用于无人机室内导航的光流与地标融合方法

针对小型无人机在无卫星导航信号条件下的导航问题,结合光流及地标定位设计了使用摄像头、惯性测量器件、超声测距仪等传感器融合的无人机室内导航方法.文章使用补偿角速率的光流微分法计算帧间像素点小位移,并用前后误差算法提取精度较高的点,避免像素点跟踪错误,提高了光流测速的精度;对得到的光流场用均值漂移算法进行寻优,得到光流场直方图峰值,以此计算光流速度.本文提出了无累积误差的连续地标定位算法,实时测量无人机位置.通过多速率卡尔曼滤波器对观测周期不一致的位置、速度信息进行最优估计.在搭建的八旋翼无人机平台上试验,将位置与速度测量结果分别与激光和PX4FLOW数据对比,结果表明该导航方法可以有效抑制定位跳变与光流测量噪声误差,给出精确的位置与速度估计.     

基于视觉传感器的移动机器人定位算法

针对移动机器人准确实时定位问题,采用视觉传感器Kinect作为信息采集源,提出将光流直接法和特征点匹配进行卡尔曼融合的视觉里程计方法。利用光流直接法快速进行帧间小规模运动估计,同时利用特征点匹配法对关键帧之间进行运动估计,纠正光流法带来的定位误差。结果表明:融合算法能够克服光流法精度差、误差累积等因素和特征点法实时处理速度慢等缺点,突出了特征点匹配精度高和局部地图消除误差累积等优点。提出的方法能提供较准确的实时定位信息,具有一定的鲁棒性。     

基于视觉的四旋翼无人机自主定位与控制系统

针对GPS(global positioning system)信号缺失环境下无人机(UAV)自主飞行控制问题,设计了一种基于视觉的自主定位与控制方法.首先通过增加特征点提取数量和优化关键帧存储来对传统视觉SLAM(simultaneous localization and mapping)算法进行改进,提高了算法的鲁棒性与通用性.其次,引入光流传感器作为视觉SLAM地图丢失情况下辅助位置信息测量单元,提高无人机飞行控制的安全性,并成功地克服视觉SLAM图像丢失问题和光流法存在的位置漂移问题.然后采用EKF(extended Kalman filter)融合无人机位置和3维加速度信息,得到了较为精确的位置信息,同时提高了信号输出频率.最后,利用上述方法获取的无人机位置信息设计PID(proportion integration differentiation)和RISE(robust integral of the signum of the error)非线性控制器,增加了算法的鲁棒性.为验证该控制策略的有效性,搭建了四旋翼无人机视觉控制系统实验平台.该平台采用嵌入式控制系统架构,使用机载计算机运行所提算法,避免了图像及控制命令在无线传输过程中引起的时间延迟和信号干扰.室外飞行实验表明,此控制方案实现了自主定位与飞行控制功能.     

基于光流传感器的旋翼无人机实时避障系统

光流的研究是计算机视觉及相关研究领域中的一个重要部分。光流在目标对象分割、跟踪、机器人导航、识别等领域有着非常重要的应用。由于旋翼无人机在民用领域的迅猛发展,使得人们对无人机行业提出了更高的要求。针对多旋翼无人机的避障问题,使用光流传感器与改进的势场法相结合的方法,实现了对飞行器的障碍物检测以及局部路径规划。光流传感器采用SAD块匹配法减小了图像计算的运算量,采集到的光流值与IMU数据进行卡尔曼滤波,可以估算出高效且有较强鲁棒性的因无人机平动产生的光流。大量实验表明,避障策略运算量较小,可以保证避障系统的实时性,在距障碍物5 m左右可以迅速规避障碍物。     

融合IMU去除运动模糊的改进光流匹配算法

为进一步提高视觉SLAM中的光流匹配精度和速度,提出一种融合惯性测量单元（inertial measurement unit,IMU）去除运动模糊的改进光流匹配算法。该算法首先利用IMU运动信息计算的点扩散函数去除运动模糊,提高特征点匹配率;其次在LK（Lucas-Kanade）光流的基础上引入梯度误差,并使用图像梯度L1范数作为正则项模拟稀疏噪声,构建代价函数;然后利用IMU预测特征点位置作为该算法初始值,并加入BB（Barzilar-Borwein）步长改进原有的高斯牛顿算法,提高计算速度。实验表明,通过两帧之间比较,该算法的效率和精度均优于LK光流法;然后将该算法集成到VINS-Mono框架,在数据集EuRoC上结果显示,该算法提高了原有框架的定位精度和鲁棒性。     

