基于VGGNet和多谱带循环网络的高光谱人脸识别系统

为了提高光谱人脸数据表征人脸特征的有效性，提出一种基于VGGNet和多谱带循环训练的高光谱人脸识别方法。首先，在光谱人脸图像的预处理阶段，采用多任务卷积神经网络（MTCNN）进行高光谱人脸图像的精确定位，并利用混合通道的方式对高光谱人脸数据进行增强；然后，基于卷积神经网络（CNN）结构建立一个面向高光谱人脸识别的VGG12深度网络；最后，基于高光谱人脸数据的特点，引入多谱带循环训练方法训练建立的VGG12网络，完成最后的训练和识别。在公开的UWA-HSFD和PolyU-HSFD高光谱人脸数据集的实验结果表明，所提方法取得了比其他深度网络（如DeepID、DeepFace、VGGNet）更好的识别性能。     

利用CNN和PCA约束优化模型实现稀疏表示分类

目的 传统的稀疏表示分类方法运用高维数据提升算法的稀疏分类能力,早已引起了广泛关注,但其忽视了测试样本与训练样本间的信息冗余,导致了不确定性的决策分类问题。为此,本文提出一种基于卷积神经网络和PCA约束优化模型的稀疏表示分类方法（EPCNN-SRC）。方法 首先通过深度卷积神经网络计算,在输出层提取对应的特征图像,用以表征原始样本的鲁棒人脸特征。然后在此特征基础上,构建一个PCA（principal component analysis）约束优化模型来线性表示测试样本,计算对应的PCA系数。最后使用稀疏表示分类算法重构测试样本与每类训练样本的PCA系数来完成分类。结果 本文设计的分类模型与一些典型的稀疏分类方法相比,取得了更好的分类性能,在AR、FERET、FRGC和LFW人脸数据库上的实验结果显示,当每类仅有一个训练样本时,EPCNN-SRC算法的识别率分别达到96.92%、96.15%、86.94%和42.44%,均高于传统的表示分类方法,充分验证了本文算法的有效性。同时,本文方法不仅提升了对测试样本稀疏表示的鲁棒性,而且在保证识别率的基础上,有效降低了算法的时间复杂度,在FERET数据库上的运行时间为4.92 s,均低于一些传统方法的运行时间。结论 基于卷积神经网络和PCA约束优化模型的稀疏表示分类方法,将深度学习特征与PCA方法相结合,不仅具有较好的识别准确度,而且对稀疏分类也具有很好的鲁棒性,尤其在小样本问题上优势显著。     

Recognizing arabic handwritten characters using deep learning and genetic algorithms

Automated techniques for Arabic content recognition are at a beginning period contrasted with their partners for the Latin and Chinese contents recognition. There is a bulk of handwritten Arabic archives available in libraries, data centers, historical centers, and workplaces. Digitization of these documents facilitates (1) to preserve and transfer the country’s history electronically, (2) to save the physical storage space, (3) to proper handling of the documents, and (4) to enhance the retrieval of information through the Internet and other mediums. Arabic handwritten character recognition (AHCR) systems face several challenges including the unlimited variations in human handwriting and the leakage of large and public databases. In the current study, the segmentation and recognition phases are addressed. The text segmentation challenges and a set of solutions for each challenge are presented. The convolutional neural network (CNN), deep learning approach, is used in the recognition phase. The usage of CNN leads to significant improvements across different machine learning classification algorithms. It facilitates the automatic feature extraction of images. 14 different native CNN architectures are proposed after a set of try-and-error trials. They are trained and tested on the HMBD database that contains 54,115 of the handwritten Arabic characters. Experiments are performed on the native CNN architectures and the best-reported testing accuracy is 91.96%. A transfer learning (TF) and genetic algorithm (GA) approach named “HMB-AHCR-DLGA” is suggested to optimize the training parameters and hyperparameters in the recognition phase. The pre-trained CNN models (VGG16, VGG19, and MobileNetV2) are used in the later approach. Five optimization experiments are performed and the best combinations are reported. The highest reported testing accuracy is 92.88%.

