A Multi-Feature Fusion Slam System Attaching Semantic Invariant to Points and Lines

Gang Li,Yawen Zeng,Huilan Huang,Shaojian Song,Bin Liu,Xiang Liao

DOI: https://doi.org/10.3390/s21041196

IF: 3.9

2021-02-08

Sensors

Abstract:The traditional simultaneous localization and mapping (SLAM) system uses static points of the environment as features for real-time localization and mapping. When there are few available point features, the system is difficult to implement. A feasible solution is to introduce line features. In complex scenarios containing rich line segments, the description of line segments is not strongly differentiated, which can lead to incorrect association of line segment data, thus introducing errors into the system and aggravating the cumulative error of the system. To address this problem, a point-line stereo visual SLAM system incorporating semantic invariants is proposed in this paper. This system improves the accuracy of line feature matching by fusing line features with image semantic invariant information. When defining the error function, the semantic invariant is fused with the reprojection error function, and the semantic constraint is applied to reduce the cumulative error of the poses in the long-term tracking process. Experiments on the Office sequence of the TartanAir dataset and the KITTI dataset show that this system improves the matching accuracy of line features and suppresses the cumulative error of the SLAM system to some extent, and the mean relative pose error (RPE) is 1.38 and 0.0593 m, respectively.

engineering, electrical & electronic,chemistry, analytical,instruments & instrumentation

What problem does this paper attempt to address?

The paper primarily aims to address several key issues in Visual Simultaneous Localization and Mapping (Visual SLAM) systems:

1. **Feature extraction in low-texture environments**: In low-texture or motion-blurred environments, traditional point features may be difficult to extract effectively, which can severely impact localization accuracy and even lead to system failure.

2. **Accuracy of line feature matching**: While line features can somewhat compensate for the shortcomings of point features in low-texture environments and provide more complete environmental structure information, the description of line features in complex scenes is not precise enough, leading to erroneous matches and introducing errors that exacerbate the problem of accumulated errors in the system.

3. **Suppression of accumulated errors**: Even though short-term trajectory drift can be reduced through local optimization methods, errors will still accumulate once constraint conditions fail; using loop closure detection to establish long-term constraints is overly dependent on the accuracy of loop closure detection.

To address the above issues, the paper proposes a stereo visual SLAM system that combines point and line feature fusion with semantic invariance. The specific contributions include:

- Proposing an improved line feature matching method that utilizes the results of semantic segmentation to enhance the accuracy of line feature data association.

- Defining a semantic reprojection error function for line features and applying it in the pose optimization process to achieve mid-term tracking of line features, thereby reducing trajectory drift and improving the robustness of the system.

In summary, the paper aims to improve the accuracy of line feature matching in point-line feature fusion SLAM systems by introducing semantic invariance constraints, thereby effectively suppressing accumulated errors and enhancing the overall stability and reliability of the system.