本文为加拿大滑铁卢大学（作者：Neda Parnian）的博士论文，共137页。
这项研究涉及开发一种智能传感系统，用于跟踪手持移动设备，使其达到毫米级精度，适用于长时间缓慢或几乎静止的场景。由于不同应用中的不同操作者可能会使用该系统，因此应提供跟踪目标的准确位置、方向和速度，而不依赖于对其操作和环境的了解，并且纯粹是基于目标所经历的运动。
本文提出将低成本的局部定位系统（LPS）和低成本的捷联惯性导航系统（SDINS）与改进的EKF相结合，在一定精度范围内确定手持工具的三维位置和方位。通过一个混合的LPS/SDINS组合，补充了两个不同导航系统的最佳特性，提供了一个独特的解决方案来更精确地跟踪和定位运动物体。SDINS提供了对运动所有组成部分的连续估计，但由于惯性传感器漂移和固有噪声，SDINS会随着时间的推移而损失精度。LPS的优点是可以独立于操作时间获得绝对位置和速度，但其鲁棒性不强，计算量大，测量速度慢。
本研究包括三个主要部分：开发一个可靠且具成本效益的多摄像机视觉系统、开发一个用于手持工具的SDINS以及开发一个用于传感器融合的Kalman滤波器。开发多摄像机视觉系统包括在工作空间周围安装摄像机、标定摄像机、拍摄图像、应用图像处理算法和从每个摄像机中提取每一帧的特征，以及从二维图像中估计三维位置。在本研究中，我们提出建立多摄影机视觉系统的具体配置，以尽可能减少视线损失。摄像机的数量、摄像机相互之间的位置以及摄像机相对于世界坐标系中心的位置和方向是这种配置的关键特征。提出了一种多摄像机视觉系统，该系统采用四个固定在导航架上的CCD摄像机，并将摄像机镜头放置在半圆上。所有摄像机都通过帧捕获器连接到PC机，该帧捕获器包括四个并行视频通道，能够同时从四个摄像机捕获图像。
基于以上设置，一个宽的圆形视场具有较少的视线损失。然而，校准比单目或立体视觉系统更困难。多摄像机视觉系统的标定包括精确的摄像机建模、每台摄像机的单摄像机标定、相邻两台摄像机的立体摄像机标定、定义唯一的世界坐标系、寻找从每个摄像机帧到世界坐标系的变换。除了校准程序外，还需要对所有四个摄像头捕获的图像进行数字图像处理，以便定位工具。
在本研究中，数字影像处理包括影像增强、边缘检测、边界检测和形态学运算。在检测到每台摄像机拍摄的每幅图像中的工具尖端后，应用三角剖分程序和优化算法，找到其相对于已知导航帧的三维位置。在单自由度惯性导航系统中，惯性传感器被刚性地直接安装在跟踪物体上，惯性测量值被计算转换到已知的导航框架下。通常，三个陀螺仪和三个加速度计，或三轴陀螺仪和三轴加速度计用于实现SDINS。惯性传感器通常集成在惯性测量单元（IMU）中。IMU通常会受到偏置漂移、非线性和温度变化导致的比例因子误差以及微小制造缺陷导致的失调的影响。由于所有这些误差都会导致SDINS在位置和方向上的漂移，因此需要一个精确的校准程序来补偿这些误差。SDINS的精度不仅取决于标定参数的精度，而且还取决于常见的运动相关误差。常见的运动相关误差是指由振动、圆锥运动、划动和旋转运动引起的误差。由于惯性传感器提供了物体在运动过程中所经历的全方位航向变化、转向率和所施加的力，因此可以建立精确的三维运动学方程来补偿常见的运动相关误差。因此，寻找工具尖端运动和取向的完整信息需要显著的计算复杂性和与特定力的分辨率、姿态计算、重力补偿以及常见运动相关误差校正有关的挑战。
卡尔曼滤波技术是改善输出估计和减小传感器漂移影响的有效方法。为了减小位置估计的误差，本文提出了改进的EKF算法，将提出的多摄像机视觉系统数据与改进的EKF相结合，有助于SDINS处理漂移问题。这种配置保证了仪器的实时位置和方位跟踪。通过卡尔曼滤波，消除了状态空间模型中引力的影响，消除了由不精确引力引起的误差。此外，生成的目标位置平滑且无纹波。混合视觉/SDINS设计的实验结果表明，工具尖端在各个方向上的位置误差约为1毫米RMS。如果视觉系统的采样率从20 fps降低到5 fps，那么定位误差在许多应用中仍然是可以接受的。
This research concerns the development of asmart sensory system for tracking a hand-held moving device to millimeteraccuracy, for slow or nearly static applications over extended periods of time.Since different operators in different applications may use the system, theproposed design should provide the accurate position, orientation, and velocityof the object without relying on the knowledge of its operation andenvironment, and based purely on the motion that the object experiences. Thisthesis proposes the design of the integration a low-cost Local PositioningSystem (LPS) and a lowcost StrapDown Inertial Navigation System (SDINS) withthe association of the modified EKF to determine 3D position and 3D orientationof a hand-held tool within a required accuracy. A hybrid LPS/SDINS combines andcomplements the best features of two different navigation systems, providing aunique solution to track and localize a moving object more precisely. SDINSprovides continuous estimates of all components of a motion, but SDINS losesits accuracy over time because of inertial sensors drift and inherent noise.LPS has the advantage that it can possibly get absolute position and velocityindependent of operation time; however, it is not highly robust, iscomputationally quite expensive, and exhibits low measurement rate. Thisresearch consists of three major parts: developing a multi-camera vision systemas a reliable and cost-effective LPS, developing a SDINS for a hand-held tool,and developing a Kalman filter for sensor fusion. Developing the multi-cameravision system includes mounting the cameras around the workspace, calibratingthe cameras, capturing images, applying image processing algorithms andfeatures extraction for every single frame from each camera, and estimating the3D position from 2D images. In this research, the specific configuration forsetting up the multi-camera vision system is proposed to reduce the loss ofline of sight as much as possible. The number of cameras, the position of thecameras with respect to each other, and the position and the orientation of thecameras with respect to the center of the world coordinate system are thecrucial characteristics in this configuration. The proposed multi-camera visionsystem is implemented by employing four CCD cameras which are fixed in the navigationframe and their lenses placed on semicircle. All cameras are connected to a PCthrough the frame grabber, which includes four parallel video channels and isable to capture images from four cameras simultaneously.
