基于非线性优化的激光雷达在线标定算法

为解决移动机器人定位过程中激光雷达位姿标定问题,保证机器人在作业过程中保持外参的精确,分析里程计与激光雷达的外参标定问题,提出基于非线性优化方式的激光雷达位姿标定算法.通过移动机器人里程计与激光帧间匹配之间的关系建立非线性优化的数学模型,根据优化方式求取较为精确的激光雷达位姿信息.该方法灵活性强,可适应不同场景进行标定,在实际机器人进行算法验证,验证结果表明,该算法有效可行,能够较为精确得到激光雷达的位姿.     

基于特征匹配的激光雷达和相机外参标定方法

智能驾驶系统的车载传感信息通常由激光雷达信息和相机信息融合而成,精确稳定的外参数标定是有效多源信息融合的基础。为提高感知系统鲁棒性,文章提出一种基于特征匹配的激光雷达和相机标定方法。首先,提出点云数据球心算法和图像数据椭圆算法提取特征点的点云三维坐标和像素二维坐标;接着,建立特征点在激光雷达坐标系和相机坐标系下的点对约束,构建非线性优化问题;最后,采用非线性优化算法优化激光雷达和相机的外参数。利用优化后的外参数将激光雷达点云投影至图像,结果显示,激光雷达和相机联合标定精度横向平均误差和纵向平均误差分别为3.06像素和1.19像素。文章提出的方法与livox＿camera＿lidar＿calibration方法相比,平均投影误差减少了40.8%,误差方差减少了56.4%,其精度和鲁棒性明显优于后者。     

移动机器人里程计系统误差及激光雷达安装误差在线标定

为了降低传感器系统误差所带来的影响,首先建立了差速移动机器人里程计系统误差及激光雷达安装误差数学模型.然后,基于拓展卡尔曼滤波算法,提出了一种里程计系统误差及激光雷达安装误差迭代标定方法,该方法能够在定位的同时对2组误差进行实时标定.通过仿真对该方法进行验证,误差估计有效地收敛到误差真值.实物实验中,误差估计能有效收敛,标定后的航迹推算误差大幅度降低.     

一种针孔相机与三维激光雷达外参标定方法

针对针孔相机和三维激光雷达的外参标定问题,提出了一种新的方法:利用相机坐标系和激光坐标系下棋盘面的对应性,将外参标定转换为三维空间中旋转、缩放矩阵的求解问题.每帧数据可以提供4个约束,最少只需3帧数据可求解外参矩阵.方法原理简单易懂,直观明了地解释了失效退化原因.仿真实验与真实数据实验表明:提出的方法可以得到高精度的外参矩阵,且在采样帧数较少的情况下依旧可以获得很好的标定效果.     

面向地下空间探测的移动机器人定位与感知方法

提出了一种面向地下空间探测的移动机器人定位与感知方法。首先,针对地下空间的结构退化问题,构建了基于因子图的激光雷达／里程计／惯性测量单元紧耦合融合框架;推导了高精度惯性测量单元／里程计的预积分模型,利用因子图算法实现对移动机器人运动状态及传感器参数的同步估计。同时,提出了基于激光雷达／红外相机融合的目标识别方法,能够对弱光照环境下的多种目标进行识别与相对定位。试验结果表明,在结构退化环境中,本文方法能够将移动机器人的定位精度提升50%以上,并对弱光照环境中的目标实现厘米级的相对定位精度。     

