Soldering defect detection in automatic optical inspection

This paper proposes an integrated detection framework of solder joint defects in the context of Automatic Optical Inspection (AOI) of Printed Circuit Boards (PCBs). Both localization and classifications tasks were considered. For the localization part, in contrast to the existing methods that are highly specified for particular PCBs, we used a generic deep learning method which can be easily ported to different configurations of PCBs and soldering technologies and also gives real-time speed and high accuracy. For the classification part, an active learning method was proposed to reduce the labeling workload when a large labeled training database is not easily available because it requires domain-specified knowledge. The experiments show that the localization method is fast and accurate. In addition, high accuracy with only minimal user input was achieved in the classification framework on two different datasets. The results also demonstrated that our method outperforms three other active learning benchmarks.     

自动光学检测在中国的应用现状和发展

本文首先概括了自动光学检测(AOI)的测试优点,然后结合具体产品讨论了AOI(包括AXI)的最新技术和发展,以及图像处理算法应用概况。接着对国外几家著名公司AOI产品作了简单介绍,并对国内目前AOI产品的研制情况进行了综述。最后介绍了几种基于AOI的电子组装综合检测技术。     

一种基于 Parzen 窗估计的鲁棒 ELM 烧结温度检测方法

在回转窑燃煤火焰视频模糊且干扰较大的情况下, 基于火焰辐射能量和燃烧稳定程度提取多帧煤粉燃烧图像的统计特征进行烧结温度判断. 为克服工业现场特征数据中的粗差干扰,将极限学习机(Extreme learning machine, ELM)与稳健估计理论相结合, 用训练误差分布的Parzen窗非参数估计构造ELM权矩阵,对其输出层权值进行稳健最小二乘估计. 基于上述火焰视频的统计特征,用该改进的鲁棒极限学习机(Robust-ELM)检测烧结带温度.实验结果表明, 在视频图像模糊、不能用常规静态图像处理方法软测量烧结带温度时,本文方法可快速有效地检测窑内烧结温度, 且检测系统不易受现场干扰,稳定性强.     

基于卷积神经网络的焊缝表面缺陷检测方法

针对工业激光焊接中,采用传统方法进行焊缝质量检测效率低下的问题,提出了一种基于卷积神经网络的工业钢板表面焊缝缺陷检测方法;首先基于卷积神经网络,搭建了一个多分类模型框架,并分析了各层中所用到的函数及相关参数;然后基于工业数控机床和工业相机进行了焊缝数据采集,并对这些数据进行了分类、增强、扩增等前期预处理;最后基于数控机器轴,采用滑动窗口检测的形式采集实际待测图像,并通过实验对比了传统的机器学习算法在该类图像数据中的性能评估;经实验证实,通过卷积神经网络训练得到的多分类模型,焊缝缺陷检测精度能达到97%以上,且每张待测图像的测试时间均在300 ms左右,远超机器学习算法,在准确性和实时性上均能达到实际工业要求。     

Cell morphology classification and clutter mitigation in phase-contrast microscopy images using machine learning

We propose using machine learning techniques to analyze the shape of living cells in phase-contrast microscopy images. Large scale studies of cell shape are needed to understand the response of cells to their environment. Manual analysis of thousands of microscopy images, however, is time-consuming and error-prone and necessitates automated tools. We show how a combination of shape-based and appearance-based features of fibroblast cells can be used to classify their morphological state, using the Adaboost algorithm. The classification accuracy of our method approaches the agreement between two expert observers. We also address the important issue of clutter mitigation by developing a machine learning approach to distinguish between clutter and cells in time-lapse microscopy image sequences.     

不变量的程序潜在错误预测

随着软件系统变得越来越复杂和庞大,软件中的安全缺陷也急剧增加,系统中的隐含错误也在逐渐增多.提出一种基于不变量的程序潜在错误预测方法,首先采用支持向量机对程序属性所产生的非函数依赖程序不变量进行学习并产生机器学习模式,然后运用该机器学习模式对需预测的程序进行属性分类,并揭示出代码可能存在的潜在错误,最后通过实验验证该方法是有效的.     

基于极限学习机与规则推理的NPC三电平逆变器二级故障诊断方法

针对NPC三电平逆变器故障诊断问题,提出一种基于极限学习机与规则推理的二级故障诊断方法。分析了依据输出电流诊断故障的可区分性,以及故障模式的分类。然后对输出电流提取故障特征,并采用极限学习机完成故障初级分类。对于初级分类结果为电流不可区分故障情况,再根据桥臂电压信息运用规则推理法实现故障二级精确诊断。诊断实验表明,该方法能够实现NPC三电平的多模式故障诊断,且故障诊断方法简单、定位精确、快速、鲁棒性强。     

An advanced auto-inspection system for micro-router collapse

In this paper, we propose an auto-optical inspection (AOI) system that can inspect micro-router (router) collapse automatically. The router is a tool used to cut a printed circuit board (PCB). A few types of defects could occur in the routers and cause unexpected damage to the PCBs. Among these defects, collapse is the most critical defect that must be detected. Currently, router manufacturing companies rely on human inspectors to control the router quality. We first extract the silhouette edges and associated features (peaks and valleys) of a router’s silhouette image by computer vision technique. Then, these silhouette edges and associated features are used to reconstruct a set of 2D isograms that correspond to the router surface. Finally, a pattern recognition method is devised to identify and classify some features of the pattern in the 2D isograms. In this study, two types of routers with different diameters are used for inspection experiments. There are 15 routers of each type. The experimental results reveal that the proposed AOI system can robustly and successfully detect the collapse of diamond-patterned routers with different sizes. The successful detection rate is above 96%. The proposed AOI system can assist in determining the quality of the routers.     

