基于遗传神经网络的颜色恒常感知计算模型

在机器视觉领域,颜色恒常性是实现计算机视觉颜色校正和保持机器对颜色识别稳定性的重要因素.该模型通过心理物理实验获得由人眼感知得到的颜色恒常感知数据,将其放入神经网络中进行样本训练,并用遗传算法优化BP神经网络的连接权值和阈值.将所建立颜色恒常感知计算模型应用到图像颜色校正,通过主观和客观两个方面对校正结果进行对比评价,结果表示所建立的颜色恒常感知计算模型计算精度和效率高、复杂度低,比几种经典算法处理误差要小,针对图像的颜色再现有着更为符合人眼感知的特性.     

卷积神经网络模型剪枝结合张量分解压缩方法

针对卷积神经网络（CNN）拥有巨大的参数量及计算量，限制了其在嵌入式系统等资源受限设备上应用的问题，提出了基于统计量的网络剪枝结合张量分解的神经网络压缩方法，其核心思想是以均值和方差作为评判权值贡献度的依据。首先，以Lenet5为剪枝模型，网络各卷积层的均值和方差分布以聚类方式分离出提取特征较弱的滤波器，而使用保留的滤波器重构下一层卷积层；然后，将剪枝方法结合张量分解对更快的区域卷积神经网络（Faster RCNN）进行压缩，低维卷积层采取剪枝方法，而高维卷积层被分解为三个级联卷积层；最后，将压缩后的模型进行微调，使其在训练集上重新达到收敛状态。在PASCAL VOC测试集上的实验结果表明，所提方法降低了Faster RCNN模型54%的存储空间而精确率仅下降了0.58%，同时在树莓派4B系统上达到1.4倍的前向计算加速，有助于深度CNN模型在资源受限的嵌入式设备上的部署。     

面向群体行为识别的非局部网络模型

目的 视频行为识别一直广受计算机视觉领域研究者的关注，主要包括个体行为识别与群体行为识别。群体行为识别以人群动作作为研究对象，对其行为进行有效表示及分类，在智能监控、运动分析以及视频检索等领域有重要的应用价值。现有的算法大多以多层递归神经网络（RNN）模型作为基础，构建出可表征个体与所属群体之间关系的群体行为特征，但是未能充分考虑个体之间的相互影响，致使识别精度较低。为此，提出一种基于非局部卷积神经网络的群体行为识别模型，充分利用个体间上下文信息，有效提升了群体行为识别准确率。方法 所提模型采用一种自底向上的方式来同时对个体行为与群体行为进行分层识别。首先从原始视频中沿着个人运动的轨迹导出个体附近的图像区块；随后使用非局部卷积神经网络（CNN）来提取包含个体间影响关系的静态特征，紧接着将提取到的个体静态特征输入多层长短期记忆（LSTM）时序模型中，得到个体动态特征并通过个体特征聚合得到群体行为特征；最后利用个体、群体行为特征同时完成个体行为与群体行为的识别。结果 本文在国际通用的Volleyball Dataset上进行实验。实验结果表明，所提模型在未进行群体精细划分条件下取得了77.6%的准确率，在群体精细划分的条件下取得了83.5%的准确率。结论 首次提出了面向群体行为识别的非局部卷积网络，并依此构建了一种非局部群体行为识别模型。所提模型通过考虑个体之间的相互影响，结合个体上下文信息，可从训练数据中学习到更具判别性的群体行为特征。该特征既包含个体间上下文信息、也保留了群体内层次结构信息，更有利于最终的群体行为分类。     

权值初始化与激励函数调整相结合的学习算法

提出了一种基于独立元分析(ICA)方法的权值初始化方法和动态调整S型激励函数的斜率相结合的神经网络学习算法。该方法利用ICA从输入数据中提取显著的特征信息来初始化输入层到隐含层权值。而且通过使神经网络的输出位于激励函数的活动区域,对隐含层到输出层的权值进行初始化。在学习过程中,再对每个隐单元和输出单元的激励函数的斜率进行自动调整。最后通过计算机仿真实际的基准问题,验证了论文提出的方法的有效性。实验结果表明,所提出的方法能有效地加快多层前向神经网络的训练过程。     

