双CCD仿人眼颈系统运动分析与控制研究

【作者】 肖宝平；

【导师】 王宣银；

【作者基本信息】 浙江大学 ， 机械电子工程， 2004， 硕士

【摘要】 90年代以来，计算机技术、网络技术、图像处理、通信技术和多媒体技术得到了飞速发展，多种技术的揉和使得机电一体化产品得到极大进步。在成功研制出全方位目标自动跟踪一体化智能球形摄像机的基础上，作者依托航天技术支撑基金项目(No．HT-2001-zjdx)研究双CCD仿人眼颈系统，主要包括机构设计、运动控制、双目立体视觉测量的相关理论和技术。为课题的进一步研究——三维立体识别和精确跟踪提供实验装置和理论基础。 第一章，阐述了双CCD仿人眼颈系统的研究意义和背景，分析了国内外双目立体视觉系统相关技术的研究和发展状况，概括了本文的主要研究内容。 第二章，介绍了双CCD仿人眼颈系统的总体方案与原理框架；设计并研制出双CCD仿人眼颈机构，建立了数学模型和求解运动学方程。 第三章，研究双CCD仿人眼颈系统测量原理和数学模型。建立了双CCD立体视觉测距光学模型；并利用双目交汇测量方法得出了双CCD仿人眼颈系统立体视觉测距方程，并进行了误差分析。 第四章，研究双CCD仿人眼颈系统的驱动和控制。介绍所研制的四自由度系统运动控制的电路和软件系统。针对步进电机的减速机构所带来的齿侧间隙，提出复位补偿算法，给出了实现方法，并进行了实验，结果表明该方法有效地消除的间隙误差。 第五章，研究摄像机的标定。利用了成像过程中的径向排列约束(RAC)分解摄像机参数，提出了一种基于径向排列约束(RAC)的摄像机标定算法，从而使得摄像机标定不需要依赖于非线性的优化，并通过实验得到了良好的标定结果。 第六章，构造了基于视觉传感器的双CCD立体测量模型，对现有的特征点匹配方法进行了改进，给出了相应的立体测量原理及软件流程设计。 第七章，总结全文并给出主要研究结论，并对一些有待于进一步探讨和研究的问题提出了展望。更多还原

【Abstract】 Computer, network, image processing, communication and multimedia technologies have developed rapidly in the decades. These technologies that are gathered and applied in mechanical and electronic products have improved industry greatly. After successfully researching and manufacturing an intelligent dome camera that can follow and track object in all orientations, supported by aerospace technology support foundation (No.HT2001-zjdx), the author researchs on Dual-CCD Vision System Simulating Human Eyes and Neck. The system uses two CCD cameras and four freedoms of rotate framework to simulate human eyes and neck. The research includes many theories and technologies such as mechanical designing, movement controlling, stereo vision measure and so on. So it supports experiment equipment and theoretic basis for our further research, 3D vision recognizing and specise object tracking.In chapter 1, the research’s significance of a dual-CCD simulating human eyes and neck vision system is introduced, as well as related research background at home and abroad. And related key technologies and main research content are summed up.In chapter 2, the whole scheme and theoretic frame of dual-CCD simulating human eyes and neck vision system are introduced. The dual-CCD simulating human eyes and neck framework is designed and its mathematic model is built, then the robot’s movement equations are deduced.In chapter 3, CCD imaging technology is introduced and the optical model for distance measuring using two CCD cameras is built. Using binocular intersection measuring method, its solid vision measuring equations are worked out and these error factors are analyzed.In chapter 4, the mathematic model of stepping motor are researched in general, subdivision control arithmetic of stepping motor are introduced. For clearance of gear side brought by decelerating framework of stepping motor being applied, its compensatingarithmetic while system resetting is put forward, and it can satisfy the system after successful experiments. The communicating ans controling software between PC and bottom CPU are also developed.In chapter 5, a new camera calibration technique for 3D machine vision with the radialalignment constrain (RAC) is described. The method separates camera parameters by using RAC, in order to provide the solution through a linear algorithm according to the reasonable order instead of traditional nonlinear optimization. After experiments, the camera calibration technique is proved to be a good method.In chapter 6, a 3D measuring model based on vision sensors is build. After existing methods of characteristic point matching being improved, the related 3D measuring theory and software flow designing are put forward.In chapter 7, the project’s main research content and solutions are summed up, and some prospect and explore for further discuss and research are described.更多还原