     

Convolutional adaptive denoising autoencoders for hierarchical feature extraction

Convolutional neural networks (CNNs) are typical structures for deep learning and are widely used in image recognition and classification. However, the random initialization strategy tends to become stuck at local plateaus or even diverge, which results in rather unstable and ineffective solutions in real applications. To address this limitation, we propose a hybrid deep learning CNN-AdapDAE model, which applies the features learned by the AdapDAE algorithm to initialize CNN filters and then train the improved CNN for classification tasks. In this model, AdapDAE is proposed as a CNN pre-training procedure, which adaptively obtains the noise level based on the principle of annealing, by starting with a high level of noise and lowering it as the training progresses. Thus, the features learned by AdapDAE include a combination of features at different levels of granularity. Extensive experimental results on STL-10, CIFAR-10, andMNIST datasets demonstrate that the proposed algorithm performs favorably compared to CNN (random filters), CNNAE (pre-training filters by autoencoder), and a few other unsupervised feature learning methods.     

Effective detection of exposed target regions based on deep learning from multimedia data

With the development of high-performance visual sensors, it has been very easy to obtain a variety of image data. Of these image data, human face regions contain personal information to distinguish one from the others. Therefore, it is important to accurately detect unhidden face regions from an input image. This paper proposes a method of robustly detecting human face regions from an input color image with the use of a deep learning algorithm, one of the machine learning algorithms. The proposed method first transforms the RGB color model of an input image to the YC_bC_r color model, and then removes other regions than face regions to segment skin regions with the use of the pre-learned elliptical skin color distribution model. Subsequently, a CNN model-based deep learning algorithm was applied to robustly detect human face regions from the detected skin regions in the previous step. As a result, the proposed method segments face regions more efficiently than an existing method. The face region detection method proposed in this paper is expected to be usefully applied to practical areas related to multimedia data processing, such as video surveillance, target blocking, image security, visual data analysis, and object recognition and tracking.

     

自适应增强卷积神经网络图像识别

目的 为了进一步提高卷积神经网络的收敛性能和识别精度,增强泛化能力,提出一种自适应增强卷积神经网络图像识别算法。方法 构建自适应增强模型,分析卷积神经网络分类识别过程中误差产生的原因和误差反馈模式,针对分类误差进行有目的地训练,实现分类特征基于迭代次数和识别结果的自适应增强以及卷积神经网络权值的优化调整。自适应增强卷积神经网络与多种算法在收敛速度和识别精度等性能上进行对比,并在多种数据集上检测自适应卷积神经网络的泛化能力。结果 通过对比实验可知,自适应增强卷积神经网络算法可以在很大程度上优化收敛效果,提高收敛速度和识别精度,收敛时在手写数字数据集上的误识率可降低20.93%,在手写字母和高光谱图像数据集上的误识率可降低11.82%和15.12%;与不同卷积神经网络优化算法对比,误识率比动态自适应池化算法和双重优化算法最多可降低58.29%和43.50%;基于不同梯度算法的优化,误识率最多可降低33.11%;与不同的图像识别算法对比,识别率也有较大程度提高。结论 实验结果表明,自适应增强卷积神经网络算法可以实现分类特征的自适应增强,对收敛性能和识别精度有较大的提高,对多种数据集有较强的泛化能力。这种自适应增强模型可以进一步推广到其他与卷积神经网络相关的深度学习算法中。     