基于卷积神经网络的图像分类模型综述

卷积神经网络在计算机视觉等领域占有一席之地，利用局部连接、权值共享以及池化操作等特性，有效地提取图像的局部特征，降低网络复杂度，具有更少的参数量和更好的鲁棒性，因此，吸引了众多研究者的关注，使分类模型朝着更轻、更快、更高效的方向迅速发展。按照卷积神经网络发展的时间线，介绍了常用的典型网络模型，剖析了其创新点与优缺点，并对其未来的发展方向进行了展望。     

面向智能客服的句子相似度计算方法

针对金融领域中智能客服的句子相似度计算方法进行了研究。利用基于词性的分词纠正模型减少中文歧义词、金融相关词汇的分词错误；通过词向量方法和循环神经网络分别提取词语级和句子级的语义特征，并且得到句子向量；用融合层计算出句子向量间的差异特征；对差异特征进行降维和归一化得到句子相似度计算结果。实验结果表明，该方法具有较高的准确率和[F1]值。     

基于布谷鸟搜索优化BP神经网络的网络安全态势评估方法

针对现有基于神经网络的网络安全态势评估方法效率低等问题,提出基于布谷鸟搜索（CS）优化反向传播（BP）神经网络（CSBPNN）的网络安全态势评估方法。首先,根据态势输入指标数和输出态势值确定BP神经网络（BPNN）的输入输出节点数,根据经验公式和试凑法计算出隐含层节点数;然后,随机初始化各层的连接权值和阈值,使用浮点数编码方式将权值与阈值编码成布谷鸟;最后,使用CS算法对权值和阈值进行优化,得到用于态势评估的CSBPNN模型并对其进行训练,将网络安全态势数据输入到CSBPNN模型中,获取网络的安全态势值。实验结果表明,与BPNN和遗传算法优化BP神经网络方法相比,基于CSBPNN的网络安全态势评估方法的迭代代数分别减少943和47且预测精度提高8.06个百分点和3.89个百分点,所提方法具有较快的收敛速度和较高的预测精度。     

采用计算机视觉进行孵化鸡蛋成活可能性的自动检测

本研究将计算机视觉技术与遗传神经网络相结合,建立一套适合于孵化鸡蛋可成活性自动检测的计算机视觉系统,通过计算机视觉技术获取了孵化鸡蛋的角度直方图,并提取了孵化鸡蛋表面颜色特征,采用遗传算法优化了多层前馈神经网络的拓扑结构与权值,提高了神经网络的学习质量和学习速度,实现了孵化鸡蛋可成活性的自动检测,实验结果表明,该方法准确率较高,并具有鲁棒性和高速度。     

基于粒计算Web文档聚类

提出了一种基于粒计算Web文档聚类（WDCGrc）方法。该方法通过TF-IDF法则计算文档词条的权值,采取设定文档阈值和平均权值相结合的方法实行降维,抽取出每篇文档的主干词;建立了文档的主干词和二进制粒之间的转换,提出了基于粒计算提取文档间的关联规则算法来获取文档间的频繁项集,由频繁项集形成初始聚类,使用优化算法对初始聚类进行优化,得到最终聚类结果。实验结果表明,该方法切实有效,聚类质量较好。     

一种基于聚类的微博关键词提取方法的研究与实现

文章提出了一种基于聚类的微博关键词提取方法。实验过程分三个步骤进行。第一步,对微博文本进行预处理和分词处理,再运用TF-IDF算法与TextRank算法计算词语权重,针对微博短文本的特性在计算词语权重时运用加权计算的方法,在得到词语权重后使用聚类算法提取候选关键词;第二步,根据n-gram语言模型的理论,取n的值为2定义最大左邻概率和最大右邻概率,据此对候选关键词进行扩展;第三步,根据语义扩展模型中邻接变化数和语义单元数的概念,对扩展后的关键词进行筛选,得到最终的提取结果。实验结果表明在处理短文本时Text Ramk算法比TF-IDF算法表现更佳,同时该方法能够有效地提取出微博中的关键词。     

