基于体素特征重组网络的三维物体识别

三维物体识别是计算机视觉领域近年来的研究热点,其在自动驾驶、医学影像处 理等方面具有重要的应用前景。针对三维物体的体素表达形式,特征重组卷积神经网络 VFRN 使用了直接连接同一单元中不相邻的卷积层的短连接结构。网络通过独特的特征重组方式,复 用并融合多维特征,提高特征表达能力,以充分提取物体结构特征。同时,网络的短连接结构 有利于梯度信息的传播,加之小卷积核和全局均值池化的使用,进一步提高了网络的泛化能力, 降低了网络模型的参数量和训练难度。ModelNet 数据集上的实验表明,VFRN 克服了体素数据 分辨率低和纹理缺失的问题,使用较少的参数取得了优于现有方法的识别准确率。     

Evaluation of Local Features for Structure from Motion

Multimedia Tools and Applications - Structure from motion (SFM) is an effective approach for reconstructing large-scale 3D scene from multiple images. In this field, many local feature methods have...     

一种基于差分进化的鲁棒音频隐写算法

音频隐写是将秘密信息隐藏到音频载体中,已成为信息隐藏领域的一个研究热点。已有研究大多聚焦最小化隐写失真,却以牺牲隐写容量为代价,且往往被一些常规信号攻击后难以正确提取秘密信息。为此,基于扩频技术,首先,分析了隐写参数（分段隐写强度和分段隐写容量）与不可感知性和鲁棒性的关系,并构建了一种以分段隐写强度、分段隐写容量为自变量,以不可感知性和隐写容量为优化目标,以信噪比为约束条件的音频隐写多目标优化模型;然后,提出了一种基于差分进化的鲁棒音频隐写算法,设计了相应的编码、适应度函数、交叉和变异算子。对比实验结果表明,所提隐写算法能够在保证不可感知性和抗隐写分析能力的前提下达到更好的鲁棒性,可以有效抵御一些常规信号处理攻击。     

基于点迹时空关系的雷达目标航迹识别

雷达目标航迹的识别有助于指挥员判断对方作战意图和作战任务,从而为战场决策提供协助作用,但目前少有文献研 究目标航迹识别问题。 根据雷达检测到的目标物体的时间、距离和方位等时空数据,研究目标航迹识别问题,并提出了一种基 于雷达点迹数据时空关系的目标航迹识别算法(TRST)。 该算法首先对雷达点迹数据进行属性选择,然后挖掘点迹数据在空间 关系上的间距值特征,最后构建递归神经网络进一步捕捉点迹数据的时空关系特征,实现对目标航迹的分类识别。 实验结果表 明,TRST 算法能够有效提高目标航迹识的准确率、精确率、召回率和 F1-Score 性能。     

现实场景中非接触式心率检测方法研究

现有的非接触式心冲击描记术(ballistocardiography,BCG)利用运动追踪捕捉头部微弱运动,存在着光照干扰、运动干扰等问题.针对该问题,提出了一种基于方向可调滤波器的头部微小振动检测方法(steerable filters schematic detection BCG,SD-BCG),以实现非接触式心...     

Path planning algorithm for WCE with joint energy replenishment and data collection based on multi-objective optimization

Considering limited energy of the wireless charging equipment (WCE) in wireless rechargeable sensor network,an energy replenishment strategy and a data collection strategy are designed.On the basis of these,a path planning model for WCE with functions of joint energy replenishment and data collection based on multi-objective optimization is constructed with two optimization objectives,maximizing the total energy utility of WCE and minimizing the average delay of data transmission of all the sensor nodes in the network.To deal with it,a multi-objective ant colony optimization algorithm based on elitist strategy was proposed,where the state transition strategy and the pheromone updating strategy were improved.Then,the Pareto set was obtained in terms of this multi-objective optimization problem.The parameter setting of ant colony algorithm’s effects on the proposed algorithm were analyzed under 20 sensor nodes.50 groups of contrastive experiments show that the average number of energy utilization obtained by ES-MOAC algorithm is 4.53% higher than that of NSGA-II algorithm.The average number of average delay of all node data transmission obtained by ES-MOAC algorithm is 5.12% lower than that of NSGA-II algorithm.     

基于向量表示和标签传播的半监督短文本数据流分类算法

社交网络平台产生海量的短文本数据流,具有快速、海量、概念漂移、文本长度短小、类标签大量缺失等特点.为此,文中提出基于向量表示和标签传播的半监督短文本数据流分类算法,可对仅含少量有标记数据的数据集进行有效分类.同时,为了适应概念漂移,提出基于聚类簇的概念漂移检测算法.在实际短文本数据流上的实验表明,相比半监督分类算法和半监督数据流分类算法,文中算法不仅提高分类精度和宏平均,还能快速适应数据流中的概念漂移.     

联合增强局部最大发生特征和k-KISSME 度量学习的行人再识别

摘要：行人再识别是一种在监控视频中自动搜索行人的重要技术,该技术包含特征表示 和度量学习2 部分。有效的特征表示应对光线和视角变化具有鲁棒性,具有判别性的度量学习 能够提高行人图像的匹配精度。但是,现有的特征大多都是基于局部特征表示或者全局特征表 示,没有很好的集成行人外观的精细细节和整体外观信息且度量学习通常是在线性特征空间进 行,不能高效地利用特征空间中的非线性结构。针对该问题,设计了一种增强局部最大发生的 有效特征表示(eLOMO)方法,可以实现行人图像精细细节和整体外观信息的融合,满足人类视 觉识别机制;并提出一种被核化的KISSME 度量学习(k-KISSME)方法,其计算简单、高效,只 需要对2 个逆协方差矩阵进行估计。此外,为了处理光线和视角变化,应用了Retinex 变换和 尺度不变纹理描述符。实验表明该方法具有丰富和完整的行人特征表示能力,与现有主流方法 相比提高了行人再识别的识别率。     

Robust visual tracking based on convolutional neural network with extreme learning machine

Multimedia Tools and Applications - Recently, deep learning has attracted substantial attention as a promising solution to many problems in computer vision. Among various deep learning...     

E-passport EAC scheme based on Identity-Based Cryptography

Extended Access Control (EAC) is a security mechanism specified to allow only authorized Inspection System (IS) to read sensitive biometric data such as fingerprints from e-passports. Although European Union EAC scheme offers more flexibility than Singapore scheme, there is clearly room for improvement. By adopting Identity-Based Cryptography (IBC) technology, a simple and secure EAC implementation scheme (IBC-EAC) is proposed. The authorization mechanism based on IBC is more trustable because the access right to sensitive data is granted directly to the IS through Authorized Smartcard. A new authentication protocol based on IBC is performed between the e-passport chip and the Authorized Smartcard. The protocol also provides an important contribution towards terminal revocation. By using IBC-EAC scheme, the complexity of deploying and managing PKI can be reduced. And the computational cost for e-passport to verify the certificate chain in EU-EAC scheme can be saved.