Multi-sensor Deep Fusion Perception Scheme
Lists
Only you see this
Abstract
This study proposes a deep object detection method based on the fusion of multi-line LiDAR point cloud data and RGB images to improve detection accuracy and system robustness in autonomous driving environments. The method first performs feature extraction and deep fusion of multi-line LiDAR and image data, combining candidate object extraction and 3D object detection to achieve precise spatial localization. Experimental results show that, compared to methods using image data alone, this fusion approach significantly improves performance across various detection tasks on the KITTI dataset, especially in terms of detection accuracy and adaptability in complex environments. Additionally, comparative analysis with state-of-the-art algorithms, YOLOv5 and Faster R-CNN, further validates the advantages of this method.
BibTeX
@misc{hua2024multi,
author = {Yiding Hua and Hang Sun and Hui Rong and Botian Mei and Chun Li},
title = {{Multi-sensor Deep Fusion Perception Scheme}},
year = {2024},
doi = {10.1109/isstc63573.2024.10824136},
}Loading related content...