall accuracy (OA) and 91% on the best class (BC). Grouping substrates into three classes improves OA (87%) and BC (94%). Open‐source tools are available in an updated version (v2.0) of PING‐Mapper, which also creates sonar image mosaics that can be viewed in geographic information system software. Additional tools remove sonar shadows, automatically measure depth, and enhance the sonar mosaics with corrections. This software improves substrate map production by providing aquatic researchers with an automated and consistent approach. Key Points Transfer learning with a repurposed SegFormer model for substrate segmentation and classification Substrate segmentation algorithms available in a free open‐source Python software package Fully automated routines ensure efficient and reproducible substrate map production