【作者】 李九灵；

【导师】 高全杰；

【作者基本信息】 武汉科技大学 ， 机械设计及理论， 2013， 博士

【摘要】 质量需求是现代企业的核心需求，产品质量的好坏直接关系着企业的市场竞争力和企业自身的存续。因此，将质量放在首位是目前所有企业的共同选择。质量控制与质量管理的方法与手段多种多样，传统检测与测量往往依赖于人工完成。本文着眼于流水线上、不间断运动的连续物体视觉可识别量的在线检测。利用机器视觉替代人工检测，运用计算机及软件替代检测工人，通过可重构的视觉检测方法实现在不同被检对象上的快速反应与装备实现。本文主要研究了可动态增减检测子系统的分布式网络拓扑，支撑采集控制硬件层、图像处理识别层、界面显示层、质量等级评价等多层面的视觉重构理论，建立了基于软件芯片的视觉重构体系，对重构流程、重构方案进行了设计，采用规范化、标准化的方式设计了各个层次间交互接口。针对市场上不同厂商的数字图像获取设备，本文均根据自身产品而定制、通用性差的缺点进行了相关研究。首先，分析了不同数字图像获取标准的优劣，建立了异构硬件环境下的图像获取通用模型；接着，创建了通用图像获取接口，设计并完善了SDK，包括初始化函数、设置函数、获取函数、图像处理回调函数、存储函数、辅助函数在内的六大类函数的定义；最后，对函数接口及其组态设计进行了研究。在图像获取的基础上，对图像处理环节进行了探讨。首先，分析了视觉检测的基本需求，梳理了视觉检测的基本流程；接着，根据视觉检测流程的主要需求，设计了机器视觉在线检测算法库，将常用的图像预处理、图像分割、图像复原、特征提取等算法进行了标准化、参数化设计，着重突出优化算法代码、执行效率、鲁棒性等问题；最后，建立并完善了一套基于配置信息的视觉检测运行环境，可实现检测算子的搜索、调用、重载等功能，有利于提高检测系统的柔性，加速针对不同检测对象的设备部署。为实现图像特征的提取与识别，本文首先研究了常用特征描述方法，分析了其优缺点，并指出单一特征或过小规模特征集在可重构视觉检测方法中的局限性。其后，针对这一核心问题展开，设计了一套基于统计与时频联合的特征提取方法，建立了通用性机器视觉在线检测特征集，并利用该特征集对三种产品的图像图形学特征进行了描述。最后，尝试设计了基于遗传算法的特征解耦方法，并就其中关键技术进行了研究。最后，利用可重构的机器视觉方法对粘扣带、导爆管、网孔织物外观质量视觉检测进行了研究，开发了具有自主知识产权的检测系统，验证了可重构的产品视觉检测方法的可行性和有效性。更多还原

【Abstract】 Quality requirements is the core of modern enterprise, the product quality has a directrelationship with the market competitiveness of enterprises and the enterprises themselvessubsisting. Sothe enterprises give top priority to quality is a common choice. Now all thecompanies’ quality control and quality management methods are diverse. And traditionaldetection and measurement is often dependent on artificial completed.This article focuses on assembly line continuous, uninterrupted movement object vision torecognize the amount of online detection. The alternative to manual inspection is machine vision,using computers and software to replace workers detection. And by using the reconfigurablevisual inspection of the implementation on different objects have been rapid response andequipment implementation.Mainly studies distributed network topology dynamically increase or decrease the detectionsubsystem, support the collection and control hardware, image processing and recognition layer,interface display layer, quality grade evaluation multifaceted visual reconstructiontheory,established a software which is based on chipvisual reconstruction system, refactoringprocess, the restructuring plan design, using a standardized design interactive interface betweenthe various levels in a standardized way.According to different vendors of digital image acquisition device in the market, all of themaccording to their own products and custom, the shortcomings of poor universality has carried onthe related research. First analysis of the pros and cons of different standard digital imageacquisition, the establishment of a generic model of heterogeneous hardware environment imageacquisition; Next, create a common image acquisition interface and well-designed SDK,including the initialization function set the function, gain the function, image processing, thecallback function, memory function and auxiliary function, the definition of the six categories offunction. Then the function interface and configuration design were studied.On the basis of image acquisition, image processing steps are discussed in this paper. Firstlyanalyzes the basic needs of visual inspection, combing the basic process of visual inspection.Then according to the visual inspection process of the main requirements, design the onlinedetection machine vision detection algorithms library,the commonly used image pretreatment,image restoration, image segmentation, feature extraction algorithms such as standardization,parametric design, especially emphasizes optimization algorithm code, execution efficiency,robustness and so on. Finally, establish and improve a set of visual detection based onconfiguration information for the runtime environment, which can realize detection operatorsearch, call, overloading, and other functions, helping improve the detection system flexible,accelerated deployment object for different detection equipment.To achieve image feature extraction and recognition, the first research to describe thecommon characteristics method, analyzes its advantages and disadvantages, points out that asingle feature or a small feature set limitations in reconfigurable visual inspectionmethod.Subsequently,he core issue is that to start design a feature extraction method based onstatistics and frequency joint,establishment of a general-purpose machine vision detection featureset of online detection and image graphics using the feature set of the three products the characteristics are described. Finally, try to design a decoupling method which ids based on thecharacteristics of genetic algorithm is presented and its key technology was studied.Finally, the use of reconfigurable machine vision method for fastening belt, detonating tube,the appearance quality of the mesh fabric visual inspection are studied, developed withindependent intellectual property rights of inspection system, verify the reconfigurable productvisual inspection method is feasible and effective.更多还原