基于有监督的核局部线性嵌入的人脸性别识别

流形学习方法可以有效地发现存在于高维图像空间的低维子流形并进行维数约简,但它是一种非监督学习方法,其鉴别能力反而不如传统的维数约简方法,而且流形学习方法大多没有明晰的投影矩阵,很难直接对新样本进行维数约简.针对这两个问题,提出一种新的有监督的核局部线性嵌入算法(SKLLE,supervised kernel local linear embedding).该算法通过非线性核映射将人脸样本投影到高维核特征空间,然后将人脸局部流形的结构信息和样本的类别信息进行有效地结合进行维数约简,提取低维鉴别流形特征用于分类.SKLLE算法不仅能发现嵌入于高维人脸图像的低维子流形,而且增强了局部类间的联系,同时对新样本有较好的泛化性,实验结果表明该算法能有效的提高人脸性别识别的性能.     

一种同步人脸运动跟踪与表情识别算法

针对单视频动态变化背景下的人脸表情识别问题,提出了一种同步人脸运动跟踪和表情识别算法,并在此基础上构建了一个实时系统.该系统达到了如下目标:首先在粒子滤波框架下结合在线外观模型和柱状几何模型进行人脸三维运动跟踪;接着基于生理知识来提取人脸表情的静态信息;然后基于流形学习来提取人脸表情的动态信息;最后在人脸运动跟踪过程中,结合人脸表情静态信息和动态信息来进行表情识别.实验结果表明,该系统在大姿态和丰富表情下具有较好的综合优势.     

张量局部判别投影的人脸识别

 经典的向量子空间学习算法是以数据流形的向量表示进行计算的,但是在现实世界中数据流形从本质上而言是以张量的形式存在,因此基于张量子空间的学习算法能够更好地揭示流形内在的几何结构.本文提出了一种新的张量子空间的学习算法:张量局部判别投影.首先构建类内和类间图,然后保持流形的局部结构并且利用数据的判别信息,推导出算法的计算公式,最后通过迭代计算广义特征向量,解得最优张量子空间.在标准人脸数据库上的实验表明该算法有效.     

核判别保局投影的人脸识别算法

针对人脸识别中的特征提取问题,提出了核判别保局投影算法,即KDLPP.该算法通过核技巧将人脸样本映射到高维空间,在高维空间中有效地结合人脸局部的流形结构和人脸的判别信息构建了新的目标函数,其优点是在保持人脸流形结构的基础上,充分利用了样本的类别信息,并采用核方法提取了人脸的非线性特征.在ORL和UMIST人脸库上的实验...     

基于异方差PLDA的外观流形建模视频人脸识别

针对现有的视频人脸识别方法不能很好地学习局部模型特定协方差的问题,为了更好地识别视频中的人脸,提出了基于异方差概率线性判别分析(PLDA)的外观流形建模(AMM)算法。首先,借助于高斯分布集合,对训练集中所有的人脸分别进行外观流形建模;然后对从视频中采集到的人脸进行聚类,并使用异方差PLDA模型学习聚类结果,从而获得表征分布的参数;最后,通过点到模型距离对测试人脸的每一帧到训练集的所有聚类进行融合匹配,并根据匹配得分最高原则完成人脸的分类。在两大通用视频人脸数据库Honda及MoBo上的实验验证了所提算法的有效性及稳定性,实验结果表明,相比其他几种较为先进的视频人脸识别算法,所提算法明显提高了识别率,并且大大降低了计算复杂度,有望应用于实时视频人脸识别系统。     

基于改进的格拉斯曼流形的模糊人脸图像识别

传统算法进行模糊人脸识别的过程中,一旦人脸表情发生变化,人脸特征也将发生改变,导致人脸识别的准确性降低。为此,提出一种基于改进的格拉斯曼流形的模糊人脸识别方法。在格拉斯曼流形上构建全部模糊人脸样本图像的近邻图来估计人脸特征分布的几何结构,然后将其作为正则化项整合到模糊人脸识别的目标函数中,从而获得更精确的人脸特征投影矩阵。仿真实验结果表明,利用改进算法进行模糊人脸识别,能够提高识别的准确率和效率,效果令人满意。     

