In this paper we present a novel featurebased RGB-D camera pose optimization algorithm for real-time 3D reconstruction systems. During camera pose estimation, current methods in online systems suffer from fast-scanned...In this paper we present a novel featurebased RGB-D camera pose optimization algorithm for real-time 3D reconstruction systems. During camera pose estimation, current methods in online systems suffer from fast-scanned RGB-D data, or generate inaccurate relative transformations between consecutive frames. Our approach improves current methods by utilizing matched features across all frames and is robust for RGB-D data with large shifts in consecutive frames. We directly estimate camera pose for each frame by efficiently solving a quadratic minimization problem to maximize the consistency of3 D points in global space across frames corresponding to matched feature points. We have implemented our method within two state-of-the-art online 3D reconstruction platforms. Experimental results testify that our method is efficient and reliable in estimating camera poses for RGB-D data with large shifts.展开更多
In this paper, new solutions for the problem of pose estimation from correspondences between 3D model lines and 2D image lines are proposed. Traditional line-based pose estimation methods rely on the assumption that t...In this paper, new solutions for the problem of pose estimation from correspondences between 3D model lines and 2D image lines are proposed. Traditional line-based pose estimation methods rely on the assumption that the noises(perpendicular to the line) for the two endpoints are statistically independent. However, these two noises are in fact negatively correlated when the image line segment is fitted using the least-squares technique. Therefore, we design a new error function expressed by the average integral of the distance between line segments. Three least-squares techniques that optimize both the rotation and translation simultaneously are proposed in which the new error function is exploited. In addition, Lie group formalism is utilized to describe the pose parameters, and then, the optimization problem can be solved by means of a simple iterative least squares method. To enhance the robustness to outliers existing in the match data, an M-estimation method is developed to convert the pose optimization problem into an iterative reweighted least squares problem. The proposed methods are validated through experiments using both synthetic and real-world data. The experimental results show that the proposed methods yield a clearly higher precision than the traditional methods.展开更多
With the rapid development in the service,medical,logistics and other industries,and the increasing demand for unmanned mobile devices,mobile robots with the ability of independent mapping,localization and navigation ...With the rapid development in the service,medical,logistics and other industries,and the increasing demand for unmanned mobile devices,mobile robots with the ability of independent mapping,localization and navigation capabilities have become one of the research hotspots.An accurate map construction is a prerequisite for a mobile robot to achieve autonomous localization and navigation.However,the problems of blurring and missing the borders of obstacles and map boundaries are often faced in the Gmapping algorithm when constructing maps in complex indoor environments.In this pursuit,the present work proposes the development of an improved Gmapping algorithm based on the sparse pose adjustment(SPA)optimizations.The improved Gmapping algorithm is then applied to construct the map of a mobile robot based on single-line Lidar.Experiments show that the improved algorithm could build a more accurate and complete map,reduce the number of particles required for Gmapping,and lower the hardware requirements of the platform,thereby saving and minimizing the computing resources.展开更多
Robots are key to expanding the scope of space applications.The end-to-end training for robot vision-based detection and precision operations is challenging owing to constraints such as extreme environments and high c...Robots are key to expanding the scope of space applications.The end-to-end training for robot vision-based detection and precision operations is challenging owing to constraints such as extreme environments and high computational overhead.This study proposes a lightweight integrated framework for grasp detection and imitation learning,named GD-IL;it comprises a grasp detection algorithm based on manipulability and Gaussian mixture model(manipulability-GMM),and a grasp trajectory generation algorithm based on a two-stage robot imitation learning algorithm(TS-RIL).In the manipulability-GMM algorithm,we apply GMM clustering and ellipse regression to the object point cloud,propose two judgment criteria to generate multiple candidate grasp bounding boxes for the robot,and use manipulability as a metric for selecting the optimal grasp bounding box.The stages of the TS-RIL algorithm are grasp trajectory learning and robot pose optimization.In the first stage,the robot grasp trajectory is characterized using a second-order dynamic movement primitive model and Gaussian mixture regression(GMM).By adjusting the function form of the forcing term,the robot closely approximates the target-grasping trajectory.In the second stage,a robot pose optimization model is built based on the derived pose error formula and manipulability metric.This model allows the robot to adjust its configuration in real time while grasping,thereby effectively avoiding singularities.Finally,an algorithm verification platform is developed based on a Robot Operating System and a series of comparative experiments are conducted in real-world scenarios.The experimental results demonstrate that GD-IL significantly improves the effectiveness and robustness of grasp detection and trajectory imitation learning,outperforming existing state-of-the-art methods in execution efficiency,manipulability,and success rate.展开更多
In this work,we propose multiple rotation averaging using only the relative rotation angle,which is a straightforward camera pose optimization method.We use the axis-angle representation to parameterize the rotation a...In this work,we propose multiple rotation averaging using only the relative rotation angle,which is a straightforward camera pose optimization method.We use the axis-angle representation to parameterize the rotation and use only relative rotation angles to constrain absolute rotations instead of complete relative rotations.When used with an inertial measurement unit(IMU),our method can obviate the need to estimate and maintain extrinsic parameters between the camera and IMU.This advantage makes our method immune to extrinsic parameters and flexible.We performed extensive evaluations on both synthetic data and publicly available real datasets,which showed that our method was comparable with the state-of-the-art method and achieved a significant gain in accuracy for the visual measurement when applied to the case in which the camera and IMU are tightly fixed.展开更多
文摘In this paper we present a novel featurebased RGB-D camera pose optimization algorithm for real-time 3D reconstruction systems. During camera pose estimation, current methods in online systems suffer from fast-scanned RGB-D data, or generate inaccurate relative transformations between consecutive frames. Our approach improves current methods by utilizing matched features across all frames and is robust for RGB-D data with large shifts in consecutive frames. We directly estimate camera pose for each frame by efficiently solving a quadratic minimization problem to maximize the consistency of3 D points in global space across frames corresponding to matched feature points. We have implemented our method within two state-of-the-art online 3D reconstruction platforms. Experimental results testify that our method is efficient and reliable in estimating camera poses for RGB-D data with large shifts.