基于循环生成对抗网络的含遮挡人脸识别

人脸图像的遮挡会严重影响人脸识别准确率,当前处理带遮挡人脸识别的方法主要有丢弃法和修复法两种。丢弃法因忽略或丢弃大量遮挡区域的有效特征易造成识别准确率不高,而当前大多数修复法需要原图的相关信息,限制了其应用。针对现有含遮挡人脸识别方法存在的问题,提出一种基于循环生成对抗网络的人脸识别算法,通过利用2对生成器和判别器进行循环训练,实现遮挡人脸图像的盲修复,此过程不需要除遮挡区域外的原图信息。在此基础上,采用ResNet-50网络对修复后人脸进行识别,该网络通过跳跃连接消除深层卷积神经网络中的网络退化问题,能够降低网络训练的难度,且不会增加额外的参数和计算量。针对盲修复后人脸特征存在类内差异大和类间差异小的特性,引入一种能够量化类间距离的分类损失函数RegularFace作为识别网络损失函数。实验结果表明,与DCGAN+CNN算法相比,所提算法对不同遮挡类型和遮挡区域的人脸图像识别准确率均有所提高,当线性遮挡面积为40%时,所提算法的识别准确率提高了14.4个百分点。     

基于深度学习的垃圾图片处理与识别

针对城市环境卫生提出的对市民生活垃圾进行分类回收的要求,考虑计算机卷积神经网络在图片分类中的强大表现,提出了基于深度学习中卷积神经网络对垃圾图片处理以及输出识别的新模型与方法。针对目前图像局部特征表达存在的复杂性,模糊性等不足,采用特征多层池化以及系统神经网络学习的方式进行优化。同时在ResNet101模型的基础上设计并构建了基于CNN(Convolutional Neural Network)算法的新模型框架,此系统模型也能实现端与端的实时识别。新模型提高了对训练样本图像信息提取的精确度以及图片识别的准确率,实验表明识别准确率平均提高了10%。为未来实现人工智能垃圾分类提供图像识别模型基础。     

Developing real-time training dataset for human racial classification

In every image processing venture the quality of data sample collected and the processing techniques poses direct impact on the result. This work explains development of these vital phases in the context to exploit it for racial classification. Here we propose a novel Indian regional face database (IRFD) consisting of large set distinctive face images of north, east, west and south regions of India to mitigate the scarcity of regional and labeled face images for future supervised classification process. The face images are collected from different universities and acquired through both online and offline mode. Due to this discrepancy the face database is exposed to challenges like varying dimension size, non-uniform background, low resolution, illumination, and pose variation. In view of addressing these problems we have proposed competent image processing techniques to enhance the quality of images. Varying size and low resolution were the main issues among others encountered while training Convolutional Neural Network (CNN) model. To handle this we have developed an expeditious compression algorithm which would reduce large size of all images to ±97% less in size without compromising the quality. Further to enhance low quality images we have proposed brightness and contrast adjusting algorithm. The efficiency of this quantitative and qualitative data set is evaluated against CNN model which has yielded ±88.21% accuracy under racial classification.

     

An adaptive gravitational search optimization (AGSO)-based convolutional neural network—long short-term memory (CNN-LSTM) approach for face recognition and classification

Face recognition and classification have gained increasing attraction in the recent decades due to their widespread adoption in real time application systems. Most of the conventional research efforts focused on developing face recognition frameworks using enhanced optimization-based classification methods, they are hampered by issues such as computational complexity, increased overhead, limited capacity to handle large datasets, and lengthy processing time. The novel contribution of this paper is to develop a highly competent and precise face recognition methodology through an innovative mechanism. In this framework, the initial step involves face detection from input images using an analytical face parts detection methodology. Subsequently, the tutor face filtering (TFF) technique is applied to preprocess the face image, enhancing its quality and filtering out noise content. Following this preprocessing step, features are extracted from the processed image using the direction-based pattern extraction (DBPE) model. To improve classifier accuracy, a novel adaptive gravitational search optimization (AGSO) technique is employed to select the optimal features during model training. Finally, an integrated deep learning model, referred to as convolutional neural network — long short-term memory (LSTM), is utilized for accurate face image recognition based on the selected optimal features. To assess and compare the system's performance, various metrics are employed in the results analysis to demonstrate the superiority of the proposed approach.     

