A neural multiclassifier system for object recognition in robotic vision applications

A robotic vision system has been developed that is based on the optical recognition of objects to be manipulated, that are located in the workspace of the robotic manipulator. The developed system has a low development and operation cost, is controlled via an external computer and operates in an unstructured complex environment. The vision system is desired to recognize the objects, which are placed in the workspace and also to identify the exact position and orientation of each particular object, in order to lead the robot manipulator system. For the recognition of objects, a high performance NEural MUlticlassifier System (NEMUS) is presented, which combines multiple classifiers that operate on different feature sets. NEMUS is characterized by a great degree of modularity and flexibility and is very efficient for demanding and generic pattern recognition applications.     

基于级联式Faster RCNN的三维目标最优抓取方法研究

机器人在三维目标识别和最优抓取方面的难点在于复杂的背景环境以及目标物体形状不规则,且要求机器人像人一样在识别不同三维目标的同时要确定该目标的最佳抓取部位的位姿。提出一种基于级联式模型的深度学习方法来识别目标物体及其最优抓取位姿。第1级提出了改进的Faster RCNN模型,该模型能识别成像小的目标物体,并能准确对其进行定位;第2级的Faster RCNN模型在前一级确定的目标物体上寻找该目标物体的最优抓取位姿,实现机器人的最优抓取。实验表明该方法能快速且准确地找到目标物体并确定其最优抓取位姿。     

基于三维检测网络的机器人抓取方法

机器人抓取任务中面对的是不同形状和大小的物体,而散落在场景中的物体会有不同的姿态和位置,这对机器人抓取中计算物体位姿任务提出了较高的挑战。针对于此,本文设计了一种基于三维目标检测的机器人抓取方法,弥补了基于二维图像识别引导机器人抓取任务中对视角要求较高的缺陷。首先,设计了一种卷积神经网络在RGB图像中识别物体,并回归出物体三维包围盒、物体中心点;其次,提出一种计算机器人抓取物体最佳姿势的策略;最后,控制机器人进行抓取。在实际场景中,使用本文设计的三维检测网络,三维目标检测精度达到88%,抓取成功率达到94%。综上所述,本文设计的系统能有效找到机器人合适的抓取姿势,提高抓取成功率,满足更高的抓取任务要求。     

一种新型主动视觉系统的研究

为了提高主动视觉平台的适用性、减小其复杂度,提出了一种新的主动视觉系统结构。利用摄像机作为传感器代替传统的跟踪器计算头盔的位姿。对头盔识别采用了合作目标的方式,提高了识别率和位姿计算速度。采用了一种合作目标亚像素精度定位的方法以提高合作目标三维空间的定位精度。利用PID算法对主动视觉平台进行姿态的控制。最后,实验结果证明了该方法的可行性。     

Vision Based Simulation and Experiment for Performance Tests of Robot

The feature-based visual servoing approach has been used to control robot through vision. In order to find the position of the end effector by vision and through robot performance tests, computational kinematic approach has been used. The software carries out the duty of environment simulation and operation of an industrial robot. The disputes related to image capturing, image processing, target recognition, and how to control robot by vision system ability have been carried out in the simulation tests. The vision based program has been defined in such a way that it can be carried out by a real robot with the least changes. In the experiment, the vision system will recognize the target and control the robot by obtaining images from environment and processing them. At the first stage, images from environment are changed to a grayscale mode then it can diverse and identify objects and noises by using a threshold objects which are stored in different frames and then the main object will be recognized. This will control the robot to achieve the target. Finally, the issues of robot performance tests based on the two standards ISO 9283 and ANSI-RIA R15.05-2 have been accomplished through simulator program using vision system over the 3P robot for evaluating the path-related characteristics of the robot. To evaluate the performance of the proposed method experimental test is carried out.     

