Point-Plane SLAM Using Supposed Planes for Indoor Environments.

Area of Science:

• Robotics
• Computer Vision
• Artificial Intelligence

Background:

• Simultaneous Localization and Mapping (SLAM) is crucial for autonomous systems.
• Indoor environments often feature dominant planes, offering robust features for SLAM.
• Existing SLAM methods can be sensitive to noise and may not fully exploit geometric properties of indoor spaces.

Purpose of the Study:

• To propose a novel point-plane SLAM system tailored for indoor environments using RGB-D cameras.
• To improve pose estimation accuracy and robustness by leveraging plane features and their geometric constraints.
• To enhance the performance of SLAM systems by incorporating novel constraints derived from plane edges and structural relationships.

Main Methods:

• Extracting feature points from RGB images and planes from depth images.
• Utilizing plane contours for global map correspondences and exploiting plane edge constraints.
• Generating perpendicular planes from edge lines to increase observations and constraints.
• Incorporating structural constraints (parallel/perpendicular) based on indoor environment geometry.
• Constructing a factor graph and minimizing cost functions for pose and map estimation.

Main Results:

• The proposed point-plane SLAM system demonstrates robust and accurate camera pose estimation.
• The novel use of plane edge constraints and structural information effectively reduces estimation errors.
• Performance evaluation on public RGB-D benchmarks shows superiority over existing state-of-the-art SLAM systems.

Conclusions:

• The developed point-plane SLAM system offers a significant advancement for indoor localization and mapping.
• Exploiting geometric properties of planes and their edges provides a more robust SLAM solution.
• The system's accuracy and robustness make it suitable for various real-world indoor applications.