【作者】 王亮；

【导师】 付先平；

【作者基本信息】 大连海事大学 ， 计算机科学与技术， 2017， 硕士

【摘要】 在工业检测流水中应用机器视觉系统,对样本进行检测,以提高制造业、印刷业等工厂的机械自动化能力,提升工业生产效率,已经成为生产中的迫切需求。本文提出了一种高速特征检测算法,解决了图像检测算法应用于工厂环境中,速度慢与稳定性差的问题,完全满足工业检测实时性与准确性要求。本文主要工作如下:(1)针对流水线待检样本的实际运行环境,以及机器视觉系统的光源特性,提出了一种基于小波变换的图像匀光方法,消除了的机械视觉系统中点光源对于输入图像造成的影响,获得了稳定的输入图像。(2)提出了一种基于迭代逼近的模板配准方法,提升了模板定位的速度,采用改进的LBP方法提取图像整体纹理特征,保证了定位的准确度。(3)采用基于KNN的ORB特征匹配方法满足了实时性要求,提出了一种基于块划分的相似度计算方法,保证了相似度计算的稳定性。另外,本文设计开发了工业样本自动检测系统对本文算法进行验证。此系统应用于印刷业配页机械进行样本页检测,经验证本文提出的算法对于样本页检测最大耗时低于250ms,检测准确率在99.5%左右,完全可以满足工业检测的实时性与准确性要求。更多还原

【Abstract】 The application of machine vision system in industrial inspection has become an urgent need in production,for it can improve the mechanical automation capabilities and productivity in both the manufacturing and printing industries.A high-speed feature detection method has been proposed in this paper to solve slow speed and poor stability problems of image detection method which is applied in a factory environment.It meets requirements of the real-time and accuracy industrial inspection.The main work of this paper includes:(1)According to the running environment of the production line and the features of machine vision system,a uniform light method based on wavelet is proposed in this paper to eliminate the effects of light source in machine vision system so that it can obtain a stable input image.(2)An iterative template registration method was proposed to ensure the speed of positioning the image,and the improved LBP method has been chosen to extract the whole contour features of the image to ensure the accuracy of positioning the image.(3)An ORB feature detection method and an match method based on KNN to meet the real-time requirement are adopted in this paper guaranteeing to calculate the similarity between the sample image and the template image with a similarity method based on blocks.In addition,an automatic detection system for industrial samples is designed based on the algorithm proposed in this paper.The system is used in collating machinery for printing industry.It has been tested that the maximum time consumed,according to the algorithm proposed in this paper for sample detection,is less than 250ms,and the accuracy rate is about 99.5%.Therefore,it can completely meet the real-time and accuracy requirements of industrial inspection.更多还原