基于径向小波神经网络的混沌系统鲁棒自适应反演控制

设计了一种具有自适应性和鲁棒性的反演控制律, 实现了对含有系统不确定性的类Rossler系统的控制. 首先通过小波神经网络辨识系统的非线性部分, 将系统转化为含有结构不确定性和参数不确定性的参数化模型; 然后, 对于系统中的参数不确定性, 设计自适应控制律, 在线估计未知参数; 对于系统中的结构不确定性, 设计鲁棒控制律, 使得系统具有鲁棒性. 最后, 通过仿真实现, 验证了以上控制方法的有效性.     

De Broglie Tired Light Model and the Reality of the Quantum Waves

In the early 1960s of the 20th century de Broglie was able to explain the cosmological observable red shift, without ad hoc assumptions. Starting from basic quantum considerations he developed his tired light model for the photon. This model explains in a single and beautiful causal way the cosmological redshift without need of assuming the Big Bang and consequently a beginning for the universe. Evidence coming from Earth sciences seems also to confirm these ideas and furthermore concrete proposal of laboratorial scale experiments that can test the model are reviewed.     

基于平稳小波和相空间重构的激光混沌预测

提出了一种激光混沌时间序列预测算法.该算法通过平稳小波分解，将原始数据序列分解为与原序列等长的尺度系数和小波系数，利用坐标延迟理论，重建各级尺度系数和各级小波系数的相空间,再根据混沌吸引子的稳定性和分形性,在相空间中对尺度系数和小波系数进行预测,进而通过平稳小波重构算法,实现了时间序列的非线性预测.该算法对数据可以进行更平滑的处理，比无小波算法预测的时间范围更长.通过仿真试验说明，原始时间数据序列被成功的重建，说明算法能够有效的对非线性动态系统的时间序列进行建模和预测.     

适于遥感图像实时压缩的小波基的选择

遥感图像相关性弱,纹理细节丰富,故对其采用基于小波变换的图像编码方法进行编码压缩,而在对图像进行小波变换时,小波基的选取是至关重要的,它直接影响到变换速度和编码效率。在详细分析小波基的基本性质及其与图像编码的关系的基础上,选出数种典型小波基进行实验比较。实验结果表明D5/3双正交小波基最适合于遥感图像实时压缩,最后经实验验证其选取是合理的。     

On the scale-to-scale coupling between a full-scale wind turbine and turbulence

The scale-dependent response of an instrumented full-scale wind turbine is studied under neutrally stratified conditions. The analysis is focused on the linkage between the incoming flow, turbine power output and foundation strain. Wind speed, measured from sonic anemometers installed on a meteorological tower, and foundation strain were sampled at 20 Hz, while the turbine power was sampled at 1 Hz. A wavelet framework and structure function are used to obtain cross correlations among flow turbulence, turbine power and strain across scales as well as to quantify intermittent signatures in both flow and turbine quantities. Results indicate that correlation between the streamwise velocity component of the wind flow and turbine power is maximised across all scales larger than the rotor radius for wind measured at the turbine hub height. The characteristic time lag associated with maximum correlation is shown to be consistent with the Taylor’s hypothesis for turbulent scales smaller than the separation between the meteorological tower and the turbine. However, it decreases with increasing scale size and diminishes to zero at scales on the order of the boundary layer thickness. Turbine power and strain fluctuations exhibited practically the same behaviour at scales larger than two rotor diameters. At those scales, the cross correlation between these quantities resulted ～0.99 and remains still over 0.9 at the scale of rotor radius. Below this scale, the correlation decreases logarithmically with scale. The strong linkage between power and strain for all the relevant scales would eventually allow the analysis of dynamic forcing on the foundation based on the power output. Intermittency on the flow is shown to be transferred and amplified by the turbine, leading to highly intermittent power output.     

微分Haar-Gaussian小波变换在大景深物体边缘测量中的应用

提出了微分Haar_Gaussian小波变换算法并将其应用于边缘检测,在空域和频域中进行变换,并应用带宽匹配方法和远心测量光学系统,提高了运算速度,通过对实际大景深物体离焦像的边缘测量,得到了准确的结果。     

基于小波分解系数的贝叶斯人脸识别方法

本文给出了贝叶斯人脸识别方法中匹配准则的多个近似表达式及一种实用的快速计算方法,在此基础上,利用反对称双正交小波变换的微分算子功能,提出了一种利用两幅人脸图像的小波变换系数差作为模式矢量的贝叶斯人脸识别方法,并利用AR人脸图象库进行了实验,实验结果表明本文方法与基于图像灰度的类似方法相比,识别率提高8％左右,此外本文方法也提供了一条在图像压缩数据域中实现人脸识别的可能途径。     

红外运动小目标的检测

通过分析天空背景下红外运动小目标、噪音以及背景的特征,提出一种检测方法·首先利用向量小波变换对运动图像进行预处理;其次采用图像差分进行目标的粗检测,提取出候选目标;最后可根据运动目标和噪音的特征对候选目标进行识别,检测出真实的运动小目标·实验证明,该方法可有效检测天空背景下红外运动小目标·     

提升小波变换与分形相结合的图像压缩

提出了一种提升小波变换与分形相结合的图像压缩方案.充分利用小波变换后系数能量的分布特性,对提升小波变换后的低频部分采用改进的分形图像压缩编码,其余部分采用集合分裂嵌入块(SPECK)编码算法.试验结果表明,该方法在提高了压缩效率的同时,获得了较高的恢复图像质量.     

基于HVS的小波域信息隐藏方法

为了提高信息隐藏的不可感知性和鲁棒性,结合人眼视觉系统(HVS)的相关知识,提出了一种新的小波域信息隐藏算法。将载体图像的中等亮度区域设为不可用区域,不在该区域隐藏信息,同时选择在载体图像小波变换的高频系数细节子图隐藏信息,并优先选择在对角高频系数细节子图中隐藏。计算局部子块的方差值作为隐藏信息的嵌入系数,并通过计算高频子图的噪声敏感模型ωθ(i,j)决定隐藏信息的位置。充分考虑了人眼的各种视觉掩蔽效应,使得隐藏信息后的结果图像相比隐藏之前没有出现明显的降质。实验结果表明该方法能够有效提高信息隐藏的不可知性,达到了很好的隐藏效果,同时能够有效地抗击噪声、裁剪和压缩等攻击影响,具有很好的鲁棒性,在信息安全领域有较好的应用前景。