无下采样轮廓波变换纹理图像检索系统

针对基本轮廓波变换纹理检索系统检索率较低的问题,提出了一种无下采样轮廓波变换(NSCT)纹理图像检索系统.该系统采用的轮廓波变换由无下采样拉普拉斯金字塔级联无下采样方向滤波器构成,特征向量采用子带系数的能量和标准偏差连接而成;以Canberra距离为相似度度量标准.比较了基于同样架构的基本轮廓波变换和NSCT纹理检索系统的性能.实验结果表明:在特征向量长度,检索时间、所需存储空间基本相同的情况下,NSCT检索系统比基本轮廓波变换检索系统具有更高的检索率;NSCT分解结构参数以及图像类型对于平均检索率也有较大的影响.     

基于拉普拉斯金字塔的图像压缩与重构研究

目的研究利用深度学习辅以拉普拉斯金字塔来完成图像压缩与重构。方法利用卷积神经网络提取图像的主要特征,利用双三线性插值法来减少特征尺寸,使用拉普拉斯金字塔来构建分层体系,从而逐步地减少图像大小以达到压缩的目的。在重构端上,对此系统则进行卷积操作,并采用上采样过程,进行图像的恢复重构过程,得到重构图。结果采用来自法国贝尔实验室的set 5与set 14数据集进行验证,使用2层金字塔即在16倍的高倍率压缩下进行实验结果验证,结果表明在主观评价上使用深度学习的方法在清晰度和还原度上要优于PCA,DCT和SVD,同时在客观评价上文中方法取得了标准差(52.73)与信息熵(7.44)的最好结果,高于PCA的49.70与7.38。SVD变换法与DCT变换法,在标准差上只有48.69和49.02,远不如文中方法,同时图片的信息熵只有7.34与7.35,低于文中的7.44。结论利用拉普拉斯金字塔结构来设计卷积神经网络结构来完成图像压缩与重构取得了不错的效果。     

Tetrolet域卫星云图分块压缩感知(英文)

针对卫星云图数据量大,但传输通道和存储空间相对狭小的问题,本文提出了一种基于Tetrolet变换的卫星云图分块压缩感知方法。该方法将Tetrolet变换引入压缩感知的稀疏表示环节,以刻画卫星云图细节丰富,纹理复杂的特性,而且将分块压缩感知与平滑投影Landweber迭代方法结合用于云图重构,以提高计算效率。同时,为了进一步提高重构云图的质量,本文对云图的稀疏表示提出了另一种改进方案,首先对原始云图进行拉普拉斯金字塔分解,将得到的低频分量和高频分量分别进行分块及采样,并对低频及高频分量分别进行离散小波变换(DWT)及Tetrolet变换以实现稀疏表示,此不仅可以发挥不同稀疏变换各自的优点,而且充分利用了Tetrolet变换在表示云图方向纹理和边缘等重要信息方面的优势。实验结果表明,在相同采样率下,本文方法的重构结果明显优于直接用Tetrolet,DWT,Contourlet和DCT变换对卫星云图进行稀疏表示的重构结果。     

基于CUDA的点匹配合成算法

纹理合成在计算机动画制作中具有重要地位。为克服传统串行点匹配纹理合成算法效率低下的缺陷，提出一种基于计算统一设备架构（CUDA）的并行合成算法。通过合理安排CPU和GPU之间的数据传输，用GPU进行繁琐耗时的计算，明显地提高了算法效率。     

利用哈尔变换分析机械振动冲击信号

本文首先介绍了哈尔变换的数学基础：哈尔函数和哈尔级数及其特点；接着介绍哈尔交换及其逆变换和快速算法（ＦＨＴ＆ＩＦＨＴ）的实现；最后在波谱分析方面与傅里叶变换作了比较，指出了在分析机械振动冲击信号时，哈尔交换更能突出信号特征，更易识别。     

