Improved Point-Line Feature Based Visual SLAM Method for Complex Environments.

Area of Science:

• Computer Vision
• Robotics
• Artificial Intelligence

Background:

• Traditional visual simultaneous localization and mapping (SLAM) systems often struggle in environments with weak textures or illumination changes.
• Reliance on point features alone limits the robustness of existing SLAM algorithms.

Purpose of the Study:

• To develop a more robust monocular visual SLAM system by integrating line segment features with traditional point features.
• To enhance camera trajectory estimation accuracy and reduce drift in challenging environments.

Main Methods:

• Proposed an ORB algorithm with a local adaptive threshold for feature extraction.
• Optimized line feature extraction and implemented a screening step to combine point and line feature matching.
• Introduced a weighted cost function considering feature richness and dispersion for improved SLAM performance.

Main Results:

• The integrated point-line feature method demonstrated competitive performance against state-of-the-art SLAM techniques.
• Evaluations on public datasets showed a significant reduction in trajectory drift and loss.
• The proposed system exhibited increased robustness in complex visual environments.

Conclusions:

• Combining point and line features significantly enhances the robustness of monocular visual SLAM.
• The developed system offers improved localization accuracy and reliability for robotic applications.
• This approach effectively addresses the limitations of traditional feature-based SLAM systems.