【作者】 王磊；

【导师】 沈昱明；

【作者基本信息】 上海理工大学 ， 检测技术与自动化装置， 2010， 硕士

【摘要】 铜条工业生产中由于各种原因导致铜条表面产生缺陷，通常情况下工厂采用人工检测的方法，不仅效率低，而且准确率不是很高。针对上述问题，本文研究了已有的机器视觉技术，提出了一种铜条缺陷在线高速检测装置，该装置对复杂环境的适应性较好，同时具有实时性和准确性等特点。本文中的铜条缺陷在线高速检测装置是由照明系统、工业相机、工控机、图像处理软件等组成。照明系统采用同轴光源照明技术，获取到质量较高的铜条图像，采集到图片后，将图片传入工控机进行实时处理，最后输出结果和对缺陷进行记录并保存。本装置采用LabVIEW并且配合NI公司的机器视觉模块进行编程，其算法流程是：首先对图像进行预处理，采用高斯滤波对图像进行去噪处理，利用数学形态学中的闭运算对不属于缺陷的小瑕疵进行去除，然后采用最大类间方差法对图像进行阈值分割，再采用Sobel算子对铜条图像进行边缘检测得到缺陷以及铜条的边界，利用铜条的形态，在铜条范围中设定检测线来识别缺陷，根据缺陷的有无显示出Pass/Fail，并且将缺陷出现的系统时间以及与开始位置的距离显示并记录保存。通过大量实验证明，本装置具有准确率高、实时性好、自动化程度高等优点。更多还原

【Abstract】 In copper bars Industrial production, various reasons make copper strips surfacehave defects, usually the factory use manual inspection method, it is inefficient and notvery high accuracy. Response to these questions, we studied the existing machinevision technologies, and gave a applications in online detection of defects onhigh-speed bar copper, the device is better on the complex environment, also istimeliness, accuracy and so on.This application in online detection of defects on high-speed bar copper includeslighting systems, industrial cameras, industrial computer, image processing softwareand other components. Lighting system uses coaxial illumination technology to gethigher quality copper strips images, then transfer the collected images to IPC and doreal-time processing, output results, record and save defects at last. This device usedLabVIEW and NI’s Machine Vision modules to program, the algorithm process is:firstly, image preprocessing, using Gaussian filter for image denoising, using theclosing operation of mathematical morphology to remove small defects that do notbelong to defects, and then using the maximum between-class variance thresholdingthe image, then in edge detection, using Sobel operator to select the edge of copperstrips and defects in the images, Setting range test line to identify the defects based onthe form of copper strips, showing Pass/Fail according to presence or absence ofdefects, and displaying, recording and saving the system time when defects appear andthe distance to the beginning. A large amount of experimental results implied that thedevice is high accuracy, real time, automatization.更多还原