Dynamic obstacle detection method based on U-V disparity and residual optical flow for autonomous driving

Area of Science:

• Computer Vision
• Robotics
• Autonomous Systems

Background:

• Dynamic obstacle detection is crucial for autonomous driving, impacting obstacle avoidance and path planning.
• Existing methods face challenges in efficiency and accuracy for real-time applications.

Purpose of the Study:

• To propose a robust dynamic obstacle detection method for autonomous driving.
• To improve detection efficiency and accuracy by integrating drivable area detection with U-V disparity and optical flow.

Main Methods:

• Drivable area detection using U-V disparity images.
• Obstacle detection within the drivable area based on U-V disparity and geometric relationships.
• Motion likelihood estimation via ego-motion compensation and residual optical flow.

Main Results:

• Significantly narrowed search range for moving obstacles, enhancing efficiency.
• Demonstrated high detection precision and low missed detection rates on KITTI and campus datasets.
• Achieved reduced time consumption compared to existing methods.

Conclusions:

• The proposed method effectively detects dynamic obstacles in autonomous driving scenarios.
• Integrating drivable area detection with U-V disparity and optical flow offers a promising approach for enhanced safety and performance.
• The method shows potential for real-world deployment in autonomous vehicles.