一种基于同质映射的视频图像中文本检测方法

提出了一种基于同质映射域的纹理特征的文本检测方法,并通过实验验证了该方法的性能.该方法与传统的文本检测方法的不同之处在于,首先将图像映射到同质性空间域中,在此空间域中计算纹理特征,然后通过支持向量机(SVM)分类器确定文本区域.与直接在图像空间域中提取纹理特征的方法相比,该方法对复杂背景下的文本检测更为有效,能有效地解决场景纹理特征与文本区域相近似造成的文本检测错误.     

一种基于马尔可夫随机场的运动目标检测算法

本文提出了一种基于马尔可夫随机场的运动目标检测算法.针对传统时间分割使用主观固定阁值的缺点,使用马尔可夫随机场模型对差分图像建模,并提出了一种新的模型阶次选择算法,以及一种可以加速收敛过程的随机场迭代算法.采用期望最大算法获取高斯分布参数并检测运动变化区域,利用形态学运算修正时间分割模板;空间分割部分提出了基于人眼视觉特征的改进分水岭算法,有效地解决了过分割问题;最后对时、空间分割结果进行信息融合处理,从而得到完整的运动目标.仿真实验结果证明了本文算法可以有效地分割视频运动目标.     

基于卷积神经网络的森林火灾烟雾探测算法研究

为解决森林火灾烟雾检测过程中受外界干扰且由于烟雾存在多种静、动态特性导致识别难度高的问题,提出一种基于卷积神经网络的火灾烟雾视频探测算法提取可疑区域特征并进行模式分类,进而检测出火灾烟雾。实验结果表明:该算法在各种视频场景下均具有良好的烟雾识别性能,并能与灭火装置通信对初期林火进行扑灭,为森林火灾探测扑救装备的智能化、高效化提供了新思路。     

基于蚁群算法和LBP-HF的场景文本定位

针对复杂背景下文本误检率较高的问题,提出了一种基于蚁群聚类和LBP-HF特征验证的复杂场景文本定位算法。该算法首先利用小波高频系数统计特征表达文本模式,采用蚁群聚类算法对文本像素和背景像素进行分类,得到所有可能的文本区域;然后提取更具区分力的LBP-HF纹理特征对侯选的文本区进行验证,获得文本的准确位置。实验结果表明,所提出的基于LBP-HF特征的验证方法能够有效区分文本和非文本区域,使复杂背景下的文本误检率明显下降。     

基于深度学习的机器视觉目标检测算法及在票据检测中应用

基于深度学习的目标检测是机器视觉应用的重要方面。该文系统总结基于区域候选的目标检测算法、基于回归方法的目标检测算法及其他优化算法的算法思想、网络架构、演进过程、技术指标、应用场景,指出在机器视觉系统应用中,应充分考虑检测对象、检测精度、实时性能要求,结合不同目标检测算法特点,选择最合适的检测算法。最后,面向票据检测需求,分析目标检测算法在票据图像位置检测、防伪特征检测、文本信息检测中的应用。     

基于位置的特征项权重算法

Web文本分类是Web文本挖掘的主要内容,而特征项权重的计算是web文本分类中一个非常重要的步骤。Web文本一般由标题、描述和正文三部分组成。根据Web文本的这一特点,本文提出了一种基于位置的特征项权重算法,并使用此算法对Web文本进行了分类实验。实验结果表明该算法有效提高了Web文本分类系统的分类性能。     

一种动态模板匹配的卡尔曼滤波跟踪方法

在视频跟踪系统中,运动目标检测是实现跟踪的前提和难点.为了能够有效地检测出目标,提出了一种基于动态模板匹配和卡尔曼滤波的目标跟踪算法.首先将前两帧图像差分检测运动目标区域,提取特征点;然后利用卡尔曼滤波在搜索区域中找到与目标模型最匹配的候选目标位置并与当前帧目标模板进行匹配;最后将特征点流失率作为限定阈值,采用模板更新策略动态更新模板.跟踪实验表明,该算法具有很好的匹配精度与实时性,对目标姿态变化、大小变化、遮挡问题等有很好的鲁棒性.     

基于二阶矩显著性估算的局部不变特征提取

借鉴图像显著性区域的检测思想,提出一种基于局部二阶矩显著性估算的局部不变特征提取算法.利用二阶矩矩阵对尺度空间下局部图像的各向异性程度的估算作用,在图像尺度空间中对局部特征提取区域的信息显著性进行评估,并根据显著性进行局部不变特征的提取,提取出拥有较高显著性的局部不变特征,增加了匹配特征点对的数量和尺度跨度.真实图像实验证明,该算法在保持局部特征各种不变性的基础上有效地提高了特征提取和匹配算法的性能.     

区域边缘直方图的目标搜索算法

视频目标搜索是智能视频监控领域一大挑战,提出一种基于灰度图像区域边缘直方图的目标搜索算法.首先,在固定场景的视频数据中,对选定目标进行特征提取,即区域边缘直方图(REH)特征向量;接着在同一场景的未知视频数据中进行前景检测并提取前景目标的特征向量;经滤波处理后,与选定目标特征向量进行匹配,通过相似性度量评判是否搜索成功.实验得到了最佳72.4%的匹配成功率,验证了32维的区域边缘特征向量为最佳描述特征.实验结果表明,本算法能有效地实现目标搜索.     