结合深度Q学习和注意模型的视频人脸识别

针对视频人脸识别中存在的动态人脸信息捕捉困难和局部人脸特征提取粗糙的问题,提出了一种基于深度Q学习和注意模型结合的视频人脸识别方法。首先,采用卷积神经网络(Convolutional Neural Network,CNN)训练视频数据可提取多维特征;其次,将视频特征输入注意模型,根据视频数据时间连续性信息得到局部人脸特征、人脸位置和时间记忆单元;最后,采用Q学习迭代计算注意模型的输出,找到含人脸的最优帧序列,并以此计算视频匹配准确度。实验结果表明,该方法有效提高了复杂背景下视频人脸识别的准确性。     

视频监控系统中基于PCA算法的人脸识别

PCA算法提供了一个高维和低维间的线性变换矩阵,这个变换矩阵可以通过求取协方差矩阵的特征向量获得。特征值较大的特征向量反映人脸最大差异性;根据脸部固定结构特点构造人脸平均模板,利用模板匹配来检测图像中的人脸,计算待测图像与特征空间的距离进一步判别是否是数据库中人脸。实验表明,PCA算法在视频监控系统的人脸识别中可以很好地实现人脸特征提取和检测。     

基于LLE的视频摘要提取方法研究

为了自动获取主要视频信息且冗余信息较少的视频摘要,本文提出了LLE-自适应FCM和LLE-自适应阈值FCM算法.这两种方法首先利用流形学习算法局部线性嵌入（LLE）提取视频帧的特征向量,然后将得到的特征向量输入到自适应FCM和自适应阈值FCM中,得出分类效果和聚类中心.自适应FCM通过聚类有效性函数来确定分类类别数,而自适应阈值FCM是通过阈值的自动变化来确定分类类别数.最后把离聚类中心最近的视频帧作为视频摘要.实验的结果表明,在不需要人工干预的情况下,所提取的视频摘要既反映了视频的主要内容,而且冗余信息少.     

基于非线性流形学习的人脸面部运动估计

由于人脸面部运动变化分布在一个低维非线性流形中,基于线性假设的主动外观模型采用主成分分析算法描述人脸形状的变化必然带来额外的估计误差.为降低或消除这一误差,该文提出一种改进的局部线性嵌入算法构建人脸形状-纹理空间,并将其应用于主动外观模型中.实验结果表明,不仅对于面部形变不大的人脸形状,局部线性嵌入-主动外观模型拥有更...     

Gender Classification Based on Multi-Scale and Run-Length Features

Human faces can convey substantial information about a person, such as his or her age, race, identity, gender, and emotions. Such facial information can be obtained through techniques like human facial tracking and detection, facial recognition, gender classification, emotion recognition, as well as age estimation. Of these, gender classification is particularly important due to its diverse applications in the fields such as video surveillance and commercial advertising. In this thesis, we propose a method of gender classification based on run-length histograms. The proposed method uses a run-length histogram to record the position information of pixels, thereby efficiently improves the recognition rate and makes the technique suitable for a big-data multimedia database. The experimental results show that the proposed method can achieve better accuracy than a multi-scale based method can.     

Application of robust face recognition in video surveillance systems

In this paper, we propose a video searching system that utilizes face recognition as searching indexing feature. As the applications of video cameras have great increase in recent years, face recognition makes a perfect fit for searching targeted individuals within the vast amount of video data. However, the performance of such searching depends on the quality of face images recorded in the video signals. Since the surveillance video cameras record videos without fixed postures for the object, face occlusion is very common in everyday video. The proposed system builds a model for occluded faces using fuzzy principal component analysis (FPCA), and reconstructs the human faces with the available information. Experimental results show that the system has very high efficiency in processing the real life videos, and it is very robust to various kinds of face occlusions. Hence it can relieve people reviewers from the front of the monitors and greatly enhances the efficiency as well. The proposed system has been installed and applied in various environments and has already demonstrated its power by helping solving real cases.     

Subclass linear discriminant analysis for video-based face recognition

We present a novel subclass Linear Discriminant Analysis algorithm for feature extraction that copes with the severe pose, expression and illumination changes present in faces extracted from far-field video streams with subjects unconstrained in their motion and uncooperative to the system. Our novelty lies on the efficient automatic generation of subclasses from the gallery faces, by exploiting their different visual appearance and not constrained by their numbers per class. The proposed feature extraction algorithm is integrated in our complete face recognition system, with modules for preprocessing, classification, and decision fusion. We demonstrate the capability of the new algorithm to automatically generate discriminable subclasses and the resulting improved classification accuracy on a challenging video-based dataset, comprising low quality and resolution faces, as well as large variations in visual appearance. Our results indicate superior recognition rate compared to any systems in the CLEAR 2007 evaluation, running on that dataset.     

