FLS-dataset

We collected a total of 381 forward-looking sonar images from two distinct locations. These locations were carefully selected to ensure diversity in the dataset, representing both structured and unstructured underwater environments. By incorporating these varied environments, the dataset encompasses a broad spectrum of obstacle types, enhancing its applicability to real-world underwater navigation scenarios. Structured obstacles, such as piers and vessels, have defined geometric shapes and predictable sonar reflections, whereas unstructured obstacles, like rocks and sloped surfaces, exhibit irregular contours and varying sonar signatures. The combination of these elements contributes to a more comprehensive dataset, enabling underwater robots to handle diverse navigational challenges effectively.本研究共收集了来自两个不同采集点位的381幅前视声呐（forward-looking sonar）图像。采集点位经过精心筛选，以保障数据集的多样性，涵盖结构化与非结构化两类水下环境。通过纳入这类多样化的水下环境，该数据集覆盖了丰富多样的障碍物类型，提升了其在实际水下导航场景中的应用价值。结构化障碍物（如码头、船舶）具有明确的几何形状与可预测的声呐反射特征；而非结构化障碍物（如礁石、斜坡表面）则呈现不规则轮廓与多变的声呐特征。两类障碍物的结合使得该数据集更具全面性，可助力水下机器人高效应对多样化的导航挑战。

搜集汇总

背景与挑战

背景概述

FLS-dataset是一个包含381张前视声纳图像的数据集，来自两个不同地点，覆盖结构化和非结构化的水下环境，适用于水下机器人导航和避障研究。数据集大小为332.13 KB，包含184个文件，发布于2025年。