烟条外观质量视觉检测系统的研究

发布时间：2018-03-14 09:46

本文选题：烟条外观质量检测系统　切入点：平滑算法　出处：《南京理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：烟条外观质量检测系统是基于机器视觉技术,专门用于检测缺陷烟条的一种自动化设备,综合了包括光、电子、机械、计算机在内多种学科内容,它的使用不仅有助于提高生产线自动化程度、提高生产力和生产效率,而且还可以保证产品的品质,拥有很好的前景。本文首先研究了系统的工作流程和系统的硬件构成,分析了系统硬件结构的优缺点,搭建了适用于烟条外观检测的硬件系统。然后研究了图像的平滑算法和图像的锐化算法。针对图像的平滑算法,分析了均值平滑滤波器、双边滤波器、非局部均值滤波器的优缺点,改进了非局部均值滤波器,实现了保留边缘信息情况下的图像平滑操作;针对图像的锐化算法,分析了一阶、二阶微分算子锐化和LoG算子锐化,然后改进了 LoG算子,实现了锐化边缘自然过渡的图像锐化操作。其次,针对烟条图像的缺陷检测,研究了烟条图像定位、烟条拉线检测和烟条端面检测的检测算法。在烟条端面检测中,引入了人工神经网络的检测方法,利用样本训练的方法学习端面的特征,构建神经网络结构,从而实现了端面缺陷的检测,此方法具有更广泛的适用性。最后,本文基于以上研究内容,开发了与硬件设备相匹配的软件系统,分析了系统主界面、加载牌号模块、图像记录模块、工作记录模块、系统管理的工作流程和相应的组成结构,实现了具有用户友好性和美观性的软件界面。
[Abstract]:Appearance quality detection system based on machine vision technology, specifically for a defect detection of cigarette automation equipment, including integrated light, electronic, mechanical, computer, a variety of disciplines, it not only helps to improve the degree of automation of production line, improve productivity and efficiency, but also can guarantee the quality of the product and have a good prospect. This paper studies the workflow system and the hardware structure, analyzes the advantages and disadvantages of the hardware structure of the system, the hardware system is built for the appearance detection. Then the image sharpening algorithm and image smoothing algorithm for image smoothing algorithm, analysis the mean value smoothing bilateral filter, filter, advantages and disadvantages of non local mean filter, improved non local mean filter, the image smoothing and edge preserving information under the condition of operation As for image sharpening algorithm; the analysis of one order, two order differential operator and LoG operator, then the improved LoG operator, to achieve image sharpening operation sharp edge natural transition. Secondly, aiming at the defect detection of the cigarette image, image positioning, line detection and detection algorithm of cigarette cigarette end detection in the face detection. In the detection method of artificial neural network, using the method of training the learning characteristics of the face, to construct the neural network structure, so as to realize the detection of surface defects, this method has more extensive applicability. Finally, this paper based on the above, the software system is developed to match with the hardware device, analyzed the system main interface, grade loading module, image recording module, record module, system management work process and the corresponding structure has been realized The user friendliness and beauty of the software interface.

【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】