Relative Orientation Estimation Research Articles

RGB-D 데이터를 이용한 비전 기반 로봇 간 상대 위치 및 자세 추정 기법 연구

In this paper, a vision-based robot-to-robot relative pose estimation method is proposed for markerless wheeled mobile robots using an RGB-D camera. The proposed method comprises three parts: (i) object detection using a machine vision technique with RGB data, (ii) relative position estimation averaging over the point cloud with outlier rejection, and (iii) relative orientation estimation using point cloud scan matching through iterative closest point (ICP) methods. The neighboring or target robots are detected using YOLOv3-tiny so that the position of the robot can be determined in pixel coordinates. To estimate the 3D position of the target robot in the camera coordinates of the ego-robot, the object detection results are used along with the depth information and the precalibrated pinhole camera model. To estimate the orientation of the target robot, a point cloud is generated using the depth image, downsampled, and then used for scan matching by using ICP methods. For experimental verification, the proposed relative pose estimation method was implemented as a robot operating system package and tested on a small-scale differential-driving wheeled mobile robot, TurtleBot3; the estimation results were compared with the ground truth in real-world environments.

Hand-Finger Pose Estimation Using Inertial Sensors, Magnetic Sensors and a Magnet

Hand-finger motion tracking based on inertial and magnetic measurement units (IMMUs) has become a well-accepted method in many applications such as rehabilitation and virtual reality. However, existing systems require an IMMU on each segment of the whole hand and known length of each segment, which is hard to achieve for users. Therefore, we proposed a small sensor configuration with three IMMUs and a magnet to estimate poses of interested fingertips relative to the hand. The relative orientations were estimated by integrating relative angular velocity and fused with relative orientation estimation during time intervals when the whole hand moved or rotated as one object. The relative positions were estimated by using a magnetic dipole model of the magnet and prior estimated relative orientations. The performance has been demonstrated in multiple experiments. For whole hand rotation and functional grasping or writing experiments, the estimated errors of index fingertip position and orientation relative to the hand were 8.0~9.8 mm and 5.7° ~11.2°. For index finger flexion and pinching, the estimated errors of index fingertip position and orientation relative to the hand were 16.2~25.2 mm and 20.6° ~ 21.7°.

ENHANCEMENT STRATEGIES FOR FRAME-TO-FRAME UAS STEREO VISUAL ODOMETRY

Autonomous navigation of indoor unmanned aircraft systems (UAS) requires accurate pose estimations usually obtained from indirect measurements. Navigation based on inertial measurement units (IMU) is known to be affected by high drift rates. The incorporation of cameras provides complementary information due to the different underlying measurement principle. The scale ambiguity problem for monocular cameras is avoided when a light-weight stereo camera setup is used. However, also frame-to-frame stereo visual odometry (VO) approaches are known to accumulate pose estimation errors over time. Several valuable real-time capable techniques for outlier detection and drift reduction in frame-to-frame VO, for example robust relative orientation estimation using random sample consensus (RANSAC) and bundle adjustment, are available. This study addresses the problem of choosing appropriate VO components. We propose a frame-to-frame stereo VO method based on carefully selected components and parameters. This method is evaluated regarding the impact and value of different outlier detection and drift-reduction strategies, for example keyframe selection and sparse bundle adjustment (SBA), using reference benchmark data as well as own real stereo data. The experimental results demonstrate that our VO method is able to estimate quite accurate trajectories. Feature bucketing and keyframe selection are simple but effective strategies which further improve the VO results. Furthermore, introducing the stereo baseline constraint in pose graph optimization (PGO) leads to significant improvements.

ENHANCEMENT STRATEGIES FOR FRAME-TO-FRAME UAS STEREO VISUAL ODOMETRY

Abstract. Autonomous navigation of indoor unmanned aircraft systems (UAS) requires accurate pose estimations usually obtained from indirect measurements. Navigation based on inertial measurement units (IMU) is known to be affected by high drift rates. The incorporation of cameras provides complementary information due to the different underlying measurement principle. The scale ambiguity problem for monocular cameras is avoided when a light-weight stereo camera setup is used. However, also frame-to-frame stereo visual odometry (VO) approaches are known to accumulate pose estimation errors over time. Several valuable real-time capable techniques for outlier detection and drift reduction in frame-to-frame VO, for example robust relative orientation estimation using random sample consensus (RANSAC) and bundle adjustment, are available. This study addresses the problem of choosing appropriate VO components. We propose a frame-to-frame stereo VO method based on carefully selected components and parameters. This method is evaluated regarding the impact and value of different outlier detection and drift-reduction strategies, for example keyframe selection and sparse bundle adjustment (SBA), using reference benchmark data as well as own real stereo data. The experimental results demonstrate that our VO method is able to estimate quite accurate trajectories. Feature bucketing and keyframe selection are simple but effective strategies which further improve the VO results. Furthermore, introducing the stereo baseline constraint in pose graph optimization (PGO) leads to significant improvements.

Relative Pose Estimation for Alignment of Long Cylindrical Components Based on Microscopic Vision

In this paper, an efficient relative pose estimation method based on multimicroscopic vision is presented for the alignment of long cylindrical components in six degree-of-freedom in the 3-D space. First, the relative pose estimation method measures the relative orientation errors between components in a coarse-to-fine manner. The coarse relative orientation estimation reflects the average orientation difference among multiedges and it promises the relative orientation errors can converge to a limited extent rapidly. The fine relative orientation estimation is to detect whether the components will interfere with each other in the insertion process and it provides higher orientation measurement accuracy under multilateral constraint circumstance. The method to determine the invisible feature lines of long cylindrical components of prism shape with odd edges for relative position error estimation is discussed in detail. Afterward, the relative position between components is measured by side-view cameras using the feature lines. In addition, the method to calibrate the optical axis of microscopic camera is also presented. The relative pose estimation method can reliably estimate the relative pose between long cylindrical components of prism shape with odd edges. Alignment experiments and results demonstrate the effectiveness of the proposed relative pose estimation method.

Linear Approach for Initial Recovery of the Exterior Orientation Parameters of Randomly Captured Images by Low-Cost Mobile Mapping Systems

Abstract. In this paper, we present a novel linear approach for the initial recovery of the exterior orientation parameters (EOPs) of images. Similar to the conventional Structure from Motion (SfM) algorithm, the proposed approach is based on a two-step strategy. In the first step, the relative orientation of all possible image stereo-pairs is estimated. In the second step, a local coordinate frame is established, and an incremental image augmentation process is implemented to reference all the remaining images into a local coordinate frame. Since our approach is based on a linear solution for both the relative orientation estimation as well as the initial recovery of the image EOPs, it does not require any initial approximation for the optimization process. Another advantage of our approach is that it does not require any prior knowledge regarding the sequence of the image collection procedure, therefore, it can handle a set of randomly collected images in the absence of GNSS/INS information. In order to illustrate the feasibility of our approach, several experimental tests are conducted on real datasets captured in either a block or linear trajectory configuration. The results demonstrate that the initial image EOPs obtained are accurate and can serve as a good initialization for an additional bundle adjustment process.