基于相位光流法的结构振动微小位移测量

随着数码相机的成本降低和普及,基于计算机视觉的非接触式结构健康监测技术越来越受到重视.传统的接触式测量仪器可能导致轻质结构上的质量负荷,并且在大型土木结构上安装和维护成本高且耗时,特别是对于长期应用.作为一种替代的非接触方法,使用数码相机的计算机视觉方法成本相对较低、灵活.针对传统光流方法在微小位移测量精度和稳定性不足的问题,本文提出了一种改进的基于相位的光流法来计算结构的振动位移.在这种方法中,通过引入基于相位的视频运动放大方法中构造的部分复数可操纵滤波器,与视频序列图像的卷积得到结构的局部空间相位信息.接着通过基于相位的光流法计算结构振动的像素位移.最后通过降噪处理和比例因子法将像素位移转换为真实位移.实验室简支梁振动实验结果证明,本文的方法相比于其他光流方法在微小位移计算上具有更高的精确度和稳定性.     

面向无人机自主导航的立体匹配算法

无人机自主导航是无人机发展的必然趋势,立体视觉技术作为一种优秀的环境信息测量技术能够为无人机自主导航提供关键信息.但是,导航图像存在幅度失真,现有立体匹配方法的匹配精度较低,针对Census变换舍弃了图像像素色彩信息而造成的误匹配问题,本文提出了一种Census变换和图像色彩信息相结合的联合匹配算法,并经过理论分析提出了正交积分的方法以提高算法的实时性.首先,将Census变换和图像色彩信息联合,构造初始匹配代价;然后,采用改进的自适应窗口作为代价累积窗口,并使用正交积分思想提高累积速度;最后,经过视差提精,获得最终的视差图.实验结果表明:本文算法对幅度失真图像的匹配误差比单独使用Census变换提高了40%～50%,算法的运算时间提速了3～12倍,与Census变换和图像灰度单独作为匹配代价时相比,该方法具有更高的匹配精度,对幅度失真有很强的鲁棒性,能够较好地应用于无人机自主导航场景中.     

基于人脸识别与光流追踪的移动机器人视觉导航方法

针对移动机器人视觉导航中跟踪目标丢失的问题,提出了基于人脸识别与稀疏光流算法(KLT)结合的移动机器人视觉导航方法(FR-KLT视觉导航方法)。采用OpenCV库中的Haar特征提取人脸识别算法实时检测识别目标人脸,通过Harris角点检测获取目标人体特征点,对目标人体进行精准定位;KLT光流追踪法测算目标移动趋势,并预测目标下一刻大致位置。目标人体位置变动时移动机器人对目标进行实时追踪导航。通过Pioneer-LX机器人在真实环境下试验,验证了该方法准确识别并跟踪目标的实时性和有效性。     

融合双目视觉与惯导信息的高效视觉里程计算法

为提高视觉里程计(VO)在大尺度环境下运行的实时性,提出一种融合双目视觉与惯导信息的视觉里程计算法,主要由前端位姿跟踪和后端局部地图优化两个线程组成.位姿跟踪线程首先使用惯导信息辅助光流法进行帧间特征点跟踪并估计相机初始位姿;接着通过最小化图像光度误差获取当前帧像素点与局部地图点的对应关系;而后最小化当前帧上局部地图点的重投影误差和惯性测量单元(IMU)预积分误差,得到当前帧准确的位姿估计.后端局部地图优化线程对滑动窗口内的关键帧提取特征点并三角化新地图点,使用光束平差法(BA)对逆深度参数化表示的地图点位置、关键帧位姿、速度以及陀螺仪和加速度计零偏进行滑窗优化,为前端提供更加精确的局部地图相机位姿和环境信息.在EuRoC数据集上的实验表明,相比于ORB-SLAM2、ORB-SLAM3算法,该融合双目视觉与惯导信息的视觉里程计算法的定位精度略有下降,但可以较大程度地提高位姿跟踪的实时性.     