基于AR自相关峰态值的一类轴承故障检测方法

针对轴承故障检测系统中异常样本数据不易收集以及异常样本数据分布不均导致传统分类算法出现过适应现象等现实应用问题,提出了一种基于自回归(AR)模型自相关系数峰态特征的一类故障检测方法.该方法利用正常样本生成AR模型参数,其他样本在该模型的投影形成残差序列,计算残差序列的自相关系数并取其峰态特征作为相似性的度量.实验结果表明该方法能有效地克服以AR模型参数为特征计算复杂度高且检测性能易受样本大小影响的不足.同时,文章给出了单一故障诊断模型并提出基于粒子群优化算法的阈值设定决策方法.实验中将本方法同其他以AR模型为特征的多层感知机(MLP)及自组织映射(SOM)方法进行比较,实验结果验证了本文建议方法的正确性和有效性.     

Two types of learning effects on maintenance activities

This paper studies a manufacturer with a system prone to failure. The manufacturer performs two types of maintenance activities: preventive maintenance (PM), performed periodically, resets the system, and Minimal Repair (MR), performed after breakdowns, restores the system to working condition. It is assumed that two different types of learning take place: (i) repetition learning: due to the repetitive nature of PM, the manufacturer gains experience and learns to perform the PM activities faster and at a lower cost and (ii) failure learning: each failure gives the manufacturer the opportunity to find the root causes, to learn from mistakes and to improve the system. This paper, the first one to quantify failure learning in maintenance literature, assumes that such learning can then be applied during the next PM activity, which brings down the failure rate for the next PM cycle. For the increasing failure rate case, repetition learning increases the PM frequency, whereas failure learning causes the manufacturer to reduce the optimal number of PM activities. However, for the constant failure rate, repetition learning has no effect on the PM frequency, whereas failure learning may actually increase it.     

Self‐directed Learning Readiness Among Engineering Undergraduate Students

Two studies related to readiness for self‐directed learning of engineering students were performed using the Self‐directed Learning Readiness Scale (SDLRS). A cross‐sectional study of students in the first through final years of study showed that their SDLRS scores are significantly correlated with academic year of study and with grade point average, but not with gender. However, neither academic year of study nor grade point average is a good predictor of SDLRS scores; together they account for less than 5 percent of the observed variance. A second study investigated the effect of a problem‐based learning experience on students' readiness for self‐directed learning. It showed that the average readiness for self‐directed learning increased significantly for students in the problem‐based learning courses. However, investigation of the changes for individual students revealed that only nine of eighteen students showed significant increases in their SDLRS scores, and two showed significant decreases. Potential underlying causes are explored.     

基于深度强化学习的图像去模糊方法

目的为了有效地去除多种图像模糊,提高图像质量,提出基于深度强化学习的图像去模糊方法。方法选用GoPro与DIV2K这2个数据集进行实验,以峰值信噪比(PSNR)和结构相似性(SSIM)为客观评价指标。通过卷积神经网络获得模糊图像的高维特征,利用深度强化学习结合多种CNN去模糊工具建立去模糊框架,将峰值信噪比(PSNR)作为训练奖励评价函数,来选择最优修复策略,逐步对模糊图像进行修复。结果通过训练与测试,与现有的主流算法相比,文中方法有着更好的主观视觉效果,且PSNR值与SSIM值都有更好的表现。结论实验结果表明,文中方法能有效地解决图像的高斯模糊和运动模糊等问题,并取得了良好的视觉效果,在图像去模糊领域具有一定的参考价值。     

美国大学电子化教育的考察

美国大学电子化教育的应用面已经相当广泛。电子化教育注重以人为本，满足人们终身学习的需要；重视师生的教学互动，并采用多种媒体混合的形式，取得了良好的效果。本文根据作者对美国大学教育信息化的考察，对美国大学实行电子化教育的基本依据和主要做法做了介绍。     

MI-STEG: A Medical Image Steganalysis Framework Based on Ensemble Deep Learning

Medical image steganography aims to increase data security by concealing patient-personal information as well as diagnostic and therapeutic data in the spatial or frequency domain of radiological images. On the other hand, the discipline of image steganalysis generally provides a classification based on whether an image has hidden data or not. Inspired by previous studies on image steganalysis, this study proposes a deep ensemble learning model for medical image steganalysis to detect malicious hidden data in medical images and develop medical image steganography methods aimed at securing personal information. With this purpose in mind, a dataset containing brain Magnetic Resonance (MR) images of healthy individuals and epileptic patients was built. Spatial Version of the Universal Wavelet Relative Distortion (S-UNIWARD), Highly Undetectable Stego (HUGO), and Minimizing the Power of Optimal Detector (MIPOD) techniques used in spatial image steganalysis were adapted to the problem, and various payloads of confidential data were hidden in medical images. The architectures of medical image steganalysis networks were transferred separately from eleven Dense Convolutional Network (DenseNet), Residual Neural Network (ResNet), and Inception-based models. The steganalysis outputs of these networks were determined by assembling models separately for each spatial embedding method with different payload ratios. The study demonstrated the success of pre-trained ResNet, DenseNet, and Inception models in the cover-stego mismatch scenario for each hiding technique with different payloads. Due to the high detection accuracy achieved, the proposed model has the potential to lead to the development of novel medical image steganography algorithms that existing deep learning-based steganalysis methods cannot detect. The experiments and the evaluations clearly proved this attempt.