纹理块图拼接算法研究

为提高纹理块图的拼接质量和速度,基于多分辨率合成算法的思想,用多层纹理图像金字塔来保存样本纹理的结构特征,采用KD树存储结构以加快最佳匹配块的搜索速度;为加强合成过程的约束,采用螺旋顺序逐块合成纹理图像,避免了合成过程中不连续区域的出现。算法较好的保持了样本图的结构特征,在保证纹理合成质量的同时,减少了计算量,提高了块图的纹理合成速度。     

基于拉普拉斯金字塔的数字水印防伪技术

目的为了提高数字水印的不可见性,运用拉普拉斯金字塔对原始图像进行拉普拉斯分解后使用DCT变换将水印嵌入其拉普拉斯残差图像的中高频。方法对原始图像的Y通道灰度图像进行多层拉普拉斯分解后得到第2层拉普拉斯残差图像,然后通过余弦变换将水印图像嵌入原始图像第2层拉普拉斯残差图像,最后将所有拉普拉斯残差图重建为含水印图像。结果嵌入水印图像的PSNR能够达到40.3 d B,在图像细节上有少许多噪声,并且对一定的图像攻击具有鲁棒性。结论该算法能够改善数字水印技术的不可见性。     

改进型抗混叠轮廓波的图像超分辨率重建

针对Contourlet变换中存在的频谱混叠现象,采用了基于抗混叠轮廓波变换的算法进行图像超分辨率重建.该算法首先用抗混叠塔式滤波器组替换掉Contourlet变换中的拉普拉斯塔式变换,对图像进行尺度变换;然后,根据尺度变换后不同尺度的高频子带之间的相似性,对相关高频分量用双三次插值做高频外推相似变换,再通过方向滤波器...     

一种基于SPIHT和匹配追踪的图像编码方法——SPMP算法

为了更好地利用图像的结构特征,提高图像重建的质量,提出了一种基于多级树集合划分(SPIHT)和匹配追踪(MP)的分层图像编码方法--(SPMP)算法.该方法首先采用拉普拉斯金字塔(Laplacian Pyramid)算法将原始图像分解成低频平滑层和高频细节层,然后使用离散小波变换和SPIHT算法编码图像的低频成分,使用...     

小波变换的实时性应用研究

Ｍａｌｌａｔ塔式分解算法是一个被广泛应用的小波变换快速算法，但运算速度慢，实时性差的缺点阻碍了其在实际应用中的进一步发展。本文介绍了的小波变换快速算法，结合系统的实时怀要求，对Ｍａｌｌａｔ算法进行了优化，将小波与快速傅立叶变换有机结合使图象的实时变换成为可能。在保留图象细节的基础上该算法的运算量比Ｍａｌｌａｔ算法减少了一半以上，并在此基础上设计了该算法的相应硬件，实验结果表明该的确能快速有铲地进行     

A Distributed LRTCO Algorithm in Large-Scale DVE Multimedia Systems

In the large-scale Distributed Virtual Environment (DVE) multimedia systems, one of key challenges is to distributedly preserve causal order delivery of messages in real time. Most of the existing causal order control approaches with real-time constraints use vector time as causal control information which is closely coupled with system scales. As the scale expands, each message is attached a large amount of control information that introduces too much network transmission overhead to maintain the real-time causal order delivery. In this article, a novel Lightweight Real-Time Causal Order (LRTCO) algorithm is proposed for large-scale DVE multimedia systems. LRTCO predicts and compares the network transmission times of messages so as to select the proper causal control information of which the amount is dynamically adapted to the network latency variations and unconcerned with system scales. The control information in LRTCO is effective to preserve causal order delivery of messages and lightweight to maintain the real-time property of DVE systems. Experimental results demonstrate that LRTCO costs low transmission overhead and communication bandwidth, reduces causal order violations efficiently, and improves the scalability of DVE systems.     