A multi-biometric iris recognition system based on a deep learning approach

Multimodal biometric systems have been widely applied in many real-world applications due to its ability to deal with a number of significant limitations of unimodal biometric systems, including sensitivity to noise, population coverage, intra-class variability, non-universality, and vulnerability to spoofing. In this paper, an efficient and real-time multimodal biometric system is proposed based on building deep learning representations for images of both the right and left irises of a person, and fusing the results obtained using a ranking-level fusion method. The trained deep learning system proposed is called IrisConvNet whose architecture is based on a combination of Convolutional Neural Network (CNN) and Softmax classifier to extract discriminative features from the input image without any domain knowledge where the input image represents the localized iris region and then classify it into one of N classes. In this work, a discriminative CNN training scheme based on a combination of back-propagation algorithm and mini-batch AdaGrad optimization method is proposed for weights updating and learning rate adaptation, respectively. In addition, other training strategies (e.g., dropout method, data augmentation) are also proposed in order to evaluate different CNN architectures. The performance of the proposed system is tested on three public datasets collected under different conditions: SDUMLA-HMT, CASIA-Iris-V3 Interval and IITD iris databases. The results obtained from the proposed system outperform other state-of-the-art of approaches (e.g., Wavelet transform, Scattering transform, Local Binary Pattern and PCA) by achieving a Rank-1 identification rate of 100% on all the employed databases and a recognition time less than one second per person.     

复杂背景图像的字符识别算法研究

为了解决字符识别过程中的局部曝光、印刷字符的断裂以及变形和自然环境下的背景污染等问题, 提出了一种分块处理与卷积神经网络(CNN)相结合的字符图像识别算法. 首先利用OpenCV机器视觉库, 结合分块处理、伽马运算、参数调整等方法对产品零件表面印刷字符进行预处理, 初步解决图像局部曝光和字符断裂问题; 其次为了获得单个字符图像, 利用数学形态学算法对局部曝光处理后的二值化图像进行分步分割, 进而去掉字符间的无用信息; 最后利用Keras模块为字符识别提供的API搭建CNN模型, 经过对100多张字符的识别训练, 准确率高达96.9%, 为某汽车零部件自动化生产中的字符识别提供了可靠的依据.     

基于低秩约束的极限学习机高效人脸识别算法

复杂应用场景中,光照变化、遮挡和噪声等干扰使得将像素特征作为相似性度量的识别算法的图像类内差大于类间差,降低了人脸识别性能。针对这一问题,提出了一种低秩约束的极限学习机鲁棒性人脸识别算法,提升了复杂场景下的识别性能。首先,利用人脸图像分布的子空间线性假设,将待识别图像聚类到相对应的样本子空间;其次,将像素域分解为低秩特征子空间和稀疏误差子空间,依据图像子空间的低秩性对噪声鲁棒的原理,提取人脸图像的低秩结构特征训练极限学习机的前向网络;最后,实现对噪声干扰鲁棒的极限学习机人脸识别算法。实验结果表明,相比前沿的人脸识别算法,所提方法不仅识别精度高、算法时间复杂度低,且具有较好的实用性。     

基于卷积神经网络的面罩语音识别

针对带噪面罩语音识别率低的问题,结合语音增强算法,对面罩语音进行噪声抑制处理,提高信噪比,在语音增强中提出了一种改进的维纳滤波法,通过谱熵法检测有话帧和无话帧来更新噪声功率谱,同时引入参数控制增益函数;提取面罩语音信号的Mel频率倒谱系数(MFCC)作为特征参数;通过卷积神经网络(CNN)进行训练和识别,并在每个池化层后经局部响应归一化(LRN)进行优化.实验结果表明:该识别系统能够在很大程度上提高带噪面罩语音的识别率.     