基于多尺度多任务卷积神经网络的人群计数

在智能监控领域，实现人群计数具有重要价值，针对人群尺度不一、人群密度分布不均及遮挡等问题，提出一种多尺度多任务卷积神经网络（MMCNN）进行人群计数的方法。首先提出一种新颖的自适应人形核生成密度图描述人群信息，消除人群遮挡影响；其次通过构建多尺度卷积神经网络解决人群尺度不一问题，以多任务学习机制同时估计密度图及人群密度等级，解决人群分布不均问题；最后设计一种加权损失函数，提高人群计数准确率。在UCF_CC_50和World Expo'10数据库上进行了评估，验证了自适应人形核的有效性。实验结果表明：所提算法比Sindagi等的方法（SINDAGI V A，PATEL V M.CNN-based cascaded multi-task learning of high-level prior and density estimation for crowd counting.Proceedings of the 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance.Piscataway，NJ：IEEE，2017：1-6）在UCF_CC_50数据库上平均绝对误差（MAE）数值和均方误差（MSE）数值分别降低约1.7和45；与Zhang等的方法（ZHANG Y，ZHOU D，CHEN S，et al.Single-image crowd counting via multi-column convolutional neural network.Proceedings of the 2016 IEEE Conference on Computer Vision and Pattern Recognition.Washington，DC：IEEE Computer Society，2016：589-597）相比，在World Expo'10数据库上所提算法的MAE值降低约1.5，且在真实公共汽车数据库上仅0~3人的计数误差，表明其实用性较强。     

基于深度自编码网络的运动目标检测

针对从动态背景中提取前景效果较差的问题,提出了一种基于深度自编码网络的运动目标检测方法。首先,用一个三层的深度自编码网络从视频图像中提取不包含运动目标的背景图像,将背景图像作为变量构造了深度自编码网络的代价函数;然后,构造了一个分离函数得到了输入图像的背景图像,再用另一个三层的深度自编码网络学习提取出的背景图像;为了使深度自编码网络的学习能够在线地提取运动目标,还提出了一种在线学习算法,通过寻找对代价函数敏感度较低的权重进行合并,从而能够对更多的视频图像进行处理。实验结果表明,所提方法在从动态背景中提取出前景运动目标上相比Lu等的前景检测的工作(LU C, SHI J, JIA J. Online robust dictionary learning. Proceeding of the 2013 IEEE Conference on Computer Vision and Pattern Recognition, Piscataway: IEEE Press, 2013:415-422)检测的准确率提高了6%,并且误报率降低了4.5%。在实际的应用中,能够获得更好的前景背景分离效果,为视频分析等方面的研究奠定更好的基础。     

Guest Editors' Introduction to the Special Section on Award Winning Papers from the IEEE CS Conference on Computer Vision and Pattern Recognition (CVPR)

The three articles in this special section are selected papers from the IEEE CS Conference on Computer Vision and Pattern Recognition that was held in Anchorage, AL, in June 2008.     

Unifying statistical texture classification frameworks

The objective of this paper is to examine statistical approaches to the classification of textured materials from a single image obtained under unknown viewpoint and illumination. The approaches investigated here are based on the joint probability distribution of filter responses.

We review previous work based on this formulation and make two observations. First, we show that there is a correspondence between the two common representations of filter outputs—textons and binned histograms. Second, we show that two classification methodologies, nearest neighbour matching and Bayesian classification, are equivalent for particular choices of the distance measure. We describe the pros and cons of these alternative representations and distance measures, and illustrate the discussion by classifying all the materials in the Columbia-Utrecht (CUReT) texture database.

These equivalences allow us to perform direct comparisons between the texton frequency matching framework, best exemplified by the classifiers of Leung and Malik [Int. J. Comput. Vis. 43 (2001) 29], Cula and Dana [Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2001) 1041], and Varma and Zisserman [Proceedings of the Seventh European Conference on Computer Vision 3 (2002) 255], and the Bayesian framework most closely represented by the work of Konishi and Yuille [Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2000) 125].     

Using Biologically Inspired Features for Face Processing

In this paper, we show that a new set of visual features, derived from a feed-forward model of the primate visual object recognition pathway proposed by Riesenhuber and Poggio (R&P Model) (Nature Neurosci. 2(11):1019–1025, 1999) is capable of matching the performance of some of the best current representations for face identification and facial expression recognition. Previous work has shown that the Riesenhuber and Poggio Model features can achieve a high level of performance on object recognition tasks (Serre, T., et al. in IEEE Comput. Vis. Pattern Recognit. 2:994–1000, 2005). Here we modify the R&P model in order to create a new set of features useful for face identification and expression recognition. Results from tests on the FERET, ORL and AR datasets show that these features are capable of matching and sometimes outperforming other top visual features such as local binary patterns (Ahonen, T., et al. in 8th European Conference on Computer Vision, pp. 469–481, 2004) and histogram of gradient features (Dalal, N., Triggs, B. in International Conference on Computer Vision & Pattern Recognition, pp. 886–893, 2005). Having a model based on shared lower level features, and face and object recognition specific higher level features, is consistent with findings from electrophysiology and functional magnetic resonance imaging experiments. Thus, our model begins to address the complete recognition problem in a biologically plausible way.     

