改进的CAGAN在虚拟试衣中的应用

针对CAGAN（Conditional Analogy GAN）换衣后效果模糊，在目标衣服与原始衣服长短不一致时效果一般，相对目标衣服保留过少的细节等问题做了相关研究并对CAGAN进行了改进，提出了新的虚拟试衣方式。经过改进的CAGAN生成一个粗糙的结果，由该结果得到目标衣服穿在模特身上改变形状后的mask，接下来利用mask对目标衣服进行变形，综合变形的衣服和第一步的结果便得到最终的试衣图像。实验结果表明，该方法解决了前面存在的问题，而且取得了非常好的效果。     

No-reference quality assessment of dynamic sports videos based on a spatiotemporal motion model

This paper proposes an approach to improve the performance of no-reference video quality assessment for sports videos with dynamic motion scenes using an efficient spatiotemporal model. In the proposed method, we divide the video sequences into video blocks and apply a 3D shearlet transform that can efficiently extract primary spatiotemporal features to capture dynamic natural motion scene statistics from the incoming video blocks. The concatenation of a deep residual bidirectional gated recurrent neural network and logistic regression is used to learn the spatiotemporal correlation more robustly and predict the perceptual quality score. In addition, conditional video block-wise constraints are incorporated into the objective function to improve quality estimation performance for the entire video. The experimental results show that the proposed method extracts spatiotemporal motion information more effectively and predicts the video quality with higher accuracy than the conventional no-reference video quality assessment methods.     

马尔可夫跳变线性系统最优控制的研究现状与进展

马尔可夫跳变线性系统（MJLS）是一种具有多个模态的随机系统，系统在各个模态之间的跳变转移由一组马尔可夫链来决定。MJLS模型因其在表示过程中可以产生突变而更能精确的描述实际工程应用中的系统。近年来，MJLS的最优控制问题成为了研究的热点，动态规划、极大值原理以及线性矩阵不等式等成为了解决此类问题的主流方法。本文对MJLS最优控制领域的研究现状进行了综述。分别对一般情况下、带有噪声的情况下、带有时滞的情况下以及某些特定情况下的MLJS最优控制问题的国内外研究现状进行论述。最后进行了总结并提出MJLS最优控制领域未来值得关注的研究方向。     

基于深度学习的彩色以及近红外图像去马赛克

单传感器捕获的彩色-近红外（RGB-NIR）图像存在光谱干扰，从而导致重建出的标准彩色图像（RGB）图像与近红外（NIR）图像存在色彩失真以及细节信息模糊。针对这个问题提出一种基于深度学习的去马赛克方法，通过引入跳远连接与稠密连接解决了梯度消失和梯度弥散问题，使得网络更容易训练，并且提升了网络的拟合能力。首先，用浅层特征提取层提取了马赛克图像的像素相关性以及通道相关性等低级特征；然后，将得到的浅层特征图输入到连续多个的残差稠密块以提取专门针对去马赛克的高级语义特征；其次，为充分利用低级特征与高级特征，将多个残差稠密块提取到的特征进行组合；最后，通过全局跳远连接恢复最终的RGB-NIR图像。在深度学习框架Tensorflow上使用公共的图像与视觉表示组（IVRG）数据集、有植被的户外多光谱图像（OMSIV）数据集和森林（Forest）三个公开数据集进行实验。实验结果表明，所提方法优于基于多级自适应残差插值、基于卷积卷积和神经神经网络以及基于深度残差U型网络的主流的RGB-NIR图像去马赛克方法。     

Asynchronous control for two‐dimensional hidden Markovian jump systems with partly known mode observation conditional probabilities

This study focuses on the asynchronous control problem for two‐dimensional discrete‐time hidden Markovian jump systems where the mode observation conditional probability matrix is partly known. Considering the original system modes are invisible, the observed modes emitted from an observer serve as an alternative for stability analysis and controller design where a mode observation conditional probability matrix is constructed to characterize the emission between system modes and observed modes. Specially, only partly known information of the mode observation conditional probability matrix is accessible. With the introduction of the free‐connection weighting matrices, the asymptotic mean square stability criterion is firstly derived based on Lyapunov method. This introduction provides a further degree of relaxation and less conservatism is therefore achieved. Secondly, we present synthesis conditions for asynchronous state feedback controller design given in terms of a set of interconnected linear matrix inequalities. Moreover, cluster concept based on the partitions of observed modes is adopted which helps to decrease the number of controllers and simplify the design complexity. A numerical example, regarding the cases with and without clustering of the observed modes, is presented to illustrate the effectiveness of the proposed method.     

