【作者】 王建华；

【导师】 刘允才；

【作者基本信息】 上海交通大学 ， 模式识别与智能系统， 2007， 博士

【摘要】 手术导航仪能够使外科医生实时浏览病灶部位的三维解剖结构以及手术器械所处的位置和方向,规划手术器械进入的路径,减小手术创伤,提高手术精度。手术导航的出现引发了手术方式的革新,因此成为外科领域的一个应用和研究热点。根据国内对手术导航应用和研究的需要,本文在研究手术导航仪所涉及关键技术的基础上,构建了一个数字式红外手术导航仪样机。为了对手术导航仪有一个全面深入的了解,把握研究的方向,本文首先对手术导航的概念、优点和发展进行了概述,说明研究和开发手术导航仪的必要性;根据国内外技术发展差距的对比,以及国内医疗界对手术导航仪的需求,说明研究和开发具有完全自主知识产权手术导航仪的紧迫性;在这样的背景下,根据目前具备的研究基础、条件和优势,提出了本文的主要研究内容和预期目标。本文研究的主要内容如下:1.提出了一种基于平面单应矩阵中心线的摄像机标定方法:目前最具发展前景的手术导航系统是光学导航系统,其原理是双目立体视觉。应用双目立体视觉的一个关键步骤是摄像机标定,本文在对现有摄像机标定方法进行综述的基础上,针对这些方法在应用于手术导航系统时遇到的实际问题,提出一种基于平面单应矩阵中心线的摄像机标定方法,对平面单应矩阵中心线的性质进行了推导和实验验证,对标定方法的实施步骤进行了说明,并利用平面单应矩阵中心线的性质讨论了选择有效标定图像以提高摄像机标定精度的问题。2.提出了一种摄像机镜头畸变模型:摄像机镜头畸变是影响摄像机标定和立体视觉计算精度的一个重要因素,而畸变模型的选择又是进行畸变校正的一个关键问题。在采用传统镜头畸变模型进行图像校正时,虽然模型中参数较多,但相互间不独立,由于畸变参数间的耦合,会使优化过程收敛到局部最小值,影响手术导航仪的精度。针对这一不足,在分析镜头畸变产生原因和各畸变分量之间相互关系的基础上,本文提出一种新的镜头畸变模型,进行了理论分析和实验验证。3.提出了一种摄像机镜头畸变参数的估计方法:根据手术导航仪测量体积和测量距离的要求,摄像机应采用短焦距广角镜头,因此畸变不容忽略。现有摄像机标定方法缺乏对镜头畸变的初值估计,在进行迭代寻优的初始化阶段假设镜头畸变系数为零,这样的假设对于畸变较大的镜头会导致优化过程收敛到局部最小值,影响手术导航仪的精度。针对这一不足,本文提出一种镜头畸变初值的估计方法,对方法的原理进行了推导,并通过模拟实验和真实图像实验对方法进行了验证。4.提出了一种非平行轴双目立体视觉的直接计算方法:手术导航仪中双目立体视觉系统的两个摄像机通常采用非平行轴配置,它们的内部参数一般也不相同,因而不能直接应用标准平行轴双目立体视觉的计算方法。现有方法分为两类,一类通过迭代求解两条投影线在空间的交点,其缺点是需要初值估计,并且会收敛到局部最小值;另一类使左右图像分别旋转一定角度,将它们校准到满足平行轴双目立体视觉系统的要求,然后调用标准算法,其缺点是需要进行图像的几何变换,增加运算量,影响手术导航仪的实时性,并且这类方法没有考虑噪声干扰,其重建精度受噪声的影响较大。针对这两种方法的不足,本文提出一种非平行轴双目立体视觉的直接计算方法,进行了理论推导和实验验证。在上述关键技术研究的基础上,构建了一个数字式红外手术导航仪样机。这部分内容是本文研究的目的,也是上述研究结果的应用和检验。根据目标技术参数要求,设计了光学系统,构建了硬件平台,编写了相关软件,进行了标定、测试、精度评估和模拟导航实验。结果表明,所构建的数字式红外手术导航仪样机达到了预期的效果,说明了上述研究结果的正确性和实用性。更多还原

【Abstract】 Advanced navigators for surgery enable a surgeon to look through the anatomy of the patient, to see the instruement guided into the anatomy, and to plan the path of the instruement, so that the invasion is reduced and the accuracy of operation is improved. Consequently, surgical navigation has become a hot spot of application and research in medical field. Drived by the domestic demand, the thesis is intended to research into key techniques involved in navigators for surgery.In order to get sufficient information and find correct orientation of the research, a review on the technology is made. After introduction of concepts, advantages and development related to navigator for surgery, the necessity to research and develop surgical navigator is clear; through comparison of different development level in the world and in our country, the urgency of developping domestic surgical navigator with completely independent intellectual property rights is addressed. According to our condition and advantages, the main contents and objective of the research are determined. The major results of the thesis are as follows.1. A method to calibrate camera using the Center Line of planar homography matrix is proposed. The most promising surgical navigators so far use optical localizer whose principle is stereo, in which camera calibration is a key step. In practice of calibrating stereo system, some problems with the existing camera calibration methods are encountered. To solve these problems, a method using the Center Line of planar homography matrix is proposed. Some useful properties of the Center Line are presented and derived, based on which the calibration method is described. The properties of the Center Line are also used to discuss the problem on taking effective calibration images to obtain accurate calibration result and improve performance of optical navigator for surgery. Experiments validate the method. 2. A model of lens distortion is proposed. Lens distortion is one of the main factors affecting accuracy of camera calibration and stereo computation, and distortion model is an important issue in distortion correction. While calibrating cameras used in surgical navigator, it is noticed that the parameters in the traditional distortion model are not independent from one another. Based on analysis of lens distortion sources and the relation among different distortion components, a new model of lens distortion is proposed with theoretical derivation and experimental validation.3. A method to estimate initial value of lens distortion is proposed. According to the requirement for the measuring volume and the distance from the stereo rig to the instruement, wide-angle lenses with short focal length are used. Consequently, lens distortion can not be neglected. At the initializing stage of existing methods, the distortion center is assumed to be the image center, while all disdortion coefficients are assumes to be zeros. When a lens has visible distortion, the iterative process of optimization may converge to local minima. So it is desirable to estimate distortion parameters more accurately. The method is derived theoretically and validated experimentally.4. A direct method to reconstruct the scene point from nonparallel stereo is proposed. The stereo rigs used in surgical navigators are usually configured with their camera coordinate systems being not parallel to each other. In this case, the conventional reconstruction method based on the standard parallel stereo configuration cannot be used directly. Existing methods fall into two classes. One back-projects the projecting rays into space and finds their intersection. Problems with the method are initial value estimation and local minima. The other rectifies arbitrary stereo image pairs to make them satisfy standard parallel stereo and then employs the conventional algorithm. The main disadvantages of the later method are time consuming and not taking noise into account, which will affect real-time and accuracy of instruement tracking. In the thesis, a direct method to reconstruct a scene point from nonparallel stereo is proposed, bypassing the process of rectifying the images or iteratively optimizing. Theoretical derivation and experiment results from both simulated data and real images validate the method. Based on the above results of research, a prototype of digital infrared navigator for surgery is developped, which includes optical system designing, the hardware constructing, program coding, system calibrating, testing, accuracy evaluating and navigation simulating. The experiment results show that the main performance of the prototype has satisfied the aimed requirement, which validates the methods proposed above.更多还原