【作者】 薛云洋；

【导师】 俞庠； 刘秀兰；

【作者基本信息】 北京工业大学 ， 电路与系统， 2001， 硕士

【摘要】 荧光磁粉探伤是一种常用的无损检测方法，是控制产品质量的重要手段，用于检测钢制零件的表面及近表面的裂纹缺陷。铁路轮对荧光磁粉探伤半自动设备已经大量使用，但由于受现场条件的限制，检查由荧光粉显示的裂纹仍旧依靠人眼识别，因而随机性大，劳动强度高而且不易发现车轮隐蔽缺陷，极易造成漏检误检。因此，开发一套铁路轮对全自动荧光磁粉探伤系统，不仅能极大地提高车辆车轮的探伤效率，减轻探伤人员的劳动强度，同时能有效地避免人为因素对探伤质量的影响。 系统由硬件和软件两个部分组成。本文主要讲述软件部分的实现。该软件一方面控制执行机构的工作，另一方面进行图象数据的采集处理。它主要包括：图象采集、图象分割、图象的数学形态学预处理、区域标号以及特征提取和裂纹的识别。 在本套系统的软件部分，最关键的是图象处理的算法。根据系统采集的图象特点，改进了以前提出的一种快速自适应动态阈值分割算法，使之既能很好解决亮区问题又能不丢失信息。为了减轻区域标号的负担，提出了基于联结数的多次收缩法。用一种基于扫描法的快速区域标号算法，使系统能够实时处理。最后提取区域特征，根据经验值识别出裂纹。更多还原

【Abstract】 Fluorescent magnetic particle detection is a NDT method that often is used for testing trivial flaws on the surface of some steel workpiece and for controlling qualities of product. Semiautomatic device for fluorescent magnetic particle detection has been largely used. But the flaws are recognised by manual work. There axe a mass of factors which affect the testing and the work condition is atrocious. So it is important to develop a system of fluorescent magnetic particle auto-detection devices. This paper mainly presents a auto-detection flaw testing system.The system is made up of hardware and software. This thesis mainly describe how to realize software .Software section includes image sampling, image segmentation, image mathematic morphologic pretreatment, threshold division, label, eigenvalue extraction and flaw recognition.The algorithm of image processing is the core of the software. According to the characteristic of the sampled image.I have improved the algorithm of a fast adaptive dynamic threshold algorithm which was used by the former version. So it can solve light region problem and can avoid losing information. In order to relief the burden of Label,I provide a multi-shrinking method that based on the number of connection. By using a fast label algorithm which is based on the scanning method,we can make system process data in real time. At last, extract regional eigenvalues and according to these eigenvalues we can identify the flaws.更多还原