On-machine 3D vision system for machining setup modeling

In computer numerical control machine tools, using machining simulation to prevent collision becomes more and more popular due to its efficiency and low cost. However, if the entire digital model of the machining setup does not exist, the simulation is not applicable. As a result, the operator has to manually check the numerical control program, which is a time-consuming and error-prone process. In this paper, an on-machine vision system is presented to quickly construct the digital model based on the actual machining setup. The total construction for a complex setup can be done within a few minutes. Several key technologies have been developed. First, a 2D edge feature detection algorithm has been designed which will extract the edges of the object of interest by processing both the real and virtual images. Second, a stereo vision system is developed which will obtain the three-dimensional (3D) edge data of the object of interest. A new algorithm is presented to solve correspondence, which is the key problem of the stereovision system. Furthermore, the 3D object recognition algorithm is developed to let the system intelligently search for the matched solid model in the database and import it into the virtual environment with an accurate pose. Finally, experiments are carried out to test the developed system.     

基于ROS的机器人宇航员模型构建与仿真验证

随着我国空间站项目的稳步推进和机器人技术的发展,研发可以协助甚至代替宇航员进行危险舱外操作的机器人宇航员成为亟待研究的课题。首先建立了双臂仿人机器人宇航员的URDF模型,而后基于机器人操作系统提供的三维仿真工具Rviz,构建机器人宇航员的三维虚拟运动仿真环境,并进行了仿真实验。实验表明,机器人宇航员可以完成物体抓取、搬运等任务,证实了机器人宇航员模型结构的合理性。     

Quadric segmentation and fitting of data captured by a laser profile scanner mounted on an industrial robot

Applications like geometric reverse engineering, robot vision and automatic inspection require sets of points to be measured from the surfaces of objects and then processed by segmentation and fitting algorithms to establish shape parameters of interest. In industrial applications where speed, reliability and automatic operation is of interest a measuring system based on a laser profile scanner mounted on an industrial robot can be of interest. In earlier publications we have presented such a system and also a segmentation algorithm for planar surfaces using 2D profile data in combination with robot poses. Due to the data reduction offered by this approach the segmentation algorithm computes faster than algorithms based on 3D point sets alone. Encouraged by the results we have now developed a segmentation algorithm for two different quadric surfaces also based on 2D profiles in combination with robot poses. This paper presents the new algorithm together with test results and also an interesting observation that points to future work.     

Vision guided dual arms robotic system with DSP and FPGA integrated system structure

Usually, a humanoid robot has two arms and stereo vision system to execute human daily actions. It has complicate mechanism and mechatronics control system structure. The hardware control structure should be planned ingeniously to execute the complicate computation of 3D image processing and manipulate a multi degree of freedom dual arms motion control, especially for mobile robot system. Here a 7 DOF dual arms robot with FPGA hardware control structure and a digital signal processor (DSP) based CMOS stereo vision system are designed and built in our lab. The intelligent fuzzy sliding mode control strategy is employed to establish the visual guided robotic motion control software. This low cost humanoid robotic system has compact control structure and mechanism integration for mobile application purpose. Object detecting and tracking schemes in 3D space were developed for locating the target position and then guided the robot arm to pick and place objects or track the specified moving target. Experimental results show that this delicate robotic system has basic humanoid function.     

Roboscan: a combined 2D and 3D vision system for improved speed and flexibility in pick-and-place operation

We describe Roboscan, a Robot cell that combines 2D and 3D vision in a simple device, to aid a Robot manipulator in pick-and-place operations in a fast and accurate way. The optical head of Roboscan combines the two vision systems: the camera is used “stand-alone” in the 2D system, and combined to a laser slit projector in the 3D system, which operates in the triangulation mode. The 2D system, using suitable libraries, provides the preliminary 2D information to the 3D system to perform in a very fast, flexible and robust way the point cloud segmentation and fitting. Roboscan is mounted onto an anthropomorphic, 6-DOF Robot manipulator. The most innovative part of the system is represented by the use of robust 2D geometric template matching as a means to classify 3D objects. In this way, we avoid time-consuming 3D point cloud segmentation and 3D object classification, using 3D data only for estimating pose and orientation of the robot gripper. In addition, a novel approach to the template definition in the 2D geometric template matching is proposed, where the influence of surface reflectance and colour of the objects over the definition of the template geometry is minimized. We describe the procedures for 2D and 3D vision of Roboscan, together with the calibration procedures that have been implemented. We also present a set of tests that show the performance of the system and its effectiveness in a number of pick-and-place operations.     