开放道路中匹配高精度地图的在线相机外参标定

目的 相机外参标定是ADAS（advanced driver-assistance systems）等应用领域的关键环节。传统的相机外参标定方法通常依赖特定场景和特定标志物,无法实时实地进行动态标定。部分结合SLAM（simultaneous localization and mapping）或VIO（visual inertia odometry）的外参标定方法依赖于点特征匹配,且精度往往不高。针对ADAS应用,本文提出了一种相机地图匹配的外参自校正方法。方法 首先通过深度学习对图像中的车道线进行检测提取,数据筛选及后处理完成后,作为优化问题的输入;其次通过最近邻域解决车道线点关联,并在像平面内定义重投影误差;最后,通过梯度下降方法迭代求解最优的相机外参矩阵,使得像平面内检测车道线与地图车道线真值重投影匹配误差最小。结果 在开放道路上的测试车辆显示,本文方法经过多次迭代后收敛至正确的外参,其旋转角精度小于0.2°,平移精度小于0.2 m,对比基于消失点或VIO的标定方法（精度为2.2°及0.3 m）,本文方法精度具备明显优势。同时,在相机外参动态改变时,所提出方法可迅速收敛至相机新外参。结论 本文方法不依赖于特定场景,支持实时迭代优化进行外参优化,有效提高了相机外参精确度,精度满足ADAS需求。     

融合双目视觉和2D激光雷达的室外定位

高精度的定位对于自动驾驶至关重要. 2D激光雷达作为一种高精度的传感器被广泛应用于各种室内定位系统.然而在室外环境下,大量动态目标的存在使得相邻点云的匹配变得尤为困难,且2D激光雷达的点云数据存在稀疏性的问题,导致2D激光雷达在室外环境下的定位精度极低甚至无法实现定位.为此,提出一种融合双目视觉和2D激光雷达的室外定位算法.首先,利用双目视觉作为里程计提供相对位姿,将一个局部时间窗口内多个时刻得到的2D激光雷达数据融合成一个局部子图;然后,采用DS证据理论融合局部子图中的时态信息,以消除动态目标带来的噪声;最后,利用基于ICA的图像匹配方法将局部子图与预先构建的全局先验地图进行匹配,消除里程计的累积误差,实现高精度定位.在KITTI数据集上的实验结果表明,仅利用低成本的双目相机和2D激光雷达便可实现较高精度的定位,所提出算法的定位精度相比于ORB-SLAM2里程计最高可提升37.9%.     

双舵轮AGV视觉里程计的设计方法

针对双舵轮AGV在地面崎岖不平和轮胎打滑的情况下编码器失效的问题.本文提出一种使用价格低廉的RGB-D相机做视觉里程计的方案,避免了双舵轮AGV直接运动学建模导致里程计航迹推算累积误差过大的问题.本文采用ORB算子对图像进行特征提取和匹配,使用ICP的方法进行位姿估计.然后在Linux+ROS平台下搭建视觉里程计,并且和激光雷达数据融合,通过粒子滤波算法进行定位.最后,分别在不同环境下对比了编码器和视觉里程计的定位效果,并验证了整个系统的鲁棒性.     

散焦模糊量估计的相机加权标定方法

相机标定在计算机视觉领域中有着至关重要的作用.绝大多数相机标定方法假设相机为针孔模型,且需要良好聚焦的图像来保证相机内外参估计的准确性.然而,这些条件会受到相机景深的影响.在薄透镜相机模型假设下,提出了一种加权相机标定的方法,其权重考虑了控制点的模糊量信息.首先对棋盘格标定物上的每一个角点进行散焦模糊量估计,在标定过程中,将散焦模糊量的大小作为一个权重加入到标定能量函数最小化过程中,使得标定精度得到提高.该方法简单高效,不需要额外的数码设备或者特别定做的标定物.在Intel Core i7处理器的计算机下,使用合成数据以及真实数据上进行的实验结果表明,文中方法能够有效减小重投影误差,提高张正友标定方法的标定精度.     

动态环境下基于线特征的RGB-D视觉里程计

基于RGB-D的视觉SLAM(同时定位与建图)算法基本都假设环境是静态的,然而在实际环境中经常会出现动态物体,导致SLAM算法性能的下降.为此,本文提出一种基于线特征的RGB-D视觉里程计方法,通过计算直线特征的静态权重来剔除动态直线特征,并根据剩余的直线特征估计相机位姿.本文方法既可以减小动态物体的影响,又能避免点特征过少而导致的跟踪失效.公共数据集实验结果表明,与现有的基于ORB(orientedFAST and rotated BRIEF)点特征的方法相比,本文方法减小了动态环境下的跟踪误差约30%,提高了视觉里程计在动态环境下的精度和鲁棒性.     