基于上位机控制的电焊机遥控操作系统的设计

针对目前采油厂普遍使用的直流电焊机在焊接过程中,需要根据金属质地的不同改变焊接电流大小,同时受空间限制,特殊工况下手工操作不便等问题,对基于上位机控制的电焊机遥控调整焊接电流外加装置进行了研究,提出了一种基于低功耗微处理器MSP430F149控制的无线射频遥控调流装置,该装置由电焊机控制端(下位机)和手持遥控端(上位机)两部分组成.手持遥控调流装置由微处理器MSP430F149,无线通信模块(nRF905),按键电路,显示模块,稳压电源及外围电路组成;电焊机控制装置由微处理器MSP430F149,直流电焊机,环境参数传感器,无线通信模块,稳压电源,有线无线转换开关及电流调节模块组成;实现了直流电焊机工作参数的远程调控,能够节约焊接施工的时间和人力成本,扩展焊机的使用场合。     

基于机器视觉的高密度电路板缺陷检测系统

为减少高密度电路板的缺陷误报率,研究一种新型自动光学检测系统(AOI);系统采用自行研制的多色LED照明系统,利用机器视觉获取被测PCB的图像,通过图像处理软件系统快速准确地识别出各种缺陷;系统利用获取的彩色图像信息,根据各种缺陷的特征信息不同,采用OPENCV对各种缺陷的检测算法进行改进,使得系统性能有很大改进;对30块同类HDI型PCB的36300个检测点进行测试,测试结果证明,系统PCB缺陷的检出率高达99.87%,误报率只有0.32%。     

电子元器件二次筛选质量控制方法

二次筛选是电子元器件在装机使用前可靠性的重要保障过程,文章叙述了电子元器件二次筛选的基本概念,重点对在二次筛选过程中的主要筛选工序（电功率老炼和电参数测试）的工艺质量控制方法进行了阐述。     

基于LMISVM的相关反馈图像检索

相关反馈技术是基于内容图像检索研究的热点。本文针对现有SVM相关反馈中假定相关图像的所有特征为相关这一不完全准确假设,提出了MISVM短期机器学习相关反馈方法。该方法采用多示例学习方法确定图像中每个特征的相关程度来提高SVM的分类准确性;在此基础上,为进一步提高系统反馈速度与准确率,通过保存以前训练好的分类器和反馈样本,提出了基于LMISVM长期机器学习的相关反馈方法。文中提出的两种方法与其它方法进行了比较实验,结果表明该方法优于其它方法。     

Detecting stealth DHCP starvation attack using machine learning approach

Dynamic Host Configuration Protocol (DHCP) is used to automatically configure clients with IP address and other network configuration parameters. Due to absence of any in-built authentication, the protocol is vulnerable to a class of Denial-of-Service (DoS) attacks, popularly known as DHCP starvation attacks. However, known DHCP starvation attacks are either ineffective in wireless networks or not stealthy in some of the network topologies. In this paper, we first propose a stealth DHCP starvation attack which is effective in both wired and wireless networks and can not be detected by known detection mechanisms. We test the effectiveness of proposed attack in both IPv4 and IPv6 networks and show that it can successfully prevent other clients from obtaining IP address, thereby, causing DoS scenario. In order to detect the proposed attack, we also propose a Machine Learning (ML) based anomaly detection framework. In particular, we use some popular one-class classifiers for the detection purpose. We capture IPv4 and IPv6 traffic from a real network with thousands of devices and evaluate the detection capability of different machine learning algorithms. Our experiments show that the machine learning algorithms can detect the attack with high accuracy in both IPv4 and IPv6 networks.     

