Synthetic Aperture Sonar Seabed Environment Dataset (SASSED)

Approved for Public Release; distribution is unlimited.

Dataset title: Synthetic Aperture Sonar Seabed Environment Dataset (SASSED)

Date: June 2018

Description: This dataset contains 129 complex-valued, single (high frequency) channel, 1001x1001 pixel, synthetic aperture sonar snippets of various seafloor texture types. Each snippet contains one or more seabed environments, e.g., hardpack sand, mud, sea grass, rock, and sand ripple.

For each snippet there is a corresponding hand-segmented and -labeled "mask" image. The labels should not be interpreted as the ground truth for specific seafloor types. The labels were not verified by visual inspection of the actual seafloor environments or by any other method. Instead, interpret the labels as groupings of similar seafloor textures. Example code for preprocessing the data is included.

The data is stored in hdf5 format. The SAS data is stored under the hdf5 dataset 'snippets', and the hand-segmented labels are stored under 'labels'. For information on how to read hdf5 data, please visit one of the following websites: (general) https://support.hdfgroup.org/HDF5/ (python) https://www.h5py.org

Acknowledgements: Thanks go to J. Tory Cobb for curating this dataset. Please credit NSWC Panama City Division in any publication using this data.

Past Usage: Cobb, J. T., & Zare, A. (2014). Boundary detection and superpixel formation in synthetic aperture sonar imagery. Proceedings of the Institute of Acoustics, 36(Pt 1).

Approved for Public Release; distribution is unlimited.

本项目已获批公开发布，无分发限制。

数据集标题：合成孔径声呐海底环境数据集（Synthetic Aperture Sonar Seabed Environment Dataset，SASSED）

发布日期：2018年6月

数据集说明：本数据集包含129段复数值、单（高频）通道、分辨率为1001×1001像素的合成孔径声呐片段，涵盖多种海底纹理类型。每段片段包含一种或多种海底环境，例如压实沙、淤泥、海草、岩石以及沙纹。

每段声呐片段均对应一张手动分割并标注的“掩码”（mask）图像。需注意，此处标注不应被视为特定海底类型的真实标签：这些标注未通过实际海底环境的目视检查或其他任何方法进行验证，仅可将其视为相似海底纹理的分组。数据集附带数据预处理示例代码。

本数据集以HDF5格式存储：合成孔径声呐数据存储于HDF5数据集'snippets'下，手动分割的标注数据存储于'labels'下。如需了解读取HDF5数据的方法，请访问以下任一网站：（通用指南）https://support.hdfgroup.org/HDF5/ （Python接口）https://www.h5py.org

致谢：感谢J. Tory Cobb整理本数据集。若使用本数据进行任何出版物创作，请注明NSWC巴拿马城分部（NSWC Panama City Division）。

过往引用案例：Cobb, J. T. 与 Zare, A. (2014). 《合成孔径声呐图像中的边界检测与超像素生成》. 声学学会会报, 36(第1部分).

本项目已获批公开发布，无分发限制。

搜集汇总

背景与挑战

背景概述

SASSED数据集是一个包含129个合成孔径声纳图像片段的数据集，涵盖多种海底环境类型，每个片段配有手工标记的掩码图像。数据集以hdf5格式存储，适用于图像分割和声纳信号处理研究，但标记仅为相似纹理的分组而非真实标签。