As a result of this arrangement, a widecircular field of view is initiated with less loss of line-ofsight. However,the calibration is more difficult than a monocular or stereo vision system. Thecalibration of the multi-camera vision system includes the precise cameramodeling, single camera calibration for each camera, stereo camera calibrationfor each two neighboring cameras, defining a unique world coordinate system,and finding the transformation from each camera frame to the world coordinatesystem. Aside from the calibration procedure, digital image processing isrequired to be applied into the images captured by all four cameras in order tolocalize the tool tip. In this research, the digital image processing includesimage enhancement, edge detection, boundary detection, and morphologicoperations. After detecting the tool tip in each image captured by each camera,triangulation procedure and optimization algorithm are applied in order to findits 3D position with respect to the known navigation frame. In the SDINS,inertial sensors are mounted rigidly and directly to the body of the trackingobject and the inertial measurements are transformed computationally to theknown navigation frame. Usually, three gyros and three accelerometers, or athree-axis gyro and a three-axis accelerometer are used for implementing SDINS.The inertial sensors are typically integrated in an inertial measurement unit(IMU). IMUs commonly suffer from bias drift, scale-factor error owing tononlinearity and temperature changes, and misalignment as a result of minormanufacturing defects. Since all these errors lead to SDINS drift in positionand orientation, a precise calibration procedure is required to compensate forthese errors. The precision of the SDINS depends not only on the accuracy ofcalibration parameters but also on the common motion-dependent errors. Thecommon motion-dependent errors refer to the errors caused by vibration, coningmotion, sculling, and rotational motion. Since inertial sensors provide thefull range of heading changes, turn rates, and applied forces that the objectis experiencing along its movement, accurate 3D kinematics equations aredeveloped to compensate for the common motion-dependent errors. Therefore,finding the complete knowledge of the motion and orientation of the tool tiprequires significant computational complexity and challenges relating toresolution of specific forces, attitude computation, gravity compensation, andcorrections for common motiondependent errors.
The Kalman filter technique is a powerfulmethod for improving the output estimation and reducing the effect of thesensor drift. In this research, the modified EKF is proposed to reduce theerror of position estimation. The proposed multi-camera vision system data withcooperation of the modified EKF assists the SDINS to deal with the driftproblem. This configuration guarantees the real-time position and orientationtracking of the instrument. As a result of the proposed Kalman filter, theeffect of the gravitational force in the state-space model will be removed andthe error which results from inaccurate gravitational force is eliminated. Inaddition, the resulting position is smooth and ripple-free. The experimentalresults of the hybrid vision/SDINS design show that the position error of thetool tip in all directions is about one millimeter RMS. If the sampling rate ofthe vision system decreases from 20 fps to 5 fps, the errors are stillacceptable for many applications.
- 引言
- 捷联惯性导航系统
- 本地定位系统
- 扩展卡尔曼滤波器
- 结论与未来工作展望
附录A 微应变IMU详细规范
更多精彩文章请关注公众号：
