Road Detection and Autonomous Driving
发布时间：
Overview
This project involved research and development on road detection and autonomous driving systems. The work focused on developing computer vision algorithms for accurate road detection, lane recognition, and obstacle detection to support autonomous driving applications. The project primarily utilized BEVFusion (Bird’s Eye View Fusion) combined with Transformer networks for multi-modal sensor fusion and perception.
Key Features
- BEVFusion Architecture: Utilizing BEVFusion framework for multi-modal sensor fusion (cameras, LiDAR) in bird’s eye view representation
- Transformer Networks: Applying Transformer architectures for feature extraction and fusion across different sensor modalities
- Road Detection: Advanced computer vision algorithms for accurate road surface detection and segmentation
- Lane Recognition: Real-time lane detection and tracking for autonomous navigation
- Obstacle Detection: Detection and classification of obstacles on the road using fused sensor data
- Multi-modal Fusion: Effective fusion of camera and LiDAR data in BEV space for robust perception
Technologies
- BEVFusion (Bird’s Eye View Fusion)
- Transformer Networks
- Computer Vision
- Deep Learning
- Autonomous Driving
- Multi-modal Sensor Fusion (Cameras, LiDAR)
- Python, PyTorch
- OpenCV
Impact
The developed road detection and perception systems contribute to improving the safety and reliability of autonomous driving systems, enabling better navigation and decision-making in complex traffic environments.