An ultrasonic collision detection system

Robots in manufacturing are increasingly being called on to do complex tasks that require intelligence beyond merely following a preprogrammed path. In robotic assembling of mechanisms, welding, machine tending and other tasks, sensing enables robots to adapt to their environments. In this research, an ultrasonic collision detection system for an industrial robot was designed, constructed and tested. Two ultrasonic transducers and ranging modules were mounted on the robot wrist to detect and prevent collisions with objects placed in the end effector's path. Experiments were conducted to determine objectto-robot distance as a function of robot speed after the robot had stopped. Two robot motions and two loads were studied. Statistical methods of stopping distance vs. robot speed are presented and will be useful in planning robot tool paths. This ultrasonic collision detection system can be used on stationary and mobile robots, automatic-guided vehicles, and other manufacturing applications.     

铁路场景三维点云分割与分类识别算法

铁路限界侵入检测对保障高速铁路安全具有重要意义,基于激光三维点云分割与分类识别的异物侵入检测具有准确、直观的优点,在诸如隧道口和站台的铁路重点区域监测中具有广泛应用前景。设计了一种带动二维激光雷达进行俯仰运动的装置用于铁路三维点云的采集,基于法线方向一致性原则提出采用区域生长分割算法解决欧氏聚类分割和随机采样一致性(RANSAC)分割造成的过分割和欠分割问题;针对分割后的单物体点云,提出利用视点特征直方图(VFH)进行不同目标的三维点云特征提取,基于不同物体VFH建立KD树,并利用最近点搜索方法完成单物体点云分类识别。铁路场景典型物体的分类实验结果表明,本算法对铁路场景典型物体的分类识别准确率大于90%。     

基于Java3D的机械臂三维虚拟场景创建及运动学仿真

针对常规方法所构建的机械臂运动仿真可视化虚拟环境并非基于网络特性的问题,研究并采用了Java3D与VRML相结合的方法。通过Loader接口导入VRML造型文件并将其添加到Java3D场景图中进行装配,借助Java语言强大的事件处理模型,结合Java3D中的Interpolator内插器对象和Alpha对象,实现了简单的人机交互操作。最后以实验室的某型号四自由度机械臂为例,成功实现了一个基于Java3D的交互式可视化机械臂运动仿真虚拟环境创建,并采用标准的D-H参数法进行了机械臂正、逆运动学分析与仿真。研究结果表明,该基于Java3D开发的机械臂三维运动仿真系统,不仅效果逼真,而且适于网络的传输,实现了既定的效果;该研究工作在机械臂仿真中具有较大的应用价值。     

三维曲线识别技术研究

在两维曲线识别技术的基础上进行三维曲线识别技术的研究,通过简化摄像头模型,以及限制摄像头的安装位置,利用两摄像头坐标系之间的关系,推导出一种基于两维曲线识别的三维曲线识别算法,并分析了该种算法的可行}生,在机器人视觉算法研究中有一定的价值.     

