AirSLAM: An Efficient and Illumination-Robust Point-Line Visual SLAM System

Kuan Xu,Yuefan Hao,Shenghai Yuan,Chen Wang,Lihua Xie

2024-09-18

Abstract:In this paper, we present an efficient visual SLAM system designed to tackle both short-term and long-term illumination challenges. Our system adopts a hybrid approach that combines deep learning techniques for feature detection and matching with traditional backend optimization methods. Specifically, we propose a unified convolutional neural network (CNN) that simultaneously extracts keypoints and structural lines. These features are then associated, matched, triangulated, and optimized in a coupled manner. Additionally, we introduce a lightweight relocalization pipeline that reuses the built map, where keypoints, lines, and a structure graph are used to match the query frame with the map. To enhance the applicability of the proposed system to real-world robots, we deploy and accelerate the feature detection and matching networks using C++ and NVIDIA TensorRT. Extensive experiments conducted on various datasets demonstrate that our system outperforms other state-of-the-art visual SLAM systems in illumination-challenging environments. Efficiency evaluations show that our system can run at a rate of 73Hz on a PC and 40Hz on an embedded platform.

Robotics

What problem does this paper attempt to address?

This paper attempts to address the challenges posed by short-term and long-term illumination changes in Visual Simultaneous Localization and Mapping (vSLAM) systems.

Specifically, the paper focuses on the following two main issues:

1. **Short-term illumination challenges**: In a short period, due to drastic changes in lighting conditions or low-light environments, feature detection and tracking often fail, severely affecting the quality of trajectory estimation. This mainly impacts the pose estimation between two temporally adjacent frames.

2. **Long-term illumination challenges**: When the visual map is reused for relocalization, changes in lighting conditions significantly reduce the success rate of relocalization. This mainly affects the matching between the query frame and the existing map.

To address these challenges, the paper proposes a novel point-line combined vSLAM system—AirSLAM. This system improves robustness and efficiency in environments with varying illumination through the following innovations:

- **Unified Convolutional Neural Network (CNN)**: Simultaneously extracts key points and structural line features, and jointly utilizes these features during detection and matching.
- **Multi-stage relocalization method**: Utilizes point and line features, combining appearance and geometric information to achieve fast and illumination-robust relocalization.
- **Efficient feature detection and matching**: Accelerates the feature detection and matching network through C++ and NVIDIA TensorRT, enabling efficient operation on resource-constrained platforms.

With these innovations, AirSLAM not only performs excellently in environments with varying illumination but also achieves a high level of efficiency, capable of running at 73Hz on a PC and 40Hz on an embedded platform.