基于不对称双目机器视觉的嵌入式尺寸测量系统

发布时间：2018-06-01 14:39

本文选题：不对称双目机器视觉 + 嵌入式测量装置　；参考：《浙江大学》2012年硕士论文

【摘要】：传统的基于单目视觉的尺寸测量方法使用单幅图像反映待测物尺寸的全貌,当需要对某些大尺寸复杂零件的整体和局部同时进行测量,或者待测物几何尺寸会发生较大改变时,某些几何尺寸相对较小的部位由于占有像素数较少而无法实现较高精度的测量。而且,现有的视觉测量装置往往是基于PC机的,其经济性与灵活性不能满足工业测控的需求,因此对嵌入式视觉测量方法及装置的研究具有重要的现实意义。本文结合双目机器视觉和单目机器视觉的特点,设计并实现了一种基于不对称双目机器视觉的嵌入式尺寸测量系统,完成对待测物大尺寸和小尺寸部件的同时测量。整个测量过程完全在基于ARM的嵌入式软硬件平台上实现,通过客户端自带的浏览器,可以实现远程操作和监控。主要研究内容如下： 1.提出了一种基于不对称双目立体视觉技术的测量方法和几何尺寸测量步骤,并设计了基于ARM的嵌入式软硬件实现方案,对其中的重要设备选型原则作了说明。 2.在研究图像边缘提取算法时,提出了一种结合Sobel算法和Zernike矩算法的亚像素边缘提取算法,以及一种Sobel-Zernike-Forstner亚像素角点提取算法。在此基础上,运用基于平面方格点的摄像机标定方法,实现了对系统的标定。 3.利用一种基于极线约束、特征匹配与区域匹配相结合的立体匹配方法,结合本双目视觉不对称的实际,匹配若干图像边缘点对,得到其对应空间点的空间位置,结合本文提出的平面确定方法,得到测件所在平面的空间位置；一旦得到了测量平面的位置,只需合适的单摄像机,就可完成需要的平面尺寸参数测量。 4.完成了嵌入式样机的制作。硬件部分,完成了系统的选型与构建；软件部分,通过EVC开发平台将对摄像机的操作以及尺寸测量相关的各种算法封装成COM组件,采用"ASP+COM"的结构技术,实现了远程操作和监控。通过对圆参数测量的验证实验表明,本嵌入式测量系统能够实现预期的功能。
[Abstract]:The traditional dimension measurement method based on monocular vision uses a single image to reflect the whole picture of the object under test. When it is necessary to measure the whole and part of some large and complex parts at the same time, or the geometric size of the object to be measured will change greatly. Because of the small number of pixels, some parts with relatively small geometric size can not achieve high precision measurement. Moreover, the existing visual measurement devices are often based on PC, their economy and flexibility can not meet the needs of industrial measurement and control, so the research of embedded vision measurement methods and devices has important practical significance. Based on the characteristics of binocular machine vision and monocular machine vision, an embedded size measurement system based on asymmetric binocular machine vision is designed and implemented in this paper. The whole measurement process is completely implemented on the embedded software and hardware platform based on ARM, and remote operation and monitoring can be realized through the browser which is included in the client. The main contents of the study are as follows: 1. This paper presents a method of measurement and geometric dimension measurement based on asymmetric binocular stereo vision, designs the embedded hardware and software implementation scheme based on ARM, and explains the important equipment selection principle. 2. A sub-pixel edge extraction algorithm combining Sobel algorithm with Zernike moment algorithm and a Sobel-Zernike-Forstner sub-pixel corner extraction algorithm are proposed in this paper. On this basis, the calibration of the system is realized by using the camera calibration method based on plane grid points. 3. A stereo matching method based on polar line constraint, feature matching and region matching is used to match a number of edge point pairs of images, and the spatial position of the corresponding spatial points is obtained by combining the reality of this binocular vision asymmetry. Combined with the plane determination method proposed in this paper, the spatial position of the plane in which the measuring piece is located is obtained. Once the position of the measuring plane is obtained, only a suitable single camera is required to complete the measurement of the required plane dimension parameters. 4. The embedded prototype is made. In the hardware part, the selection and construction of the system are completed. In the software part, the operation of the camera and the algorithms related to the dimension measurement are encapsulated into COM components through the EVC development platform, and the structure technology of "ASP COM" is adopted. The remote operation and monitoring are realized. The experimental results show that the embedded measurement system can achieve the desired function.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP391.41;TP368.1

【参考文献】