图像反馈机器人视觉伺服系统理论与实验研究

对机器人视觉伺服系统的研究是机器人领域中的重要内容之一,其研究成果可应用在机器人自动避障、轨线跟踪和运动目标跟踪等问题中。本文分析了基于图像雅克比矩阵的机器人视觉伺服方法的基本原理,采用了基于图像的视觉伺服方法,直接利用图像特征来控制机器人运动,构建了自由度GRB-400工业机器人图像反馈视觉伺服系统。采用该系统进行了机器人跟踪两维平面运动目标的实验,结果验证了该方法的有效性。     

基于MCU和CPLD的智能移动机器人控制系统

针对移动机器人控制系统设计和开发要求的复杂性,从成本低、易开发、易调试和高集成性的角度出发,提出了一种基于微处理机控制单元（MCU）和复杂可编程逻辑器件（CPLD）的智能移动机器人控制系统设计。给出了自动避障、电机驱动控制、红外遥控等关键功能的设计实现方法,着重对基于CPLD的超声波检测模块、红外编码模块、壁障模糊控制器的设计等进行了详细的论述。同时,还给出了系统主程序和超声测距子程序软件设计流程。实践结果表明,该移动机器人控制系统可使硬件结构大大简化,并具有功能丰富、集成度高、性价比高等特点。     

The position/Orientation determination of a mobile-task robot using an active calibration scheme

A new method of estimating the pose of a mobile-task robot is developed based upon an active calibration scheme. The utility of a mobile-task robot is widely recognized, which is formed by the serial connection of a mobile robot and a task robot. To be an efficient and precise mobile-task robot, the control uncertainties in the mobile robot should be resolved. Unless the mobile robot provides an accurate and stable base, the task robot cannot perform various tasks. For the control of the mobile robot, an absolute position sensor is necessary. However, on account of rolling and slippage of wheels on the ground, there does not exist any reliable position sensor for the mobile robot. This paper proposes an active calibration scheme to estimate the pose of a mobile robot that carries a task robot on the top. The active calibration scheme is to estimate a pose of the mobile robot using the relative position/orientation to a known object whose location, size, and shape are known a priori. For this calibration, a camera is attached on the top of the task robot to capture the images of the objects. These images are used to estimate the pose of the camera itself with respect to the known objects. Through the homogeneous transformation, the absolute position/orientation of the camera is calculated and propagated to get the pose of a mobile robot. Two types of objects are used here as samples of work-pieces : a polygonal and a cylindrical object. With these two samples, the proposed active calibration scheme is verified experimentally.     

星空背景下空间目标的快速识别与精密定位

为了提高光电望远镜系统测量空间目标的实时性和定位精度,建立了空间目标快速识别与精密定位系统,讨论了空间目标运动特性、快速识别、星像质心计算和天文定位等算法。首先,介绍了空间目标光电观测系统的系统构成和工作原理;深入分析了空间目标在CCD视场中的运动特性,提出了一种空间目标快速识别算法。然后,结合帧间差分法和数学形态学等,完成了空间目标的快速识别。最后,研究了天文定位算法,采用Tycho-2星表,实现了空间目标的精密定位。实验结果表明:空间目标快速识别处理时间约为10ms,实时天文定位处理时间约为25ms,实时天文定位精度优于4″。得到的结果满足空间目标监视技术的实时性好、精度高、稳定可靠等要求。     

仿射不变的运动目标识别方法

提出了一种在三维空间内对运动目标识别的方法,该方法对目标平移、旋转、尺度及视角变化具有良好的不变性。将目标轮廓曲线用仿射不变量参数化,待识别目标的参数化轮廓线正交投影后,可使其与库存同一型号的目标轮廓线保持较强的相关性。最后,用归一化相关函数作为识别准则进行了实验。实验结果表明,对于空间姿态变化前后的目标,它们的相关结果＞95％,该方法能够以较高的识别率识别出图像序列中各种变化的同一目标。     

肌电信号控制仿生手的研究

研究的目标是利用肌电信号实时控制仿生手的运动。使用所设计的系统，实现了肌电信号采集、处理和对所设计的仿生手的控制。通过试验验证了肌电信号的统计特性，用不同操作者的不同肌肉来控制和训练仿生手抓物体，试验结果表明仿生手的失误率小于0．5％。