DIG-SLAM: an accurate RGB-D SLAM based on instance segmentation and geometric clustering for dynamic indoor scenes

Abstract

Simultaneous localization and mapping (SLAM) has emerged as a critical technology enabling robots to navigate in unknown environments, drawing extensive attention within the robotics research community. However, traditional visual SLAM ignores the presence of dynamic objects in indoor scenes, and dynamic point features of dynamic objects can lead to incorrect data correlation, making the traditional visual SLAM is difficult to accurately estimate the camera’s pose when the objects in the scenes are moving. Using only point features cannot fully extract geometric information in dynamic indoor scenes, reducing the system’s robustness. To solve this problem, we develop a RGB-D SLAM system called DIG-SLAM. Firstly, the objects’ contour regions are extracted using the YOLOv7 instance segmentation method, serving as a prerequisite for determining dynamic objects and constructing a semantic information map. Meanwhile, the line features are extracted using the line segment detector (LSD) algorithm, and the redundant line features are optimized via K-means clustering. Secondly, moving consistency checks combined with instance partitioning determine dynamic regions, and the point and line features of the dynamic regions are removed. Finally, the combination of static line features and point features optimizes the camera pose. Meanwhile, a static semantic octree map is created to provide richer and higher-level scene understanding and perception capabilities for robots or autonomous systems. The experimental results on the Technische Universität München (TUM) dataset show that the average absolute trajectory error of the developed DIG-SLAM is reduced by 28.68% compared with the dynamic semantic SLAM (DS-SLAM). Compared with other dynamic SLAM methods, the proposed system shows better camera pose estimation accuracy and system’s robustness in dynamic indoor environments and better map building in real indoor scenes.

YG-SLAM: Enhancing Visual SLAM in Dynamic Environments With YOLOv8 and Geometric Constraints

“…Ran et al [48] proposed a novel RGB-D-inertial dynamic SLAM method that enables accurate localization even when a significant portion of the camera's view is obstructed by multiple dynamic objects for an extended period.SVD-SLAM [49] has incorporated an improved YOLACT++ lightweight instance segmentation network and calculates the motion probability of each potential moving object based on the camera pose and polar constraints.DIG-SLAM [50] utilizes YOLOv7 instance segmentation to extract object contours, while optimizing line features through line segment detection and K-means clustering. It employs motion consistency checks to identify dynamic areas and removes points and line features within them.DYS-SLAM [51] Current literature mainly leans towards combining geometric constraints and deep learning for Visual SLAM in dynamic settings.…”

Hierarchical fusion based high precision SLAM for solid-state lidar

“…Based on case segmentation, Li et al [15] introduces a confidence-based matching algorithm to correlate dynamic objects and differentiate moving objects from static objects. To eliminate the interference of dynamic objects in the room, Liang et al [16] determined the dynamic area by moving one-time inspection combined with case segmentation, thus removing points and features within the dynamic area. Although the above methods can filter dynamic objects, they are not real-time.…”

Precise landmark-map for bundle adjustment LiDAR odometry

“…For example, generalized-ICP (GICP) [6] attempts to match local planar patches from both scans and integrates pointto-plane distances to ICP. Alternatively, feature-based matching methods [4,5] minimize computational efforts by only matching feature points extracted from raw point clouds.…”

“…The conventional LiDAR SLAM framework [3][4][5][6][7][8][9][10][11][12] operates by sequentially registering LiDAR scans, thereby estimating the position of the present frame for each iteration. However, it is worth noting that the information from the current frame can also enhance the estimation of historical frames, consequently improving the accuracy of the current frame itself.…”