基于CNN的监控视频中人脸图像质量评估

在公共安全领域,监控视频中的人脸识别技术是不可或缺的技术,成为研究热点.而监控视频中低质量的人脸图像会大大降低整个人脸识别系统的识别准确率,系统难以更广泛地被投入实际使用.本文提出了一种基于CNN的人脸图像质量评估方法.通过对Alexnet模型进行改进,将网络中的多个卷积层与全连接层连接,从而提取不同尺度的图像特征.通过端到端的训练过程,预测人脸图像质量分数.另外,采用人脸识别算法来标定人脸图像的质量分数,使质量分数能更有效地筛选出适合识别算法的图像.在Color FERET数据集上实验表明,本文方法能够准确地对人脸图像进行质量评估.而在实际采集的监控视频数据集上实验表明,本文方法能筛选出高质量的人脸图像用作后续人脸识别,提高人脸识别准确率.     

混合深度学习模型C-RF及其在手写数字识别中的应用

卷积神经网络（Convolutional neural network,CNN）是一种常见的深度学习模型,受人类视觉认知机制启发而来,能够从原始图像得到有效的特征表达。CNN模型在图像识别领域不断取得突破,但是在训练过程中需要花费大量时间。随机森林（Random forest,RF）在分类和回归上具有很高的精度,训练速度快并且不容易出现过拟合的问题,现有的基于RF的分类器都依赖手工选取的特征。针对以上问题,本文提出了基于CNN的C-RF模型,把CNN提取到的特征输入RF中进行分类。由于随机权值网络同样可以得到有效的结果,所以不用梯度算法调整网络参数,以免消耗大量时间。最后在MNIST数据集和Rotated MNIST数据集上进行了实验,结果表明C-RF模型的分类精度比RF有了较大的提高,同时泛化能力也有所提升。     

平稳小波域深度残差CNN用于低剂量CT图像估计

针对低剂量计算机断层扫描（LDCT）重建图像中存在大量噪声的问题,提出了一种平稳小波的深度残差卷积神经网络（SWT-CNN）模型,可以从LDCT图像估计标准剂量计算机断层扫描（NDCT）图像。该模型在训练阶段,将LDCT图像经平稳小波（SWT）三级分解后的高频系数作为输入,将LDCT图像高频系数与NDCT图像高频系数相减得到残差系数作为标签,通过深度卷积神经网络（CNN）学习输入和标签之间的映射关系;在测试阶段,利用此映射关系即可从LDCT图像的高频系数中预测NDCT高频系数,最后通过平稳小波反变换（ISWT）重构预测的NDCT图像。实验采用50对大小为512×512的同一体模的常规剂量胸腔及腹腔扫描切片和投影域添加噪声后的重建图像作为数据集,其中45对作为训练集,其余5对作为测试集。将所提模型与效果较好的非局部降噪算法、K-奇异值分解（K-SVD）算法、匹配三维滤波（BM3D）算法及图像域CNN（Image-CNN）模型对比,实验结果表明,SWT-CNN模型预测的NDCT图像信噪比（PSNR）和结构相似性（SSIM）高,且均方根误差（RMSE）小于其他算法处理结果。该模型对于提高低剂量CT图像质量是可行且有效的。     

