A Speedy Point Cloud Registration Method Based on Region Feature Extraction in Intelligent Driving Scene

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision
• Geospatial Data Processing

Background:

• Lidar point cloud data in intelligent vehicle driving presents challenges due to large scale, complex distribution, high noise, and sparsity.
• Accurate point cloud registration is crucial for environmental perception in autonomous vehicles.

Purpose of the Study:

• To propose an efficient and accurate point cloud registration algorithm for large-scale outdoor road scenes.
• To address the limitations of existing methods in handling noisy and sparse lidar data.

Main Methods:

• Selection of continuously distributed key area laser point clouds for registration.
• Extraction of feature descriptions and introduction of local geometric features for point cloud registration.
• Implementation of constrained rough and fine registration using key point clouds and their features.

Main Results:

• Achieved an average registration time of 0.5831 seconds and an average accuracy of 0.06996 meters in extensive experimental validation.
• Demonstrated significant performance improvements compared to other existing point cloud registration algorithms.
• Validated through real-vehicle experiments, confirming versatility, reliability, and efficiency.

Conclusions:

• The proposed algorithm offers an efficient and effective solution for point cloud registration in large-scale outdoor environments.
• This research has the potential to significantly enhance the environment perception capabilities of autonomous vehicles.
• The method's robustness and efficiency make it suitable for real-world applications in intelligent transportation systems.