Advanced fractal approach for unsupervised classification of SAR images

Unsupervised classification of Synthetic Aperture Radar (SAR) images is the alternative approach when no or minimum apriori information about the image is available. Therefore, an attempt has been made to develop an unsupervised classification scheme for SAR images based on textural information in present paper. For extraction of textural features two properties are used viz. fractal dimension D and Moran’s I. Using these indices an algorithm is proposed for contextual classification of SAR images. The novelty of the algorithm is that it implements the textural information available in SAR image with the help of two texture measures viz. D and I. For estimation of D, the Two Dimensional Variation Method (2DVM) has been revised and implemented whose performance is compared with another method, i.e., Triangular Prism Surface Area Method (TPSAM). It is also necessary to check the classification accuracy for various window sizes and optimize the window size for best classification. This exercise has been carried out to know the effect of window size on classification accuracy. The algorithm is applied on four SAR images of Hardwar region, India and classification accuracy has been computed. A comparison of the proposed algorithm using both fractal dimension estimation methods with the K-Means algorithm is discussed. The maximum overall classification accuracy with K-Means comes to be 53.26% whereas overall classification accuracy with proposed algorithm is 66.16% for TPSAM and 61.26% for 2DVM.     

基于小波神经网络的双足步行机器人步态规划控制器

对于传统的神经网络中神经元模型在结构和信息存储能力上存在的不足，本文提出了一种基于广义小波其函数网络的神经元集聚模型，这种小波神经网络不仅收敛速度快，非线性逼近能力更好，而且具有内部结构变尺度、自适应调整和广义信息存储等智能化特点，更符合生物原型的实际情况。静态学习和准动态学习仿真实验证明这种神经网络结构的有效性。     

基于双重注意力机制及S-BiGAN的机电设备故障诊断

标签样本少条件下机电设备的准确故障诊断对于提高复杂机电设备的健康管理能力具有重要意义。针对标签样本少条件下难以建立准确故障诊断模型的问题,在半监督生成对抗网络的基础上,将注意力模块引入生成对抗网络,并利用格拉姆角场将一维数据转换为二维图像;结合双向生成对抗网络特点,提出一种基于双重注意力机制的半监督双向生成对抗网络（S-BIGAN）机电设备故障诊断模型,以轴承数据为例进行验证。结果表明：与CNN-SVM、SGAN 等算法相比,本文提出的模型能够提高样本生成质量和故障分类特征,有效解决标签样本少情况下的故障诊断问题,极大地提高了故障诊断准确率。     

面向卫星电源系统的一种新颖异常检测方法

面向卫星电源高维周期性时序遥测数据，提出了一种新颖的代表性特征自编码器（RFAE）模型，并用于无监督的异常检测。RFAE采用改进的堆叠自编码器损失函数和训练算法，从而使模型可以学习到相位相同样本的代表性特征；然后根据代表性特征重构样本，根据重构误差来判断样本是否异常。在试验部分首先通过模拟数据校验了RFAE算法能够有效地检测出高维周期性时序数据的异常，然后又采用某卫星电源系统2014年1~12月真实遥测数据进行试验，RFAE异常检测准确率达到99%，检测效果明显优于目前的其他异常检测算法，具有较高应用价值。     

无监督深度学习单目视觉里程计研究

针对传统单目视觉里程计存在的尺度漂移和尺度不一致问题,提出了一种基于无监督深度学习的单目视觉里程计。首先,联合使用空间几何约束和图像相似性约束,得到长序列尺度一致的深度估计网络和鲁棒的光流估计网络;然后,对密集光流进行采样,得到精确的稀疏对应关系,减少尺度漂移;最后,根据改进的ORB-SLAM初始化方法,选择最优跟踪方式,结合深度信息进行尺度对齐,从而得到全局尺度一致的视觉里程计。在KITTI数据集上进行大量实验,结果表明,相较于ORB-SLAM2和基于深度学习的端到端的视觉里程计系统,该算法在通用评估指标方面性能有明显提升,验证了该算法的有效性。     

