静态目标异常检测算法研究.docVIP

PAGE PAGE 19 静态目标异常检测算法研究 摘要：当今数字图像处理技术可以说是在各个领域都发展迅速，特别是基于数字信号处理的图像缺陷异常检测有了广泛的应用。而随着数字图像处理技术的发展，以及人对产品质量日益提高的要求，传统人工检测因为各种原因无法得到较完美的检测结果，所以也使得人们越来越重视机器视觉检测处理技术的发展。现代工业生产大批量的物件时，基本上都会因为有各种因素的影响而存在着残次品。传统的人工抽样检测法很明显是无法达到产品零缺陷的，而利用图像处理技术的机器视觉对异常缺陷检测相对于人工检测有成本低、损伤少、快速、准确的优势。本文主要探讨基于数字图象处理的静态图像的异常缺陷检测算法。 本次毕设课题着重对异常缺陷检测系统对采集到的静态图像进行灰度化处理、去噪、边缘检测、及其二值化等处理，并且根据产品边界合拢的特性，得出了一种新颖的二值图像去噪的算法，该算法可以有效的去除二值图像中的斑点噪声，从而取得相对较为清楚的产品图像边界二值图像。最后经过对存在缺陷的产品的特性分析，研究检测的方法，从而实现对产品工件进行检测目的。 针对检测异常缺陷图像，文中选用了三种检测的算法：基于信息熵的异常检测算法、基于形态学图像处理的缺陷检测算法、和基于灰度信息的模版匹配的异常缺陷检测算法。 最后，在VC++软件环境下实现测试系统的一些功能，包括对产品工件图像采集、处理和分析等。 关键词：数字图像处理；异常缺陷检测；二值化；灰度化；边缘检测；VC++。 The algorithm research of static target anomaly detection Abstract: Todays digital image processing technology can be said to be in every field has developed rapidly, especially image defects in anomaly detection based on digital signal processing has a wide range of applications. But with the development of digital image processing technology, as well as the requirements of increasingly improve the quality of product, the traditional manual detection because of various reasons could not get a perfect test results, so also makes people pay more and more attention to the development of machine vision detection processing technology. Industry, automated production of large quantities of items, usually because of some factors and there are more or less incomplete. Artificial sampling test is obviously unable to achieve zero defect products, and the use of image processing technology of machine vision detection of abnormal defects had low cost compared with manual inspection, the advantages of less injury, rapid and accurate. This paper mainly discusses the static image based on digital image processing of the abnormal defects detection algorithm. The project topic of abnormal defect detection system for static images collected gray processing, denoising, edge detection and binarization processing, and according to the characteristics of the border