结构化低秩字典学习的人脸识别

目的 现实中采集到的人脸图像通常受到光照、遮挡等环境因素的影响,使得同一类的人脸图像具有不同程度的差异性,不同类的人脸图像又具有不同程度的相似性,这极大地影响了人脸识别的准确性。为了解决上述问题对人脸识别造成的影响,在低秩矩阵恢复理论的基础上提出了具有识别力的结构化低秩字典学习的人脸识别算法。方法 该算法基于训练样本的标签信息将低秩正则化以及结构化稀疏同时引入到学习的具有识别力的字典上。在字典学习过程中,首先利用样本的重建误差约束样本与字典之间的关系;其次将Fisher准则应用到稀疏编码过程中,使其编码系数具有识别能力;由于训练样本中的噪声信息会影响字典的识别力,所以在低秩矩阵恢复理论的基础上将低秩正则化应用到字典学习过程中;接着,在字典学习过程中加入了结构化稀疏使其不丢失结构信息以保证对样本进行最优分类;最后再利用误差重构法对测试样本进行分类识别。结果 本文算法在AR以及ORL人脸数据库上分别进行了实验仿真。在AR人脸数据库中,为了分析样本不同维数对实验结果造成的影响,选取了第一时期拍摄的每人6幅图像,包括1幅围巾遮挡,2幅墨镜遮挡以及3幅脸部表情变化以及光照变化（未被遮挡）的图像作为训练样本,同时选取相同组合的样本图像作为测试样本,无论哪种方法,图像的维度越高识别率越高。对比SRC （sparse representation based on classification）算法与DKSVD （discriminative K-means singular value decomposition）算法的识别率可知,DKSVD算法通过字典学习减缓了训练样本中的不确定因素对识别结果的影响;对比DLRD_SR （discriminative low-rank dictionary learning for sparse representation）算法与FDDL （Fisher discriminative dictionary learning）算法的识别率可知,当图像有遮挡等噪声信息存在时,字典低秩化可以提高至少5.8%的识别率;对比本文算法与DLRD_SR算法可知,在字典学习的过程中加入Fisher准则后识别率显著提高,同时理想稀疏值能保证对样本进行最优的分类。当样本图像的维度达到500维时人脸图像在有围巾、墨镜遮挡的情况下识别率可达到85.2%;其中墨镜和围巾的遮挡程度分别可以看成是人脸图像的20%和40%,为了验证本文算法在不同脸部表情变化、光照改变以及遮挡情况下的有效性,根据训练样本的具体图像组合情况进行实验。无论哪种样本图像组合,本文算法在有遮挡存在的样本识别中具有显著优势。在训练样本只包含脸部表情变化、光照变化以及墨镜遮挡图像的情况下,本文算法的识别率高于其他算法至少2.7%,在训练样本只包含脸部表情变化、光照变化以及围巾遮挡图像的情况下,本文算法的识别率高于其他算法至少3.6%,在训练样本包含脸部表情变化、光照变化、围巾遮挡以及墨镜遮挡图像的情况下,其识别率高于其他算法至少1.9%。在ORL人脸数据库中,人脸图像在无遮挡的情况下识别率达到95.2%,稍低于FDDL算法的识别率;在随机块遮挡程度达到20%时,相比较于SRC算法、DKSVD算法、FDDL算法以及DLRD_SR算法,本文算法的识别率最高;当随机块遮挡程度达到50%时,以上算法的识别率均不高,但本文算法的其识别率仍然最高。结论 本文算法在人脸图像受到遮挡等因素的影响时具有一定的鲁棒性,实验结果表明该算法在人脸识别方面具有可行性。     

基于深度代价敏感CNN的年龄估计算法

针对年龄估计中样本数量不平衡及不同类间发生误分类时付出代价不同的问题,将代价敏感性嵌入深度学习框架中,提出基于深度代价敏感CNN的年龄估计算法.首先为每个年龄类别分别建立损失函数,解决训练样本的不平衡问题.然后,定义代价向量以反映不同类之间发生误分类而付出的代价差异性,构造逆交叉熵误差函数.最后,融合上述方法,为卷积神经网络(CNN)构造一个损失函数,使CNN在训练阶段学习针对年龄估计的鲁棒人脸表征.在不同种族的年龄估计标准图像集上的实验验证文中算法的有效性.     

一种基于深度学习的改进人脸识别算法

针对当前许多算法在非约束条件下特征判别能力不强、人脸识别性能不佳等问题,提出一种基于深度学习的改进人脸识别算法,通过训练多任务级联卷积神经网络,完成非约束图像的人脸检测和人脸归一化,提高训练图像的人脸信息,减少对模型的干扰。同时使用Softmax损失与中心损失联合监督训练模型,优化类内聚合、类间分散。实验结果表明,该算法提高了模型的特征判别能力,在LFW标准测试集上达到了较高的识别率。