基于非线性神经网络的高清晰高光谱遥感图像分类器设计与应用

对于高光谱和空间分辨率的遥感图像而言，它具有较为复杂的地物判读特性，应用常规的监督或非监督分类方法难以达到理想的结果。为此设计了一种非线性BP网络分类器，它将纹理结构特征与地物光谱特征相结合．针对上海市某地区的卫星遥感图像．在ENVI／IDL平台上与K-means的非监督分类和最小距离的监督分类方法进行了分类的对比应用。试验结果表明．该分法较好地考虑了图像的光谱特征，能有效地提高类别辨识的精度。     

基于改进Cycle-GAN的光流无监督估计方法CSCD

卷积神经网络为光流的计算提供了一种新的方式,但作为一种数据驱动技术,用于训练网络的大规模光流真值在现实世界中不易获取。为了解决这个弊端,基于Cycle-GAN的循环对抗机制,提出了一种光流无监督估计方法。首先,引入双判别器机制在生成器生成的光流样本的底层和高层特征上进行鉴别,迫使生成器提高光流生成的精度。其次,引入Spynet作为教师网络,在生成器训练前期对其进行指导,防止网络陷入模式崩塌。最后,改进损失函数,提出了光流一致性损失和轮廓一致性损失函数,进一步提升光流估计精度。实验结果表明,与现有的先进算法相比,提出的方法达到与有监督算法相同的精度水平。     

Intrinsic component filtering for fault diagnosis of rotating machinery

Fault diagnosis of rotating machinery has always drawn wide attention. In this paper, Intrinsic Component Filtering (ICF), which achieves population sparsity and lifetime consistency using two constraints: $l_{1/2}$ norm of column features and $l_{3/2}$ -norm of row features, is proposed for the machinery fault diagnosis. ICF can be used as a feature learning algorithm, and the learned features can be fed into the classification to achieve the automatic fault classification. ICF can also be used as a filter training method to extract and separate weak fault components from the noise signals without any prior experience. Simulated and experimental signals of bearing fault are used to validate the performance of ICF. The results confirm that ICF performs superior in three fault diagnosis fields including intelligent fault diagnosis, weak signature detection and compound fault separation.     

超燃冲压发动机燃烧流场纹影图像无监督分割方法

超燃冲压发动机燃烧室流场纹影图像常存在大量噪声信号,如何高效、准确提取燃烧流场图像的主要波系结构成为当前亟需探索的问题。以超燃冲压发动机燃烧室内冷流到氢燃料点火阶段流场为研究对象,基于深度神经网络方法,发展一种燃烧室内流场的关键波系结构快速识别方法。首先,采用基于图论的超像素分割方法对纹影图像进行聚类分割,为语义信息明显相同区域分配伪标签;其次,设计了一种全卷积特征提取神经网络,并使用残差结构对各个通道进行加权,提取纹影图像高级语义特征;最后,使用交叉熵目标函数优化网络模型,并通过阈值滤波操作去除噪声像素点,提升语义分割效果。结果表明：与K-means及自适应高斯阈值方法相比,本文提出方法在准确率、召回率、F1分数和交并比指标性能明显是最优的,能够准确完成燃烧流场纹影图像语义分割任务,可以更加清晰地反应流场内的主要波系和剪切层结构     

An unsupervised classification method of flight states for hypersonic targets based on hyperspectral features

In response to the challenges of aerospace defense caused by the rapid development of hypersonic targets in recent years, the research on the unsupervised classification of flight states for hypersonic targets is carried out in this paper, which is based on the Hyperspectral Features (HFs) of hypersonic targets covered with plasma sheath during high-speed flight. First, a new concept of the super node is defined to improve classification accuracy by alleviating the intraclass variability of HFs. Then, the frequency domain information of the curve of HFs is utilized to reduce the feature redundancy according to the prior theoretical knowledge that the fluctuation characteristics of HFs of the same flight states are similar. Finally, an unsupervised classification method based on the Density Peak Clustering (DPC) for HFs is designed to class flight states after eliminating the impact of intraclass variability and feature dimension redundancy. The proposal is compared with the traditional classification algorithms on simulated hyperspectral data sets of typical flight states of the hypersonic vehicle and an actual-observation hyperspectral data set. The results indicate that the performance of our proposal has competitive advantages in terms of Overall Accuracy (OA), Average Accuracy (AA) and Kappa coefficient.