Evaluation of multi-channel LiDAR Image modalities for LiDAR-only object detection
Loading...
URL
Journal Title
Journal ISSN
Volume Title
School of Electrical Engineering |
Master's thesis
Unless otherwise stated, all rights belong to the author. You may download, display and print this publication for Your own personal use. Commercial use is prohibited.
Authors
Date
Department
Major/Subject
Mcode
Language
en
Pages
53
Series
Abstract
This thesis investigates LiDAR-only object detection using multiple image modalities derived from a single multi-channel LiDAR sensor. Range, reflectivity, signal strength, and near-infrared (NearIR) outputs from an Ouster OS1-64 LiDAR are represented as two-dimensional images and evaluated within a unified YOLOv8-based detection framework under identical training and evaluation conditions. Single-modality experiments show that NearIR and reflectivity provide the strongest detection performance, while range images alone perform poorly due to their limited visual structure. Multi-modality experiments using early and late fusion reveal that certain modality combinations improve detection accuracy, whereas others yield diminishing returns due to less complementary interactions between modalities. The results further highlight practical limitations of LiDAR-only detection in cluttered environments, at long ranges, and on low-reflectivity surfaces. In summary, this study provides a comparative assessment of LiDAR image modalities and establishes an empirical foundation for future research on intra-LiDAR fusion strategies within LiDAR-only perception systems.Description
Supervisor
Kucner, TomaszThesis advisor
Tikkanen, TommiKutila, Matti