Asynchronous adaptive output feedback sliding mode control for Takagi-Sugeno fuzzy Markovian jump systems with actuator faults

In this article, the problem of asynchronous adaptive dynamic output feedback sliding mode control (SMC) for a class of Takagi-Sugeno (T-S) fuzzy Markovian jump systems (MJSs) with actuator faults is investigated. The asynchronous dynamic output feedback control strategy is employed, as the nonsynchronization phenomenon of jump modes exists between the plant and the controller. A novel asynchronous adaptive SMC approach is proposed to solve the synthesis problem for T-S fuzzy MJSs with actuator faults. Sufficient conditions for stochastic asymptotic stability of T-S fuzzy MJSs are given. Under the designed asynchronous adaptive SMC scheme, the effects of actuator faults and external disturbance can be completely compensated and the reachability of sliding surface is ensured. Finally, an example is provided to demonstrate the effectiveness of the proposed design techniques.     

基于改进CGANs的入侵检测方法研究

近年来,机器学习算法在入侵检测系统(IDS)中的应用获得越来越多的关注。然而,传统的机器学习算法更多的依赖于已知样本,因此需要尽可能多的数据样本来对模型进行训练。遗憾地是,随着越来越多未知攻击的出现,且用于训练的攻击样本具有不平衡性,传统的机器学习模型会遇到瓶颈。文章提出一种将改进后的条件生成对抗网络(CGANs)与深度神经网络(DNN)相结合的入侵检测模型(CGANs-DNN),通过解决样本不平衡性问题来提高检测模型对未知攻击类型或只有少数攻击样本类型的检测率。深度神经网络(DNN)具有表征数据潜在特征的能力,而经过改进后的条件CGANs,能够通过学习已知攻击样本潜在数据特征分布,来根据指定类型生成新的攻击样本。此外,与生成对抗网络(GANs)和变分自编码器(VAE)等无监督生成模型相比,CGANsDNN经过改进后加入梯度惩罚项,在训练的稳定性上有了很大地提升。通过NSL-KDD数据集对模型进行评估,与传统算法相比CGANs-DNN不仅在整体准确率、召回率和误报率等方面有更好的性能,而且对未知攻击和只有少数样本的攻击类型具有较高的检测率。     

An iterative approach for the discrete‐time dynamic control of Markov jump linear systems with partial information

The , and mixed dynamic output feedback control of Markov jump linear systems in a partial observation context is studied through an iterative approach. By partial information, we mean that neither the state variable x(k) nor the Markov chain θ(k) are available to the controller. Instead, we assume that the controller relies only on an output y(k) and a measured variable coming from a detector that provides the only information of the Markov chain θ(k). To solve the problem, we resort to an iterative method that starts with a state‐feedback controller and solves at each iteration a linear matrix inequality optimization problem. It is shown that this iterative algorithm yields to a nonincreasing sequence of upper bound costs so that it converges to a minimum value. The effectiveness of the iterative procedure is illustrated by means of two examples in which the conservatism between the upper bounds and actual costs is significantly reduced.     

一类两性分枝过程条件均值增长率的极限性质

本文主要研究随机环境中配对依赖人口数两性 Galton-Watson 分枝过程的条件均值增长率的极限性质．利用上可加函数的性质，得到配对单元平均增长率的极限性质和该过程条件均值的上界和下界．文中给出了关于过程条件均值增长率的两个序列，利用配对单元平均增长率的性质，获得了这两个序列的极限性质．随机环境中配对依赖人口数两性分枝过程比较复杂，本文的结论推广了现有的研究成果．     

Stochastic stabilization and destabilization of nonlinear and time‐varying hybrid systems by noise

The aim of this article is to design a suitable strength function g(t,x,r(t)) such that the Wiener noise g(t,x(t),r(t))dw(t) either stabilizes or destabilizes a given nonlinear and time‐varying hybrid system . To this end, the basic properties, including the existence and uniqueness of the local and global solutions and the nonzero property of solutions of the nonlinear and time‐varying hybrid stochastic systems, are first investigated as the theoretical basis of the article. Second, two theorems and the corresponding corollaries on the stability and instability of the hybrid stochastic systems are established. Third, the design method for the noise strength g(t,x,r(t)) is then proposed based on the established theorems. We also point out that the Markov jump r(t) may have a stabilizing (respectively, destabilizing) effect when we design the noise strength g(t,x,r(t)) so that the introduced noise g(t,x(t),r(t))dw(t) stabilizes (respectively, destabilizes) the corresponding hybrid system. Finally, we illustrate our method using two examples. Compared with the existing literature, our method is suitable for a wider class of nonlinear and time‐varying systems with weaker conditions than quasi‐linear systems.