Scheduling real-time requests in on-demand data broadcast environments

On-demand broadcast is an attractive data dissemination method for mobile and wireless computing. In this paper, we propose a new online preemptive scheduling algorithm, called PRDS that incorporates urgency, data size and number of pending requests for real-time on-demand broadcast system. Furthermore, we use pyramid preemption to optimize performance and reduce overhead. A series of simulation experiments have been performed to evaluate the real-time performance of our algorithm as compared with other previously proposed methods. The experimental results show that our algorithm substantially outperforms other algorithms over a wide range of workloads and parameter settings. The work described in this paper was partially supported by grants from CityU (Project No. 7001841) and RGC CERG Grant No. HKBU 2174/03E. This paper is an extended version of the paper “A preemptive scheduling algorithm for wireless real-time on-demand data broadcast” that appeared in the 11th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications. Victor C. S. Lee received his Ph.D. degree in Computer Science from the City University of Hong Kong in 1997. He is now an Assistant Professor in the Department of Computer Science of the City University of Hong Kong. Dr. Lee is a member of the ACM, the IEEE and the IEEE Computer Society. He is currently the Chairman of the IEEE, Hong Kong Section, Computer Chapter. His research interests include real-time data management, mobile computing, and transaction processing. Xiao Wu received the B.Eng. and M.S. degrees in computer science from Yunnan University, Kunming, China, in 1999 and 2002, respectively. He is currently a Ph.D. candidate in the Department of Computer Science at the City University of Hong Kong. He was with the Institute of Software, Chinese Academy of Sciences, Beijing, China, between January 2001 and July 2002. From 2003 to 2004, he was with the Department of Computer Science of the City University of Hong Kong, Hong Kong, as a Research Assistant. His research interests include multimedia information retrieval, video computing and mobile computing. Joseph Kee-Yin NG received a B.Sc. in Mathematics and Computer Science, a M.Sc. in Computer Science, and a Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign in the years 1986, 1988, and 1993, respectively. Prof. Ng is currently a professor in the Department of Computer Science at Hong Kong Baptist University. His current research interests include Real-Time Networks, Multimedia Communications, Ubiquitous/Pervasive Computing, Mobile and Location- aware Computing, Performance Evaluation, Parallel and Distributed Computing. Prof. Ng is the Technical Program Chair for TENCON 2006, General Co-Chair for The 11th International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA 2005), Program Vice Chair for The 11th International Conference on Parallel and Distributed Systems (ICPADS 2005), Program Area-Chair for The 18th & 19th International Conference on Advanced Information Networking and Applications (AINA 2004 & AINA 2005), General Co-Chair for The International Computer Congress 1999 & 2001 (ICC’99 & ICC’01), Program Co-Chair for The Sixth International Conference on Real-Time Computing Systems and Applications (RTCSA’99) and General Co-Chair for The 1999 and 2001 International Computer Science Conference (ICSC’99 & ICSC’01). Prof. Ng is a member of the Editorial Board of Journal of Pervasive Computing and Communications, Journal of Ubiquitous Computing and Intelligence, Journal of Embedded Computing, and Journal of Microprocessors and Microsystems. He is the Associate Editor of Real-Time Systems Journal and Journal of Mobile Multimedia. He is also a guest editor of International Journal of Wireless and Mobile Computing for a special issue on Applications, Services, and Infrastructures for Wireless and Mobile Computing. Prof. Ng is currently the Region 10 Coordinator for the Chapter Activities Board of the IEEE Computer Society, and is the Coordinator of the IEEE Computer Society Distinguished Visitors Program (Asia/Pacific). He is a senior member of the IEEE and has been a member of the IEEE Computer Society since 1991. Prof. Ng has been an Exco-member (1993–95), General Secretary (1995–1997), Vice-Chair (1997–1999), Chair (1999–2001) and the Past Chair of the IEEE, Hong Kong Section, Computer Chapter. Prof. Ng received the Certificate of Appreciation for Services and Contribution (2004) from IEEE Hong Kong Section, the Certificate of Appreciation for Leadership and Service (2000–2001) from IEEE Region 10 and the IEEE Meritorious Service Award from IEEE Computer Society at 2004. He is also a member of the IEEE Communication Society, ACM and the Founding Member for the Internet Society (ISOC)-Hong Kong Chapter.     