基于光流和惯性导航的小型无人机定位方法

为了提高小型无人机定位精度,提出了一种基于光流系统和惯性导航系统相结合的定位方法.在介绍光流传感器工作原理,建立光流数学模型和分析光流测量误差之后,提出了采用双光流传感器测量飞行器的高度和速度,并将二维光流信息与捷联惯性导航系统(SINS)通过扩展卡尔曼滤波器(EKF)进行数据融合,得到实时的位置、速度和姿态.室内实验结果表明:该方案能够有效地减小导航中的位置、速度和姿态误差,提高定位精度.     

A visual integrated navigation for precise position estimation

In this article, we propose a visual integrated navigation method which is composed of a relative position estimation and an absolute position estimation. The aim is to estimate precise position and overcome accumulative position errors for long-distance and long-time visual navigation. The relative position estimation measures relative displacement and position by a simple optical flow method, but its accumulative position errors would increase with time. The absolute position estimation aids the relative position estimation for correcting these accumulative errors timely. The absolute position estimation employs a Kalman filter based on image entropy, which can continuously correct accumulative errors according to the actual and predicted image entropy. As a result, the proposed navigation method can supply precise position information all along the navigational path, which has been proven by some experiments in the paper.     

A mobile robot employing insect strategies for navigation

The ability to navigate in a complex environment is crucial for both animals and robots. Many animals use a combination of different strategies to return to significant locations in their environment. For example, the desert ant Cataglyphis is able to explore its desert habitat for hundreds of meters while foraging and return back to its nest precisely and on a straight line. The three main strategies that Cataglyphis is using to accomplish this task are path integration, visual piloting and systematic search. In this study, we use a synthetic methodology to gain additional insights into the navigation behavior of Cataglyphis. Inspired by the insect’s navigation system we have developed mechanisms for path integration and visual piloting that were successfully employed on the mobile robot Sahabot 2. On the one hand, the results obtained from these experiments provide support for the underlying biological models. On the other hand, by taking the parsimonious navigation strategies of insects as a guideline, computationally cheap navigation methods for mobile robots are derived from the insights gained in the experiments.     

Bioinspired Visuomotor Convergence

In this paper, wide-field integration methods, which are inspired by the spatial decompositions of wide-field patterns of optic flow in the insect visuomotor system, are explored as an efficient means to extract visual cues for guidance and navigation. A control-theoretic framework is developed and used to quantitatively link weighting functions to behaviorally-relevant interpretations such as relative orientation, position, and speed in a corridor environment. It is shown through analysis and demonstration on a ground vehicle that the proposed sensorimotor architecture gives rise to navigational heuristics, namely, centering and speed regulation, which are observed in natural systems.      

基于抑制标度因数误差的捷联旋转监控方案研究

旋转监控技术通过对惯性元件常值误差的抑制,能有效提高捷联惯导系统的导航定位精度,但由于标度因数误差的存在,降低了旋转调制效果.通过对单轴旋转方式下标度因数调制效果的分析,推导了单轴旋转下标度因数误差激励的角速率误差,并总结出旋转位置与角速率误差的对应规律,进而提出了最佳的双轴旋转方案,最大程度地抑制标度因数误差对捷联惯...     

IMU标定数学建模及误差分析

惯性测量单元（IMU）标定路径设计和数据处理方法取决于IMU标定数学模型,安装误差是决定IMU标定模型的重要因素。针对工程中加速度计和陀螺相对载体安装方式的不同,提出一种通过坐标系转换矩阵建立IMU标定数学模型的方法,推导IMU标定模型误差与载体角速度和加速度之间的关系,分析IMU标定模型误差对捷联惯性导航系统导航参数的影响,并利用转台提供的位置信息设计IMU标定路径和数据处理方法。仿真和转台实验结果表明：IMU标定数学模型误差引起捷联惯性导航系统速度误差、位置误差和姿态误差;安装误差的表现形式决定了IMU标定模型误差对系统导航精度的影响。     