基金supported by the National Basic Research Program of China(“973”Project)(Grant No.2013CB733100)National Natural Science Foundation of China(Grant No.11332012)
文摘In this paper, new solutions for the problem of pose estimation from correspondences between 3D model lines and 2D image lines are proposed. Traditional line-based pose estimation methods rely on the assumption that the noises(perpendicular to the line) for the two endpoints are statistically independent. However, these two noises are in fact negatively correlated when the image line segment is fitted using the least-squares technique. Therefore, we design a new error function expressed by the average integral of the distance between line segments. Three least-squares techniques that optimize both the rotation and translation simultaneously are proposed in which the new error function is exploited. In addition, Lie group formalism is utilized to describe the pose parameters, and then, the optimization problem can be solved by means of a simple iterative least squares method. To enhance the robustness to outliers existing in the match data, an M-estimation method is developed to convert the pose optimization problem into an iterative reweighted least squares problem. The proposed methods are validated through experiments using both synthetic and real-world data. The experimental results show that the proposed methods yield a clearly higher precision than the traditional methods.
基金National Key Research and Development of China(No.2019YFB1600700)Sichuan Science and Technology Planning Project(No.2021YFSY0003)。
文摘With the rapid development in the service,medical,logistics and other industries,and the increasing demand for unmanned mobile devices,mobile robots with the ability of independent mapping,localization and navigation capabilities have become one of the research hotspots.An accurate map construction is a prerequisite for a mobile robot to achieve autonomous localization and navigation.However,the problems of blurring and missing the borders of obstacles and map boundaries are often faced in the Gmapping algorithm when constructing maps in complex indoor environments.In this pursuit,the present work proposes the development of an improved Gmapping algorithm based on the sparse pose adjustment(SPA)optimizations.The improved Gmapping algorithm is then applied to construct the map of a mobile robot based on single-line Lidar.Experiments show that the improved algorithm could build a more accurate and complete map,reduce the number of particles required for Gmapping,and lower the hardware requirements of the platform,thereby saving and minimizing the computing resources.
基金Supported by National Natural Science Foundation of China(Grant No.52475280)Shaanxi Provincial Natural Science Basic Research Program(Grant No.2025SYSSYSZD-105).
文摘Robots are key to expanding the scope of space applications.The end-to-end training for robot vision-based detection and precision operations is challenging owing to constraints such as extreme environments and high computational overhead.This study proposes a lightweight integrated framework for grasp detection and imitation learning,named GD-IL;it comprises a grasp detection algorithm based on manipulability and Gaussian mixture model(manipulability-GMM),and a grasp trajectory generation algorithm based on a two-stage robot imitation learning algorithm(TS-RIL).In the manipulability-GMM algorithm,we apply GMM clustering and ellipse regression to the object point cloud,propose two judgment criteria to generate multiple candidate grasp bounding boxes for the robot,and use manipulability as a metric for selecting the optimal grasp bounding box.The stages of the TS-RIL algorithm are grasp trajectory learning and robot pose optimization.In the first stage,the robot grasp trajectory is characterized using a second-order dynamic movement primitive model and Gaussian mixture regression(GMM).By adjusting the function form of the forcing term,the robot closely approximates the target-grasping trajectory.In the second stage,a robot pose optimization model is built based on the derived pose error formula and manipulability metric.This model allows the robot to adjust its configuration in real time while grasping,thereby effectively avoiding singularities.Finally,an algorithm verification platform is developed based on a Robot Operating System and a series of comparative experiments are conducted in real-world scenarios.The experimental results demonstrate that GD-IL significantly improves the effectiveness and robustness of grasp detection and trajectory imitation learning,outperforming existing state-of-the-art methods in execution efficiency,manipulability,and success rate.
基金Hunan Provincial Natural Science Foundation for Excellent Young Scholars(Grant No.2023JJ20045)in part by the Science Foundation(Grant Nos.KY0505072204 and GJSD22006)。
文摘In this work,we propose multiple rotation averaging using only the relative rotation angle,which is a straightforward camera pose optimization method.We use the axis-angle representation to parameterize the rotation and use only relative rotation angles to constrain absolute rotations instead of complete relative rotations.When used with an inertial measurement unit(IMU),our method can obviate the need to estimate and maintain extrinsic parameters between the camera and IMU.This advantage makes our method immune to extrinsic parameters and flexible.We performed extensive evaluations on both synthetic data and publicly available real datasets,which showed that our method was comparable with the state-of-the-art method and achieved a significant gain in accuracy for the visual measurement when applied to the case in which the camera and IMU are tightly fixed.
