How is Visual SLAM and LiDAR sensor fusion implemented in MATLAB?

Visual Simultaneous Localization and Mapping (Visual SLAM) builds a map of an unknown environment while tracking a robot's location within it.

In MATLAB's Automated Driving Toolbox and Navigation Toolbox, the pipeline works as follows:

1. Feature Extraction: Extracts visual keypoints (ORB, SURF, or KAZE) across camera frames.
2. Pose Estimation: Matches points between frames to estimate relative camera movement.
3. Sensor Fusion: Fuses visual data with high-frequency Inertial Measurement Unit (IMU) readings using an Extended Kalman Filter (EKF) to prevent visual drift.
4. Map Optimization: Applies Pose Graph Optimization to refine trajectory accuracy and build a consistent 3D point cloud map.