超大尺寸激光测距大型测控网络研制

为了解决超大尺寸激光测距过程中所面临的测量任务大、测量点数多、数据庞大和巨型数据处理困难等问题,设计了超大尺寸激光测距大型测控网络的研制方案。提出将云计算平台与大型测控网络相结合,使得若干台具有串口通信功能的仪器协同工作,获得更强大的测量能力,实现了多种无线网络融合,并解决海量数据的存储及数据处理问题。实验结果表明:系统内存设定为2G,虚拟机硬盘设定为500G,当数据存储量超过500G时不会出现不能继续存储的问题,测量数据可以准确地被存储到云平台;一次性处理数据量超过2G时不会出现内存不足的问题,能够实现巨型数据的处理问题;Matlab和虚拟机结合能够很好的完成复杂运算任务,快速准确地确定被测关键点坐标。     

Development of a wind turbine gearbox virtual load sensor using multibody simulation and artificial neural networks

A current development trend in wind energy is characterized by the installation of wind turbines (WT) with increasing rated power output. Higher towers and larger rotor diameters increase rated power leading to an intensification of the load situation on the drive train and the main gearbox. However, current main gearbox condition monitoring systems (CMS) do not record the 6‑degree of freedom (6-DOF) input loads to the transmission as it is too expensive. Therefore, this investigation aims to present an approach to develop and validate a low-cost virtual sensor for measuring the input loads of a WT main gearbox. A prototype of the virtual sensor system was developed in a virtual environment using a multi-body simulation (MBS) model of a WT drivetrain and artificial neural network (ANN) models. Simulated wind fields according to IEC 61400‑1 covering a variety of wind speeds were generated and applied to a MBS model of a Vestas V52 wind turbine. The turbine contains a high-speed drivetrain with 4‑points bearing suspension, a common drivetrain configuration. The simulation was used to generate time-series data of the target and input parameters for the virtual sensor algorithm, an ANN model. After the ANN was trained using the time-series data collected from the MBS, the developed virtual sensor algorithm was tested by comparing the estimated 6‑DOF transmission input loads from the ANN to the simulated 6‑DOF transmission input loads from the MBS. The results show high potential for virtual sensing 6‑DOF wind turbine transmission input loads using the presented method.

     

Steganography Using Reversible Texture Synthesis Based on Seeded Region Growing and LSB

Steganography technology has been widely used in data transmission with secret information. However, the existing steganography has the disadvantages of low hidden information capacity, poor visual effect of cover images, and is hard to guarantee security. To solve these problems, steganography using reversible texture synthesis based on seeded region growing and LSB is proposed. Secret information is embedded in the process of synthesizing texture image from the existing natural texture. Firstly, we refine the visual effect. Abnormality of synthetic texture cannot be fully prevented if no approach of controlling visual effect is applied in the process of generating synthetic texture. We use seeded region growing algorithm to ensure texture’s similar local appearance. Secondly, the size and capacity of image can be decreased by introducing the information segmentation, because the capacity of the secret information is proportional to the size of the synthetic texture. Thirdly, enhanced security is also a contribution in this research, because our method does not need to transmit parameters for secret information extraction. LSB is used to embed these parameters in the synthetic texture.     

System reliability for joint minimal paths under time constraint

The quickest path in a transmission problem for a multistate flow network is considered in this article. In this problem, a given amount of data is transmitted to the destination through multiple minimal paths (MPs) simultaneously, and the probability of complete transmission within a time constraint is of interest. In particular, the MPs are joint such that a constriction occurs. An algorithm is developed by applying Monte Carlo simulation to find the system reliability. Furthermore, the proposed algorithm can be used to manage situations with either joint or disjoint MPs. Although the system reliability obtained from this algorithm is an approximate value, the experiments and inference statistics indicate that the expected value is very close to the actual system reliability.     

Prediction of cold rolling texture of steels using an Artificial Neural Network

We present an Artificial Neural Network based model for the prediction of cold rolling textures of steels. The goal of this work was to design a model capable of fast online prediction of textures in an engineering environment. Our approach uses a feedforward fully interconnected neural network with standard backpropagation error algorithm for configuring the connector weights. The model uses texture data, in form of fiber texture intensities, as well as carbon content, carbide size and amount of rolling reduction as input to the model. The output of the model is in the form of fiber texture data. The available data sets are divided into training and test sets to calibrate and test the network. The predictions of the network provide an excellent match to the experimentally measured data within the bounding box of the training set.     