Metric embedding of view-graphs

Most recent robotic systems, capable of exploring unknown environments, use topological structures (graphs) as a spatial representation. Localization can be done by deriving an estimate of the global pose from landmark information. In this case navigation is tightly coupled to metric knowledge, and hence the derived control method is mainly pose-based. Alternative to continuous metric localization, it is also possible to base localization methods on weaker constraints, e.g. the similarity between images capturing the appearance of places or landmarks. In this case navigation can be controlled by a homing algorithm. Similarity based localization can be scaled to continuous metric localization by adding additional constraints, such as alignment of depth estimates. We present a method to scale a similarity based navigation system (the view-graph-model) to continuous metric localization. Instead of changing the landmark model, we embed the graph into the three dimensional pose space. Therefore, recalibration of the path integrator is only possible at discrete locations in the environment. The navigation behavior of the robot is controlled by a homing algorithm which combines three local navigation capabilities, obstacle avoidance, path integration, and scene based homing. This homing scheme allows automated adaptation to the environment. It is further used to compensate for path integration errors, and therefore allows to derive globally consistent pose estimates based on “weak” metric knowledge, i.e. knowledge solely derived from odometry. The system performance is tested with a robotic setup and with a simulated agent which explores a large, open, and cluttered environment. This work is part of the GNOSYS (FP6-003835-GNOSYS) project, supported by the European Commission.     

Biologically‐Inspired Visual Simulation of Insect Swarms

Representing the majority of living animals, insects are the most ubiquitous biological organisms on Earth. Being able to simulate insect swarms could enhance visual realism of various graphical applications. However, the very complex nature of insect behaviors makes its simulation a challenging computational problem. To address this, we present a general biologically‐inspired framework for visual simulation of insect swarms. Our approach is inspired by the observation that insects exhibit emergent behaviors at various scales in nature. At the low level, our framework automatically selects and configures the most suitable steering algorithm for the local collision avoidance task. At the intermediate level, it processes insect trajectories into piecewise‐linear segments and constructs probability distribution functions for sampling waypoints. These waypoints are then evaluated by the Metropolis‐Hastings algorithm to preserve global structures of insect swarms at the high level. With this biologically inspired, data‐driven approach, we are able to simulate insect behaviors at different scales and we evaluate our simulation using both qualitative and quantitative metrics. Furthermore, as insect data could be difficult to acquire, our framework can be adopted as a computer‐assisted animation tool to interpret sketch‐like input as user control and generate simulations of complex insect swarming phenomena.     

Advanced state estimation for navigation of automated vehicles

For the emerging topic of automated and autonomous vehicles in all major sectors, reliable and accurate state estimation for navigation of these vehicles becomes increasingly important. Inertial navigation, aided with measurements from global navigation satellite systems (GNSS), allows high-rate and low-cost estimation of position, velocity and orientation in real-time applications. As the available satellite constellations for navigation are modernized and their number is rising, usage of multi-constellation, dual-frequency and integration of correction data lead to increased accuracy, especially in areas with partial shadowing. Different coupling methods, e.g. tightly- and loosely-coupled integrations, were developed to combine inertial and GNSS measurements. Also different error estimation filters were applied to the navigation problem, and evaluated against each other. For the typical navigation task, the objective is to choose a suitable algorithm for the specific requirements of the target application, and deploy it using an appropriate implementation strategy. This contribution gives a short introduction into the field of aided inertial navigation techniques, provides useful hints for implementation, and evaluates their performance in experiments using two different railway vehicles, an autonomous maritime vessel, and an unmanned aerial quadrotor.     

智能车辆SINS/DGPS/光电测速仪组合导航的研究

为适应自主驾驶车辆的高精度、高频率与高可靠性的导航要求,本文对智能车辆的多传感器组合导航进行了研究.提出了一种双向光电测速仪和CP-DGPS共同辅助SINS的智能车辆组合导航方法,进行了仿真试验验证.结果表明,该组合导航系统能为智能车辆提供丰富的导航信息,并具有100 Hz的高频输出、厘米级的导航精度和较强的容错能力,当GPS较长时间中断时,通过SINS/光电测速仪的组合仍能为智能车辆提供可靠的导航数据.     

Path Integration Mechanism with Coarse Coding of Neurons

Many animals can return home accurately after exploring for food using their own homing navigation algorithm. Path integration plays a critical role in homing navigation. It is believed that animals are able to recognize their relative location from the nest by accumulating both distance and direction experienced during their travel. We tested possible patterns of neuronal organization for a path integration mechanism. The neural networks consisted of a circular array of neurons, following population coding. We describe here a neural model of path integration involving a relatively small number of neurons and discuss how well the model operates for homing navigation. Robotic simulations suggest that a neural structure with only a few sensor neurons can successfully handle the path integration needed for homing navigation.