Loop Closing Detection in RGB-D SLAM Combining Appearance and Geometric Constraints

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for robots to build maps and track their position.
• RGB-D cameras provide both color and depth information, enhancing SLAM capabilities.
• Loop closing detection is vital for correcting accumulated errors in SLAM.

Purpose of the Study:

• To develop a fast and accurate loop closing detection algorithm for RGB-D SLAM.
• To improve the robustness of SLAM systems by effectively rejecting false loop closure hypotheses.

Main Methods:

• A multi-feature point matching algorithm integrating local geometric constraints.
• Encoding visual features using BRAND (binary robust appearance and normals descriptor) for RGB-D images.
• Utilizing Locality-Sensitive Hashing (LSH) for efficient feature descriptor storage and hierarchical clustering trees for searching.

Main Results:

• Demonstrated efficiency in real-time RGB-D SLAM with loop closing detection using handheld Kinect camera data.
• Comparative experiments showed superior performance against other algorithms in RTAB-Map on benchmark datasets.
• The proposed algorithm effectively rejects false loop closure hypotheses.

Conclusions:

• The proposed algorithm significantly enhances loop closing detection in RGB-D SLAM.
• Combining appearance and geometric information via BRAND descriptors improves feature matching accuracy.
• The integration of LSH and hierarchical clustering enables efficient and robust loop closure detection.