Decentralized Searching of Multiple Unknown and Transient Radio Sources with Paired Robots

In this paper, we develop a decentralized algorithm to coord inate a group of mobile robots to search for unknown and transient radio sources. In addition to limited mobility and ranges of communication and sensing, the robot team has to deal with challenges from signal source anonymity, short transmission duration, and variable transmission power. We propose a two-step approach: First, we decentralize belief functions that robots use to track source locations using checkpoint-based synchronization, and second, we propose a decentralized planning strategy to coordinate robots to ensure the existence of checkpoints. We analyze memory usage, data amount in communication, and searching time for the proposed algorithm. We have implemented the proposed algorithm and compared it with two heuristics. The experimental results show that our algorithm successfully trades a modest amount of memory for the fastest searching time among the three methods.     

Minimizing data collection for field calibration of steady-state virtual sensors for HVAC equipment

The paper presents an algorithm to minimize the amount of data collection for calibration of steady-state virtual sensors when obtained during normal operation of HVAC equipment in the field. If virtual sensors were calibrated using laboratory data, the amount of the calibration data could be minimized using design of experiment techniques. However, these techniques are not applicable to data from field operating equipment since most of the operating conditions are not controllable (e.g., ambient temperature, loads). In this paper, an algorithm to minimize the data collection period for calibration of steady-state virtual sensors is developed based on sensor accuracy, reliability and applicability. Application of the method is demonstrated for calibration of hourly electricity consumption virtual sensors for 3 packaged air conditioners that are serving a small commercial building, and the method terminated the calibration process after about 3 weeks of data collection for all units with very accurate estimates of electricity consumption.     

A Novel Collective User Web Behavior Simulation Method

A collective user web behavior simulation is an import means for generating a large-scale user network behavior in a network testbed or cyber range. Existing studies almost focus on individual web behavior analysis and prediction, which cannot simulate human dynamics that widely exist in large-scale users’ behaviors. To address these issues, we propose a novel collective user web behavior simulation method, in which an algorithm for constructing a connected virtual social network is proposed, and then a collective user web behavior simulation algorithm is designed on the virtual social network. In the simulation method, a new epidemic information dissemination algorithm based on the SIR model is proposed to drive the user web behavior with Breadth—First Search algorithm on the connected virtual social network. We specially build an experiment environment with 12 servers by using Docker container technology and then perform a wide range of experiments with different user scales to evaluate the method. The experimental results demonstrate that not only the degrees of the social network but also the time intervals of the collective users’ web behavior can be well fitted to a power-law distribution and show that our simulation method can well simulate a collective user web behavior.

     

Multi-VMs Intrusion Detection for Cloud Security Using Dempster-shafer Theory

Cloud computing provides easy and on-demand access to computing resources in a configurable pool. The flexibility of the cloud environment attracts more and more network services to be deployed on the cloud using groups of virtual machines (VMs), instead of being restricted on a single physical server. When more and more network services are deployed on the cloud, the detection of the intrusion likes Distributed Denial-of-Service (DDoS) attack becomes much more challenging than that on the traditional servers because even a single network service now is possibly provided by groups of VMs across the cloud system. In this paper, we propose a cloud-based intrusion detection system (IDS) which inspects the features of data flow between neighboring VMs, analyzes the probability of being attacked on each pair of VMs and then regards it as independent evidence using Dempster-Shafer theory, and eventually combines the evidence among all pairs of VMs using the method of evidence fusion. Unlike the traditional IDS that focus on analyzing the entire network service externally, our proposed algorithm makes full use of the internal interactions between VMs, and the experiment proved that it can provide more accurate results than the traditional algorithm.