TSQ-2024: a categorized dataset of 2D LiDAR images of moving dump trucks in various environment conditions

摘要

LiDARs are powerful and common sensors used for complex 3D environment analysis tasks. They are widely utilized as data providers in industrial measurements, robotics and unmanned technologies. Due to physical properties of such sensors they suffer from environmental heterogenity resulting in distortions of obtained measurements because of natural clutter such as cloud dust, snow and fog. Nevertheless the problem of robust clutter filtering algorithms is of significant importance, we highlight the lack of LiDAR measurements datasets containing natural clutter. The current study presents a novel publicly available and expertly annotated 2D LiDAR measurements dataset TSQ-2024 that contains 120 measurements of passing trucks under natural conditions including sand dust clutter data. The proposed dataset was utilized in the task of truck body segmentation. The experiments section provides the performance evaluation of the proposed truck body segmentation algorithm on the TSQ-2024 dataset which is also published along with the data.