基于惯性测量单元与视觉融合的三维空间定位技术研究

发布时间：2019-02-13 20:04

【摘要】：虚拟现实技术(简称:VR)是当前技术研究的热点之一。检测人体行为操作的惯性测量单元(简称IMU)是虚拟现实技术的核心感应元器件,把IMU惯性测量与机器视觉测量进行融合,从而实现抑制惯性测量的累计误差是本文的主要研究内容。在现实应用中,IMU惯性测量会引起的累计误差最终导致IMU惯性测量失真,这个问题一直困扰着很多国内外的VR设备厂家。在这种背景下,有关IMU惯性测量校正的相关概念应运而生,它的目标是借用外部信息作为参照来更新IMU惯性测量的状态,降低和消除IMU惯性测量的累积误差。目前IMU惯性测量校正算法已被广泛地应用于航天、导航仪器、虚拟现实等领域。本文首先讨论有关IMU惯性测量和视觉测量的相关理论技术基础,以及国内外在该领域的相关研究成果。接着分别对IMU惯性测量和视觉测量两种测量方法进行推导和实现。最后将两种测量的结果进行融合,得到校正后的IMU惯性测量结果。本文主要工作分为以下3点:(1)基于陀螺仪和加速度计、磁强计的IMU惯性组合算法:通过组合陀螺仪和加速度计、磁强计三种惯性元器件,得到混合磁强计的IMU惯性测量单元。这种IMU惯性测量单元的惯性坐标系与地球坐标系重合,可以监测元器件在移动时所产生的的偏移角度,最后利用罗德里格斯变换(Rodrigues)解出与视觉测量相匹配的数据。(2)基于视觉测距的方法获取现实中被测物体的方向向量:首先对双目相机进行视觉标定,得到相机的焦距和旋转矩阵等相关参数,接着用特征点检测方法检测目标特征点,最后利用双目测距定理解算目标深度信息,并求解出被测标志物的三维空间坐标。(3)基于卡尔曼算法将IMU测量向量和视觉测量向量的融合:将高频率的IMU测量作为滤波系统的状态向量,低频率的视觉测量作为更新向量,最终系统输出的偏移向量不仅频率高,并且具有自适应校准累积误差的功能。
[Abstract]:Virtual reality (: VR) technology is one of the hotspots of current technology research. The Inertial Measurement Unit (IMU), which detects human behavior, is the core inductive component of virtual reality technology, which combines IMU inertial measurement with machine vision measurement. It is the main research content of this paper to reduce the cumulative error of inertial measurement. In the practical application, the cumulative error caused by IMU inertial measurement will eventually lead to the distortion of IMU inertial measurement. This problem has been puzzling many domestic and foreign VR equipment manufacturers. In this context, the related concepts of IMU inertial measurement correction emerge as the times require. Its goal is to update the state of IMU inertial measurement by using external information as reference, and to reduce and eliminate the cumulative error of IMU inertial measurement. At present, IMU inertial measurement correction algorithm has been widely used in aerospace, navigation instruments, virtual reality and other fields. This paper first discusses the theoretical and technical basis of IMU inertial measurement and visual measurement, as well as the domestic and foreign research results in this field. Then the IMU inertial measurement and visual measurement are deduced and implemented. Finally, the results of the two measurements are fused to get the corrected IMU inertial measurement results. The main work of this paper is as follows: (1) the IMU inertial combination algorithm based on gyroscopes and accelerometers, magnetometers: through the combination of gyroscopes and accelerometers, magnetometer three kinds of inertial components, The IMU inertial measurement unit of the mixed magnetometer is obtained. The inertial coordinate system of the IMU inertial measurement unit coincides with the Earth coordinate system, which can monitor the deviation angle of the components when they move. Finally, the Rodrigues transform (Rodrigues) is used to solve the data matching with the visual measurement. (2) based on the visual ranging method, the direction vector of the measured object in reality is obtained: firstly, the binocular camera is visually calibrated. The relative parameters such as focal length and rotation matrix of the camera are obtained, then the target feature points are detected by the feature point detection method, and the target depth information is calculated by binocular ranging. The three-dimensional coordinates of the markers are solved. (3) based on the Kalman algorithm, the IMU measurement vector and the visual measurement vector are fused. The high-frequency IMU measurement is used as the state vector of the filtering system. The low-frequency vision measurement is used as the update vector, and the offset vector output by the system is not only high frequency, but also has the function of self-adaptive calibration cumulative error.
【学位授予单位】：中国地质大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.9;TP391.41

【相似文献】