Advances in Sonar Image Segmentation: A Comprehensive Review

Research in sonar image segmentation, a critical technology for ocean exploration, has seen significant advancements in recent years. A team of researchers from the Zhejiang University of Water Resources and Electric Power has conducted a comprehensive review of the current state of sonar image segmentation, highlighting the challenges and opportunities in this field. The study, which has been peer-reviewed, provides a detailed analysis of traditional and deep-learning-based methods, as well as a classification and comparison of various segmentation algorithms.

Key Takeaways:

• Sonar image segmentation is a crucial technology in ocean exploration, used to separate foreground, shadow, and background from sonar images.
• Traditional image processing algorithms are commonly used in sonar image segmentation, but deep-learning based methods have shown great potential in recent years.
• The study analyzed the research progress in sonar image segmentation from the last century to the present, using cluster analysis and a detailed classification of traditional and deep-learning-based methods.
• Despite the increasing popularity of deep-learning-based methods, traditional analysis-based methods still have a significant impact in the field of sonar segmentation.
• The study proposed several research directions to be studied in the future, including the development of multi-modal datasets and the investigation of new segmentation methods.

Statistics:

• The study reviewed over 100 research papers on sonar image segmentation published in the last century.
• The number of research papers on sonar image segmentation using deep-learning methods has increased significantly in recent years, from 5 papers in 2015 to 25 papers in 2022.
• The study found that sonar images have unique characteristics, such as widespread noise sources and uneven intensity distribution, which require special segmentation algorithms.
• The study proposed a multi-modal dataset based on sonar and camera collected in an experimental pool, which could be used to evaluate and compare different segmentation methods.

Sources:

• Zhejiang University of Water Resources and Electric Power, College of Electrical Engineering, Hangzhou 310018, People's Republic of China
• National Natural Science Foundation of China (NSFC)
• Open Fund Project Hanjiang National Laboratory
• Key Research and Development Program Zhejiang Province
• Key Laboratory of Multimodal Perception and Intelligent Systems of Zhejiang Province
• Elsevier, Radarweg 29, 1043 Nx Amsterdam, Netherlands (www.elsevier.com; www.journals.elsevier.com/neurocomputing/)