Assessing Ideational and Evaluative Skills and Creative Styles and Attitudes

This article describes the rationale for the multiphase creative problem solving process, and reports the findings from an empirical investigation conducted to facilitate the problem solving of managers. The ideational skills of the managers were assessed before and after training in a complete process of creative problem solving, along with their ideational attitudes, creative problem solving style (i.e., generator, conceptualizer, optimizer, or implementor), and evaluative skill (i.e., ability to recognize original ideas). The most important findings indicated that the training had a significant impact on the evaluative accuracy of the managers. They were significantly more accurate in their judgments about original ideas after training, both in their identification of original ideas and their recognition of unoriginal ideas. After training, the managers also gave more solutions and more original solutions to problems. Finally, several variables (e.g., the “preference for active divergence” attitude, and the conceptualizer process style) seemed to moderate the impact of training. Training was therefore effective, with specific effects that can be predicted from pre-training individual differences in attitudes and process style.     

高等院校行政分析与决策系统的设计与实现

为提高高等院校的管理水平和决策水平,充分利用校园网资 源,开发了高等院校行政财政分析与决策系统。解决了诸如数据的动态查询、自动生成报表 、网络环境下数据共享等技术问题,具有网上数据共享、图形界面友好和安全的保密措施等 特点。     

Semantics and verification of monitors and systems of monitors and processes

In this paper we present a sound and complete semantics for the monitor concept of C.A.R. Hoare. First a method for specification of monitors, introduced by O.-J. Dahl, is reviewed. This method is based on the relation between the historic sequence of monitor procedure calls and the historic sequence of monitor procedure exits. Based on such specifications and our new monitor semantics we present a method by which it is possible to prove that a concrete monitor is an implementation of an abstract one. In the last part of the paper an axiomatic semantics for systems of concurrent processes and monitors is introduced. The method supports verification by separation of concerns: Properties of the communication to and from each process are proven in isolation by a usual Hoare style axiomatic semantics, while abstract monitors are also specified in isolation by the method reviewed in the first part of the paper. These properties of the components of the system are then used in a new proof rule to conclude properties of the complete system. Stein Gjessing received a Ph.D. (actually a Dr. philos.) from the University of Oslo (Norway) in 1985. Presently he is an Associate Professor at the Institute of informatics, University of Oslo, Norway. Dr. Gjessings research interests are in the area of concurrent and distributed programming, operating systems, formal specification and verification and programming languages.     

Stochastic balancing and approximation-stability and minimality

A new method of stochastic model reduction has recently been introduced by Desai et al. [1], [2]. The stability of the reduced order model has not previously, been considered. In this paper, we show that the stability of the reduced order model follows directly from the results of Pernebo and Silverman [3]. It is also shown that the reduced order model is minimal, in the controllability/observability sense. The relevance of this notion of minimality to stochastic minimality is made clear.     

Integrating and automating analysis and optimization

The development of an interface coupling program on personal computers for an analysis software system such as ANSYS, SAP, etc. and an optimization software system, MOST, is presented. By controlling and directing the communications the interface coupler integrates the two programs while retaining their versatility and interactive features. The integrated system is used to solve a numerical example of active noise control for a three-dimensional enclosure, in which an energy density level of control points is minimized by adding the sound source to cancel the unwanted noise. The interface coupling program automates with relatively low cost the iterative process for designing an engineering system, remaining flexible in acoustical modelling and efficient in equation solving. Also, the coupling interface is developed in a general-purpose way so that it can be expanded easily to integrate more analysis software packages of different kinds.