基于随机子空间核极端学习机集成的高光谱遥感图像分类

结合随机子空间和核极端学习机集成提出了一种新的高光谱遥感图像分类方法。首先利用随机子空间方法从高光谱遥感图像数据的整体特征中随机生成多个大小相同的特征子集;然后利用核极端学习机在这些特征子集上进行训练从而获得基分类器;最后将所有基分类器的输出集成起来,通过投票机制得到分类结果。在高光谱遥感图像数据集上的实验结果表明:所提方法能够提高分类效果,且其分类总精度要高于核极端学习机和随机森林方法。     

基于深度学习的单幅图片超分辨率重构研究进展

图像超分辨率重构技术是一种以一幅或同一场景中的多幅低分辨率图像为输入, 结合图像的先验知识重构出一幅高分辨率图像的技术. 这一技术能够在不改变现有硬件设备的前提下, 有效提高图像分辨率. 深度学习近年来在图像领域发展迅猛, 它的引入为单幅图片超分辨率重构带来了新的发展前景. 本文主要对当前基于深度学习的单幅图片超分辨率重构方法的研究现状和发展趋势进行总结梳理: 首先根据不同的网络基础对十几种基于深度学习的单幅图片超分辨率重构的网络模型进行分类介绍, 分析这些模型在网络结构、输入信息、损失函数、放大因子以及评价指标等方面的差异; 然后给出它们的实验结果, 并对实验结果及存在的问题进行总结与分析; 最后给出基于深度学习的单幅图片超分辨率重构方法的未来发展方向和存在的挑战.     

An algorithm for semi-supervised learning in image retrieval

We study the problem of image retrieval based on semi-supervised learning. Semi-supervised learning has attracted a lot of attention in recent years. Different from traditional supervised learning. Semi-supervised learning makes use of both labeled and unlabeled data. In image retrieval, collecting labeled examples costs human efforts, while vast amounts of unlabeled data are often readily available and offer some additional information. In this paper, based on support vector machine (SVM), we introduce a semi-supervised learning method for image retrieval. The basic consideration of the method is that, if two data points are close to each, they should share the same label. Therefore, it is reasonable to search a projection with maximal margin and locality preserving property. We compare our method to standard SVM and transductive SVM. Experimental results show efficiency and effectiveness of our method.     

Line balancing of PCB assembly line using immune algorithms

Printed Circuit Boards (PCBs) are widely used in most electronic devices. Typically, a PCB design has a set of components that needs to be assembled. In a broad sense, this assembly task involves placing PCB components at designated location on a PCB board; fixing PCB components; and testing the PCB after assembly operation to ensure that it is in proper working order. The stringent requirements of having a higher component density on PCBs, a shorter assembly time, and a more reliable product prompt manufacturers to automate the process of PCB assembly. Frequently, a few placement machines may work together to form an assembly line. Thus, the application of more than one machine for component placement on a PCB presents a line-balancing problem, which is basically concerned with balancing the workload of all the machines in an assembly line. This paper describes the application of a new artificial intelligence technique known as the immune algorithm to PCB component placement as well as the line balancing of PCB assembly line. It also includes an overview of PCB assembly and an outline of the assembly line balancing problem. Two case studies are used to validate the IA engine developed in this work. The details of IA, the IA engine and the case studies are presented.     

基于卷积去噪自编码器的芯片表面弱缺陷检测方法

芯片表面缺陷会影响芯片的外观和性能,因此表面缺陷检测是芯片生产过程中的重要环节。具有缺陷与背景对比度低、缺陷较小等特点的弱缺陷给传统检测方法带来了挑战。因为近年来深度学习在机器视觉领域展现出了强大的能力,所以文中采用基于深度学习的方法来研究芯片表面弱缺陷的检测问题。该方法将芯片表面缺陷看作噪音,首先应用卷积去噪自编码器(Convolutional Denoising Auto-encoders,CDAE)重构无缺陷图像,然后用重构的无缺陷图像减去输入图像,获得包含缺陷信息的残差图。因为残差图中已经消除了背景的影响,所以最后可以基于残差图较容易地进行缺陷检测。由于基于CDAE重构芯片背景的无缺陷图像时存在随机噪音,导致弱缺陷可能会湮没在重构噪音中,为此,文中提出了重叠分块策略抑制重构噪音,以便更好地检测弱缺陷。因为CDAE是无监督学习网络,所以训练时无需进行大量的人工数据标注,这进一步增强了该方法的可应用性。通过对真实芯片表面数据进行测试,验证了所提方法在芯片表面检测上的有效性。     

基于AOI的PCB孔缺陷检测法

PCB(Printed Circuit Board,印制电路板)是提供电子零件安装与插接的基板,实现电子元器件之间的电气互连,是电子产品不可缺少的基本构成部分。过孔是PCB的重要组成部分,在PCB生产线上由于钻孔和电镀等工艺技术的限制,造成孔内缺陷,影响元器件的电气连接及焊接质量。因此为了保证过孔的可靠性必须进行缺陷检测。基于AOI(Automatic Optical Inspection,自动光学检测)原理,用计算机对PCB孔透光过程进行光学衍射建模仿真。将缺陷孔和无缺陷孔的仿真衍射图样进行比较,可判断出PCB孔是否存在缺陷。结果证明,这种缺陷检测方法具有可行性且操作简便。     

基于统计分析的PCB组装缺陷特征学习方法*

为了减少自动光学检测系统对用户经验的依赖,提出了一种基于统计分析的PCB组装缺陷特征学习方法。该方法通过对良品和不良品样本图像的统计学习优选出分类能力强的特征,再采用最小风险贝叶斯决策得到特征分类参数。实验结果表明,该算法有效地简化了用户检测程序的编程和调试,提高了AOI的使用效率和准确率。