【作者】 赵立强；

【导师】 孔令富；

【作者基本信息】 燕山大学 ， 计算机应用技术， 2010， 博士

【摘要】 机器视觉融合视觉传感技术、图像理解技术、模式识别技术以及计算机控制技术,用机器代替人眼来做测量和判断,是实现系统智能化、自动化、信息化的先进技术。机器视觉系统的应用,能大大提高装备使用效率、精度及可靠性。针对目前并联机器人无法感知工作状态、环境的变化,处于“盲”工作状态的问题,本文从并联机器人机构的特点和机器人工作空间视觉监测出发,提出并建立了一种基于圆形导轨的双目主动视觉监测平台,在国家高技术研究发展计划(863计划)的支持下,研制了并联机器人双目主动视觉监测平台的实验样机,并对视觉监测平台的机构运动理论、视觉测量计算理论和关键应用技术等问题进行了深入研究。主要研究内容如下:(1)提出了新型并联机器人双目主动视觉监测平台的设计,建立平台机构的运动学模型。打破传统视觉监测平台两摄像机基线相对固定的设计,首次提出双摄像机在圆形轨道灵活运动的视觉机构设计,可以实现基线可调、光轴可调、大范围多角度观察、灵活避障的双目视觉的能力;针对实验样机,建立了视觉系统的机构学模型,给出了视觉系统的位置、速度、加速度运动(正、反解)模型,实验测试结果验证了模型的正确性。(2)建立了并联机器人双目主动视觉系统二级标定体系。其中,以平面棋盘靶作为一级标定靶,采用非线性摄像机标定算法,用于视觉系统摄像机内外参数的标定,实现系统的初始化;为解决并联机器人操作过程中出现机器人操作部件和加工部件对棋盘靶的遮挡问题,满足视觉系统动态标定的需要,设计了一种均匀分布在圆形导轨内侧,由六组立体标靶块构成的二级标定靶。然后,利用建立的标定体系,对并联机器人双目主动视觉机构的各结构参数、控制当量以及机构精度进行了测定。(3)通过两支链的协调控制,并联机器人双目主动视觉监测平台实现对环境和目标的双目最佳观测。为此,定义了系统正视观测模式,提出了摄像机观测注意点转移的支链控制策略,建立了正视模式下实现视觉跟踪的双目协调运动控制模型。实验结果验证了该观测模式和控制策略的有效性和可行性。(4)针对视觉监测系统在并联机器人中的应用,研究了视觉导航和运动监测问题。首先,以并联雕刻机器人刀具导向理想加工点为背景,提出了利用两摄像机采集的刀具的图像信息,直接计算刀具运行轨迹的简化方法。其次,为实现并联机器人运动检测,提出了基于圆台型标靶的加工刀具的位姿检测和基于扩展卡尔曼滤波的运动参数分析方法,并实验验证方法的有效性和可行性。(5)提出了并联机器人双目主动视觉系统软件的总体结构设计方案。软件系统由前台电气控制系统和后台视觉服务系统组成,前台电气控制系统负责视觉平台的电气设备控制和图像采集控制,后台视觉服务系统负责图像分析、控制策略生成,前后台计算机通过共享格式化文本文件实现信息交换。更多还原

【Abstract】 Machine vision is a comprehensive technology fusing vision sensor, image understanding, pattern recognition and computer control etc. Using machine vision to replace human eyes with measurement and judgment, It is advantageous to achieve system intelligence, automation and information. The application of machine vision system can greatly improve the level of intelligence and automation, and enhance efficiency, accuracy and reliability of equipments. To solve the problem that parallel robot is unable to perceive the changes in working conditions and environment, this paper puts forward a binocular active visual monitoring platform based on circular orbit adapt to the parallelism of parallel mechanism and the requirement of robot workpace monitoring. With the support of the National High Technology Research and Development Program(863), a prototype machine of binocular active visual monitoring platform of parallel mechanism is developed, and the problems including kinematic theory, vision measurement and others key technologies have been deeply studied. The main research contents are as follows:(1)A new design of binocular active visual monitoring platform of parallel Mechanism is proposed, and its motion model is also established, which breaks the traditional design that the baseline of the camera in monitoring platform must be fixed. The design that the two cameras freely move along circle orbit is presented for the first time, which can realize more flexible binocular vision such as baseline adjustable capacity, optical axis adjustable capacity, large-scale multi-angle observations and vision avoidance. Aiming at experimental prototype, the mechanism model of visual system is established, and the position, velocity, acceleration kinematics model (forward and inverse) of vision system is also given. The experimental results verify the correctness of the model. (2)Two levels calibration architecture of binocular active visual monitoring platform of parallel mechanism is established. A planar chess target is the first level calibration, and a non-linear calibration algorithm is proposed to solve intrinsic parameters and extrinsic parameters of cameras and realize system initialization. To solve the occlusion issue caused by the end effectors and machined body in the calibration process by a planar chess target, the second calibration target, which consist of six groups of 3D target distributing uniformly inboard circular orbit, is designed to satisfy the requirement of dynamic calibration for vision system. Then using this two level calibration architecture, it is implemented measuring structural parameters and control equivalent and machining accuracy of binocular active visual monitoring platform of parallel mechanism.(3)Binocular active visual monitoring platform of parallel mechanism can achieve the best observation for the object in the common view of two cameras though harmoniously adjusting its two chains. For this purpose, the binocular emmetropization mode is defined, and A single chain control strategy is improved to transfer a camera attentive focus to the object position, and the binocular cooperative kinematics model is developed to implement visual tracking in binocular emmetropization mode. The experimental results show the effectiveness and feasibility of the observation model and control strategy.(4)Considering some application questions of binocular active visual monitoring platform for parallel mechanism, visual navigation and motion detection are preliminarily explored. Firstly, based on guiding cutting tool of engraving parallel robot to expected positions, a simplified direct method is proposed to plan the path of the cutting tool using its extracted visual information from two cameras. Secondly, the pose detection method of the cutting tool base on a truncated cone-shapetarget and motion parameters analysis method of the cutting tool based on Extended Kalman Filter are successively developed. The experimental results show that the methods are effective and practicable.(5)The overall software design of binocular active visual system of parallel mechanism is given. The whole system is constituted by a foreground control system and a background visual service system front electrical control system. The foreground control system provides some functions of electric equipment control and image acquisition. The background visual service system provides some functions of Image analysis and control strategy generating. Information exchange is implemented through sharing the same format Text file between the foreground and background computers.更多还原