基于云计算的视频推荐系统的设计

通过对现有视频网站的调查研究,发现大部分都存在信息过载的问题。所以对视频网站来说拥有推荐系统是有必要的。通过对现有视频推荐系统的分析研究,利用开源云计算技术—Hadoop,及其部分相关组件Hive、Hbase等,设计了一种基于云计算的个性化视频推荐系统,此系统仅适用于以专业视频为主的网站。     

基于视频的天气类别识别系统

针对目前尚未有基于视频的天气类别自动识别系统，本文设计并实现了基于视频的天气类别自动识别系统框架。对于给定的一个测试视频，该系统根据其场景是否与现有的场景相同，选择相应的分类器进行识别，最终输出对应的天气类别识别结果。     

Diverse Features Fusion Network for video-based action recognition

The two-stream convolutional network has been proved to be one milestone in the study of video-based action recognition. Lots of recent works modify internal structure of two-stream convolutional network directly and put top-level features into a 2D/3D convolution fusion module or a simpler one. However, these fusion methods cannot fully utilize features and the way fusing only top-level features lacks rich vital details. To tackle these issues, a novel network called Diverse Features Fusion Network (DFFN) is proposed. The fusion stream of DFFN contains two types of uniquely designed modules, the diverse compact bilinear fusion (DCBF) module and the channel-spatial attention (CSA) module, to distill and refine diverse compact spatiotemporal features. The DCBF modules use the diverse compact bilinear algorithm to fuse features extracted from multiple layers of the base network that are called diverse features in this paper. Further, the CSA module leverages channel attention and multi-size spatial attention to boost key information as well as restraining the noise of fusion features. We evaluate our three-stream network DFFN on three public challenging video action benchmarks: UCF101, HMDB51 and Something-Something V1. Experiment results indicate that our method achieves state-of-the-art performance.     

Augmented two stream network for robust action recognition adaptive to various action videos

In video-based action recognition, using videos with different frame numbers to train a two-stream network can result in data skew problems. Moreover, extracting the key frames from a video is crucial for improving the training and recognition efficiency of action recognition systems. However, previous works suffer from problems of information loss and optical-flow interference when handling videos with different frame numbers. In this paper, an augmented two-stream network (ATSNet) is proposed to achieve robust action recognition. A frame-number-unified strategy is first incorporated into the temporal stream network to unify the frame numbers of videos. Subsequently, the grayscale statistics of the optical-flow images are extracted to filter out any invalid optical-flow images and produce the dynamic fusion weights for the two branch networks to adapt to different action videos. Experiments conducted on the UCF101 dataset demonstrate that ATSNet outperforms previously defined methods, improving the recognition accuracy by 1.13%.     

基于视频的人脸验证

本文提出了一种基于视频的人脸验证方法.采用立体视觉方法初步将人脸区域与背景分割开,再根据多关联模板匹配方法精确定位人脸.对定位后的人脸区域抽取特征器官位置,再依此裁剪出人脸样本.从视频流中收集人脸样本,训练支持向量机(SVM)作为验证器.实验表明该方法在复杂的现场环境下是有效的、鲁棒的.     

具有遮挡鲁棒性的监控视频人脸再识别算法

传统身份识别技术需要将待识别人员信息预先录入，同时未考虑识别过程中的遮挡问题，不能满足公共场所基于监控视频的再识别需求。现有行人再识别算法多依赖于服饰等外观特征，难以进行长期追踪与再识别。针对以上问题，本文提出了一种对遮挡具有鲁棒性的人脸再识别算法。首先，对监控视频中的人脸进行检测与对齐，并判断人脸中存在的遮挡位置；其次，根据遮挡位置查找掩码字典并选择对应掩码，再用掩码排除遮挡元素；最后，使用注意力机制对多帧图片分配权重以更新特征，再使用分区域匹配方法得到识别结果。为验证该方法的有效性，本文分别在COX数据集和人工合成遮挡的数据集上对所提方法进行了测试。其中，在COX数据集上的rank-1准确率为95.2%， 在合成遮挡的数据集上rank-1准确率为73.0%，相比现有方法有明显优势。