Stel Component Analysis: Joint Segmentation, Modeling and Recognition of Objects Classes

Models that captures the common structure of an object class have appeared few years ago in the literature (Jojic and Caspi in Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), pp. 212?C219, 2004; Winn and Jojic in Proceedings of International Conference on Computer Vision (ICCV), pp. 756?C763, 2005); they are often referred as ??stel models.?? Their main characteristic is to segment objects in clear, often semantic, parts as a consequence of the modeling constraint which forces the regions belonging to a single segment to have a tight distribution over local measurements, such as color or texture. This self-similarity within a region in a single image is typical of many meaningful image parts, even when across different images of similar objects, the corresponding parts may not have similar local measurements. Moreover, the segmentation itself is expected to be consistent within a class, although still flexible. These models have been applied mostly to segmentation scenarios. In this paper, we extent those ideas (1) proposing to capture correlations that exist in structural elements of an image class due to global effects, (2) exploiting the segmentations to capture feature co-occurrences and (3) allowing the use of multiple, eventually sparse, observation of different nature. In this way we obtain richer models more suitable to recognition tasks. We accomplish these requirements using a novel approach we dubbed stel component analysis. Experimental results show the flexibility of the model as it can deal successfully with image/video segmentation and object recognition where, in particular, it can be used as an alternative of, or in conjunction with, bag-of-features and related classifiers, where stel inference provides a meaningful spatial partition of features.     

Bi-Real Net: Binarizing Deep Network Towards Real-Network Performance

International Journal of Computer Vision - In this paper, we study 1-bit convolutional neural networks (CNNs), of which both the weights and activations are binary. While being efficient, the...     

A novel online boosting algorithm for automatic anatomy detection

This paper presents a novel online learning method for automatically detecting anatomic structures in medical images. Conventional off-line learning methods require collecting a complete set of representative samples prior to training a detector. Once the detector is trained, its performance is fixed. To improve the performance, the detector must be completely retrained, demanding the maintenance of historical training samples. Our proposed online approach eliminates the need for storing historical training samples and is capable of continually improving performance with new samples. We evaluate our approach with three distinct thoracic structures, demonstrating that our approach yields performance competitive with the off-line approach. Furthermore, we investigate the properties of our proposed method in comparison with an online learning method suggested by Grabner and Bischof (IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2006, vol. 1, pp. 260–267, 2006), which is the state of the art, indicating that our proposed method runs faster, offers more stability, improves handling of “catastrophic forgetting”, and simultaneously achieves a satisfactory level of adaptability. The enhanced performance is attributed to our novel online learning structure coupled with more accurate weaker learners based on histograms.     

A review of deformable surfaces: topology, geometry and deformation

Deformable models have raised much interest and found various applications in the fields of computer vision and medical imaging. They provide an extensible framework to reconstruct shapes. Deformable surfaces, in particular, are used to represent 3D objects. They have been used for pattern recognition [Computer Vision and Image Understanding 69(2) (1998) 201; IEEE Transactions on Pattern Analysis and Machine Intelligence 19(10) (1997) 1115], computer animation [ACM Computer Graphics (SIGGRAPH'87) 21(4) (1987) 205], geometric modelling [61][Computer Aided Design (CAD) 24(4) (1992) 178], simulation [Visual Computer 16(8) (2000) 437], boundary tracking [ACM Computer Graphics (SIGGRAPH'94) (1994) 185], image segmentation [Computer Integrated Surgery, Technology and Clinical Applications (1996) 59; IEEE Transactions on Medical Imaging 14 (1995) 442; Joint Conference on Computer Vision, Virtual Reality and Robotics in Medicine (CVRMed-MRCAS'97) 1205 (1997) 13; Medical Image Computing and Computer-Assisted Intervention (MICCAI'99) 1679 (1999) 176; Medical Image Analysis 1(1) (1996) 19], etc. In this paper we propose a survey on deformable surfaces. Many surface representations have been proposed to meet different 3D reconstruction problem requirements. We classify the main representations proposed in the literature and we study the influence of the representation on the model evolution behavior, revealing some similarities between different approaches.