Related Experiment Video
Updated: May 7, 2026
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
8.9K
Fine-grained vehicle recognition under low light conditions using EfficientNet and image enhancement on LiDAR point
Guanqiang Ruan1, Tao Hu1, Chenglin Ding2
11Automotive Structure and Energy Storage Engineering Center, School of Mechanical Engineering, Shanghai Dianji University, Shanghai, 201306, China.
Scientific Reports
|February 7, 2025
Summary
This study presents a new method for fine-grained vehicle model recognition using LiDAR, achieving 98.88% accuracy even in low-light conditions. The approach enhances LiDAR data for precise autonomous driving perception.
Area of Science:
- Computer Vision
- Robotics
- Sensor Fusion
Background:
- Environmental perception is vital for autonomous driving, with cameras and LiDAR as common sensors.
- Camera performance degrades in adverse lighting, while LiDAR offers robustness but limited classification detail.
- Existing LiDAR methods struggle with fine-grained vehicle model recognition, especially in low-light scenarios.
Purpose of the Study:
- To introduce a novel method for fine-grained vehicle model recognition using LiDAR data.
- To address the limitations of current sensors in low-light autonomous driving perception.
- To improve the accuracy and detail of vehicle classification from LiDAR point clouds.
Main Methods:
- Collected LiDAR data for various vehicle models.
- Applied projection transformation to the LiDAR data.
- Enhanced data using contrast limited adaptive histogram equalization and Gamma correction.
- Implemented vehicle model recognition utilizing the EfficientNet deep learning architecture.
Main Results:
- Achieved 98.88% accuracy in fine-grained vehicle model recognition.
- Obtained an F1-score of 98.86% for the proposed method.
- Demonstrated robust performance in low-light conditions.
Conclusions:
- The proposed LiDAR-based method significantly improves fine-grained vehicle model recognition.
- The data enhancement techniques are effective for improving LiDAR data quality in low-light.
- This approach offers a promising solution for enhanced perception in autonomous driving systems.
