【作者】 耿世勇；

【导师】 张五一；

【作者基本信息】 中原工学院 ， 控制理论与控制工程， 2011， 硕士

【摘要】 本课题的研究目的是实现基于图像处理卡的帘子布疵点检测硬件平台,并探索快速、准确的在线帘子布疵点检测算法。帘子布具有幅宽大,运动速度快且变动剧烈的特点,因此对其进行清晰、稳定的图像采集难度较大,这也是该类机器视觉系统的一个关键问题。该系统的硬件设计分为三个部分:一、光学成像系统,包括线阵相机,镜头及光源。二、运动同步系统,包括编码器,分频器。该子系统要求能准确地检测帘子布的运动速度,精确控制线阵相机的同步采集。三,图像获取系统,包括图像采集卡及计算机。本文根据系统的设计要求,对该系统进行了硬件选型及参数计算,对运动同步系统进行了软硬件设计,并在VC平台下,利用MIL库编写了图像采集程序。该图像采集系统已在现场安装并完成调试。实验结果表明,该系统能够在工业现场进行高质量的,稳定的帘子布图像采集。在快速判断帘子布有无疵点方面,本文比较了基于灰度直方图和基于灰度双阈值的判断方法。其中基于直方图检测的方法是比较基本的方法,由于其本身的缺陷,在误检率上还不太满足系统的要求;基于灰度双阈值的方法是根据帘子布的纹理结构特性,选取大量正常图像的最大灰度值和最小灰度值做为此方法的双阈值,通过区域对比和纵向结合的方法判断图像的灰度是否有异常。在最大灰度值和最小灰度值两个阈值的约束下,能准确快速检测出疵点,所以更加适合快速判断有无疵点的检测。更多还原

【Abstract】 This study aims to build an online cord fabric’s defect detection hardware platform based on image processing board, and explore real time defect detection methods with high accuracy. The challenge of acquiring high quality images in this kind of system comes from the large width of the cord fabric and the movement velocity which is high and varies in a large extent. The hardware platform is subdivided into three parts: first, the optical imaging subsystem including the line camera, lens and lamp-house; second, motion synchronization subsystem composed of an encoder and the frequency division controller, in which accurate speed measurement of the cord fabric is implemented in order to synchronize the acquisition frequency of the line camera; third, the image acquisition subsystem composed of image processing board and PC. In this paper, hardware is elaborately selected and their parameters are calculated subsequently. Implementation of the motion synchronization subsystem, both hardware and software, is conducted. By aid of the MIL library, an image acquisition program is realized in Visual Studio. The whole image acquisition system has been installed and adjusted in the field. Experiments approve that our system can obtain images with high quality stably in practice.For quickly judging whether the cord fabric has defects, methods based on histogram and two thresholds in grayscale are compared. Of which based on the histogram detection method is a more basic way, because of its own shortcomings, in the false detection rate is not yet meet the requirements of the system. According to the structural characteristics of texture cord fabric, the method based on grayscale double thresholds selects a large number of maximum grayscale values and minimum grayscale values in normal image as this method’s double thresholds, and then comparing with regional and longitudinal, to judge whether there is abnormity in grayscale. Under the constraints between maximum values and minimum values, it can accurately detect defects quickly, so it is more suitable for quick judgment.更多还原