BROWSE
Detail
Conference Paper
안개 환경에서의 LiDAR-Camera 융합 기반 3차원 객체 검출 성능 향상 방법
Cited - time in 
Share 
Share
- Issue Date
- 2025-11
- Citation
- 대한전자공학회 학술 대회 (추계) 2025, pp.536-539
- Publisher
- 대한전자공학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- This paper proposes a LiDAR Camera fusion method for 3D object detection in foggy environments based on the Seeing Through Fog (STF) dataset. The proposed approach enhances robustness in dense fog conditions by integrating multi-scale image features using a ResNet50- FPN backbone within a voxel-based fusion framework. Experiments conducted on the STF dataset demonstrate that the proposed method achieves approximately 7.9 points higher mean Average Precision (mAP), corresponding to about 17.7% relative improvement compared to the FocalsConv-based baseline, with notable gains for the Car and Pedestrian classes under dense fog conditions. These results validate the effectiveness of leveraging both LiDAR and multiscale image cues under adverse weather environments.
- KSP Keywords
- 3D object detection, Average Precision, Image Features, Multi-scale, Voxel-based, adverse weather