This work presents a novel RGB-D dynamic Simultaneous Localisation and Mapping (SLAM) method that improves the precision, stability, and efficiency of localisation while relying on lightweight deep learning in a dynamic environment compared to the traditional static feature-based visual SLAM algorithm. Based on ORB-SLAM3, the GCNv2-tiny network instead of the ORB method, improves the reliability of feature extraction and matching and the accuracy of position estimation; then, the semantic segmentation thread employs the lightweight YOLOv5s object detection algorithm based on the GSConv network combined with a depth image to determine potentially dynamic regions of the image. Finally, to guarantee that the static feature points are used for position estimation, dynamic probability is employed to determine the true dynamic feature points based on the optical flow, semantic labels, and the state in last frame. We have performed experiments on the TUM datasets to verify the feasibility of the algorithm. Compared with the classical dynamic visual SLAM algorithm, the experimental results demonstrate that the absolute trajectory error is greatly reduced in dynamic environments, and that the computing efficiency is improved by 31.54% compared with the real-time dynamic visual SLAM algorithm with close accuracy, demonstrating the superiority of DLD-SLAM in accuracy, stability, and efficiency.

Visual–inertial SLAM algorithms empower robots to autonomously explore and navigate unknown scenes. However, most existing SLAM systems heavily rely on the assumption of static environments, making them ineffective when confronted with dynamic objects in the real world. To enhance the robustness and localization accuracy of SLAM systems in dynamic scenes, this paper introduces a visual–inertial SLAM framework that integrates semantic and geometric information, called D-VINS. This paper begins by presenting a method for dynamic object classification based on the current motion state of features, enabling the identification of temporary static features within the environment. Subsequently, a feature dynamic check module is devised, which utilizes inertial measurement unit (IMU) prior information and geometric constraints from adjacent frames to calculate dynamic factors. This module also validates the classification outcomes of the temporary static features. Finally, a dynamic adaptive bundle adjustment module is developed, utilizing the dynamic factors of the features to adjust their weights during the nonlinear optimization process. The proposed methodology is evaluated using both public datasets and a dataset created specifically for this study. The experimental results demonstrate that D-VINS stands as one of the most real-time, accurate, and robust systems for dynamic scenes, showcasing its effectiveness in challenging real-world scenes.

Perspective-n-point is a classical computer vision problem that uses three-dimensional points and image pixels to estimate camera pose. The visual robot often loses its position when the camera moves too fast or the environment changes. Perspective-n-point is used to relocate robot position, but the distribution of three-dimensional points in the world frame and different choices of coordinates affect the perspective-n-point performance and make perspective-n-point results less robust and inaccurate. In this study, we review the previous perspective-n-point algorithms and provide their disadvantages when facing three-dimensional points with large variances. According to the drawbacks of previous perspective-n-point methods, we propose a normalization method inspired by the homogeneous matrix calculation process to increase perspective-n-point algorithm accuracy and robustness. The experimental results demonstrate that the proposed perspective-n-point method is robust to different choices of coordinates and is thus better than other state-of-art perspective-n-point methods. Considering that the true camera pose is difficult to obtain, the former perspective-n-point solution validation experiment is mostly based on simulated image data. In this study, we design a new experiment based on total station and chessboard to verify the robustness and accuracy of the perspective-n-point algorithm.

Visual inertial SLAM algorithms enable robots to autonomously explore and navigate in unknown scenes. However, most of the current SLAM systems highly rely on static environment assumptions, which fails in the exsitence of motional objects in the real environment. To improve the robustness and localization accuracy of SLAM systems in dynamic scenes, this paper proposes a visual-inertial SLAM framework that fuses semantic and geometric information, called DA-VINS. First, this paper presents a dynamic object classification method based on feature’s current motion state, which obtains temporary static features in the environment. Secondly, a features dynamics check module based on IMU prior and adjacent frame’s geometry constraint is designed to calculate dynamic factors. It also verifies the classification results of temporary static features. Finally, a dynamic adaptive bundle adjustment module based on the features’ dynamic factors is designed to adjust the weights of features in nonlinear optimization. We evaluated our method in public and our dataset. The results show that D-VINS is one of the most real-time, accurate, and robust systems in dynamic scenes.

In geodetic surveying, input data from two coordinates are needed to compute rigid transformations. A common solution is a least-squares algorithm based on a Gauss–Markov model, called iterative closest point (ICP). However, the error in the ICP algorithm only exists in target coordinates, and the algorithm does not consider the source model’s error. A total least-squares (TLS) algorithm based on an errors-in-variables (EIV) model is proposed to solve this problem. Previous total least-squares ICP algorithms used a Euler angle parameterization method, which is easily affected by a gimbal lock problem. Lie algebra is more suitable than the Euler angle for interpolation during an iterative optimization process. In this paper, Lie algebra is used to parameterize the rotation matrix, and we re-derive the TLS algorithm based on a GHM (Gauss–Helmert model) using Lie algebra. We present two TLS-ICP models based on Lie algebra. Our method is more robust than previous TLS algorithms, and it suits all kinds of transformation matrices.

The global positioning system real-time kinematic (GPS-RTK) instrument is a commonly used instrument to obtain GPS coordinates in cadastral surveying, but its use is limited to unshielded environments. This study proposes a new method based on a visual simultaneous localization and mapping (SLAM) algorithm that uses a stereo-camera to replace traditional methods. The benefits of our method are a higher measurement efficiency, fewer required professionals, and fewer restrictions on the operating environment compared with the traditional RTK and total station method. First, we propose a robust non-planar camera calibration algorithm that uses three non-planar chessboards to obtain intrinsic parameters of a stereo-camera to improve the accuracy of the measurement. Second, we propose a detailed algorithm that uses a stereo-camera and known GPS coordinates to obtain unknown GPS coordinates. Using experiments, we demonstrated the effectiveness and stability of our method to provide accurate GPS coordinates.