EPL-VINS: Efficient Point-Line Fusion Visual-Inertial SLAM With LK-RG Line Tracking Method and 2-DoF Line Optimization

Duofeng Zeng, Xiaotao Liu, Kangjin Huang, Jing Liu

Abstract: The performance of a visual SLAM system based on point features significantly diminishes in low-textured environments due to the challenges in extracting sufficient and reliable points. The fusion of line and point features improves SLAM system performance by providing additional visual constraints. To improve the efficiency and accuracy of the point-line-based SLAM system, this letter introduces EPL-VINS, an efficient point-line fusion visual-inertial SLAM system. We present the LK-RG line segment tracking method, which combines the Lucas-Kanade (LK) algorithm with the Region Growing (RG) algorithm from the Line Segment Detector (LSD). Moreover, we introduce a novel representation for spatial lines, based on which we construct line reprojection residuals and conduct a 2-degrees-of-freedom (2-DoF) optimization of spatial lines in the back-end. The proposed system is built upon VINS-Fusion, and supports the original three sensor suites: a monocular with an IMU, stereo cameras, and stereo cameras with an IMU. The experimental results show that the LK-RG method exhibits rapid processing and a high success rate in line segments matching. Furthermore, the entire system obtains better localization accuracy than the state-of-the-art algorithm.