Research on Intelligent Car Navigation Technology Based on Multi-Sensor Fusion

This study is based on the ORB-SLAM3 framework, integrating LiDAR point cloud information, and proposes a deep learning-based feature extraction and matching method, as well as a loop closure detection method based on point cloud completion. The approach improves feature extraction speed, accuracy, and loop closure detection robustness, outperforming ORB-SLAM3 in complex outdoor scenarios.