图像数据统计过程控制方法研究

发布时间：2019-02-23 13:36

【摘要】：随着机器视觉系统在工业中的应用越来越广泛,产品信息越来越多地以图像数据的形式呈现。研究如何利用图像数据对生产过程进行监控已成为统计过程控制面临的新课题。本文以具有一致性或存在特定模式的产品图像为研究对象,针对偏移区域形状未知、偏移数目未知、图像中相邻像素的灰度值存在高度相关关系等情况下的图像数据监控方法分别进行了研究。首先,针对偏移区域形状未知的情况,提出了基于EWMA和区域生长的图像数据监控方法,给出了实施该方法的详细步骤。利用仿真试验考察了该方法在检测偏移区域形状规则、不规则时的效果,并与Megahed等提出的基于最大广义似然比的图像数据监控方法进行了对比。结果表明,所提方法不仅能快速检测出偏移,还能更准确地估计出偏移发生的区域。接着,针对偏移数目未知的情况,提出了基于广义似然比之和的图像数据监控方法,介绍了实施该方法的详细步骤。利用仿真试验考察了该方法在检测单个、多个偏移时的效果,并在检测单个偏移时,与Megahed等提出的基于最大广义似然比的图像数据监控方法进行了对比。结果表明,所提方法检测效果优于最大广义似然比的图像数据监控方法。然后,针对图像中相邻像素的灰度值存在高度相关关系的情况,提出了基于多元广义似然比的图像数据监控方法,给出了实施该方法的详细步骤。利用仿真试验考察了该方法在检测单个、多个偏移时的效果,并在检测单个偏移时,与Megahed等提出的基于最大广义似然比的图像数据监控方法进行了对比。结果表明,所提方法检测效果优于最大广义似然比的图像数据监控方法。最后,将基于多元广义似然比的图像数据监控方法应用于实际案例中,应用结果表明,该方法不仅能快速检测出偏移,还能提供偏移发生的时间、区域等信息,帮助员工对过程进行诊断,查找异常原因,尽早恢复过程。
[Abstract]:With the increasing application of machine vision system in industry, product information is presented in the form of image data more and more. How to use image data to monitor the production process has become a new subject of statistical process control. In this paper, the product images with consistent or specific patterns are taken as the research object. The shape of the migration region is unknown, and the number of the offsets is unknown. The methods of image data monitoring are studied under the condition that the gray values of adjacent pixels are highly correlated. Firstly, an image data monitoring method based on EWMA and region growth is proposed for the unknown shape of the offset region, and the detailed steps to implement the method are given. The effectiveness of this method in detecting irregular shape and shape of offset region is investigated by simulation experiments, and compared with the image data monitoring method based on maximum generalized likelihood ratio proposed by Megahed et al. The results show that the proposed method can not only detect the migration quickly, but also estimate the migration area more accurately. Then, an image data monitoring method based on the sum of generalized likelihood ratio is proposed, and the detailed steps to implement the method are introduced. The effectiveness of the method in detecting single or multiple offsets is investigated by simulation, and compared with the image data monitoring method based on the maximum generalized likelihood ratio proposed by Megahed et al. The results show that the proposed method is superior to the maximum generalized likelihood ratio method for image data monitoring. Then, a method of image data monitoring based on multivariate generalized likelihood ratio is proposed, and the detailed steps to implement the method are given in view of the high correlation between the gray values of adjacent pixels in the image. The effectiveness of the method in detecting single or multiple offsets is investigated by simulation, and compared with the image data monitoring method based on the maximum generalized likelihood ratio proposed by Megahed et al. The results show that the proposed method is superior to the maximum generalized likelihood ratio method for image data monitoring. Finally, the image data monitoring method based on multivariate generalized likelihood ratio is applied to practical cases. The application results show that the method can not only detect the migration quickly, but also provide information about the time and region of migration. Help employees diagnose the process, find the cause of the exception, and resume the process as soon as possible.
【学位授予单位】：天津大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP13

【相似文献】