拉普拉斯约束低秩表示的高光谱图像异常检测

伴随高光谱图像的广泛使用,高光谱图像技术得到长足的发展,其中高光谱图像异常检测技术越发受到重视。为了解决传统高光谱图像异常检测技术的实用性和检测效果不佳的问题,提出一种新颖的低秩表示检测算法。对于高光谱图像,大部分背景像元均可以被少量主要的背景像元组合近似地表示,且它们的表示系数将会位于低秩的空间中。在剩下无法被主要背景像元表示的稀疏部分中存在着异常像元,则可以被检测算法提取出来。在低秩表示中,背景像元字典的构建将会影响高光谱图像中背景像元的表示。如直接从现有高光谱图像中提取背景像元构建字典,会导致异常像元对背景像元字典的污染。而利用待检测高光谱图像观测数据和由光谱组成原理可合成的潜在未观测数据来构建背景像元字典,提取出背景像元的主要特征,有利于更好地分离出稀疏异常像元的信息。并且高光谱图像数据存在高维几何结构特点,通过引入拉普拉斯矩阵来约束空间中局部相似的像元对于待检测像元的表示作用,获得更接近于真实的表示系数。实验结果分别在仿真数据和真实数据上验证,与传统方法相比,提出的方法通过有效地突出异常像元提高了检出率和抑制了背景像元,降低了误检率。     

GTAD: Graph and Temporal Neural Network for Multivariate Time Series Anomaly Detection

The rapid development of smart factories, combined with the increasing complexity of production equipment, has resulted in a large number of multivariate time series that can be recorded using sensors during the manufacturing process. The anomalous patterns of industrial production may be hidden by these time series. Previous LSTM-based and machine-learning-based approaches have made fruitful progress in anomaly detection. However, these multivariate time series anomaly detection algorithms do not take into account the correlation and time dependence between the sequences. In this study, we proposed a new algorithm framework, namely, graph attention network and temporal convolutional network for multivariate time series anomaly detection (GTAD), to address this problem. Specifically, we first utilized temporal convolutional networks, including causal convolution and dilated convolution, to capture temporal dependencies, and then used graph neural networks to obtain correlations between sensors. Finally, we conducted sufficient experiments on three public benchmark datasets, and the results showed that the proposed method outperformed the baseline method, achieving detection results with F1 scores higher than 95% on all datasets.     

基于线阵探测器的单圈绝对轴角编码器

为了解决轴角编码器尺寸与角度分辨率之间的矛盾,同时提高编码器的响应频率,提出了一种基于线阵探测器的单圈绝对式光电轴角编码器.该编码器使用了一种新型单圈绝对式编码盘,整个码盘只有一个码道,粗码被直接刻在这个单圈的码道上,码盘图像经光学放大后被线阵探测器接收.利用FPGA控制电路将数据传送至计算机,并对数据进行译码处理,分别利用图像处理技术读取粗码和利用像素细分技术获得细码,两者相结合得到角度信息.通过该技术设计一个码盘直径为40 mm的绝对式轴角编码器,其分辨率为15位.     

适于干涉多光谱图像压缩的自适应率控制算法

提出了一种基于优化截取内嵌码块编码（EBCOT）的感兴趣区域（ROD编码干涉多光谱图像压缩方法。小波变换后,对1级分解的高频系数感兴趣区域即包含光谱信息区域进行垂直方向的分解,再对感兴趣区域进行比特平面提升。T1编码器对不同比特平面的编码过程（Codingpass）赋予不同的重要性权值,由高到低依次编码,T2编码器根据所得的比特率自适应地反馈控制T1的编码深度,最后进行率失真优化截取。实验结果表明,该方法提高了恢复图像质量,有效地减少了优化截取内嵌码块编码算法的计算量和内存使用量（bpp-1时,测试图像的整体、感兴趣区域和背景区域平均峰值信噪比均提高0．1dB以上,计算量和内存使用量平均减少40％和60％以上）,编码方式适合干涉多光谱图像压缩系统硬件实现。     

基于多尺度分解的超光谱图像异常检测

提出了一种基于多尺度分解的超光谱图像异常检测算法。在目标和背景均未知的前提下,利用光谱和空间两种信息完成对异常目标信号的定位,从而实现超光谱遥感数据中异常目标检测。首先利用非下采样塔式变换对超光谱图像进行分解,将其划分为不同尺度子块;然后依据超光谱图像同一波段不同尺度空间内相邻系数的相关性,采用不同波段各个尺度空间的反锐化掩模方法优化背景数据分布,从而抑制异常数据对背景的干扰;最后利用设计的核RX算子得到异常目标检测结果。为验证方法的有效性,利用真实和模拟的AVIRIS数据进行了实验,并与经典RX算法相比较,实验结果表明,基于非下采样塔式分解的异常检测方法具有更好的检测性能和较低的虚警。     

A low complexity block-based adaptive lossless image compression

For low power and lossless image compression, in this paper, a low complexity, block-based decomposition of subbands technology is proposed for embedded compression (EC) algorithm, which is ready for being implemented on a single-chip of FPGA. The proposed algorithm is based on high-speed pipeline architecture of 2-D lossless integer wavelet transformation (IWT) with 2-D Lossless Hadamard Transformation (LHT). In the proposed algorithm, the coefficients of a 2-D IWT are decomposed by 4 × 4 blocks to further remove redundancy, compared with direct encoder by EBCOT of JPEG2000. Considering the feature of the 2-D IWT, a different strategy is designed for LL-subband and non-LL subbands, which denotes DC prediction (DCP) and adaptive transformation method (ATM), respectively. DCP is used to remove the correlation between two adjacent blocks of LL-subband, and ATM is used to transform non-LL subbands by 2-D LHT selectivity. After further transformation, the coefficients are decomposed as truncated integer part (TIP) and truncated residue parts (TRP), considering the complexity of hardware implementation, TIP is encoded by Zero Running Length (ZRL) and Exp-Golomb (EG). TRP is encoded by a fixed length (FL) encoder after removed redundancy by the feature of 2-D LHT, when seen as bit patterns [1]. Experimental results show that the proposed EC algorithm can achieve a good compression performance as JPEG2000, and the coding latency can be decreased at an average of 43.9%. Another innovation of this paper is EC's hardware-friendly feature and easy hardware implementation, which are presented by a simple addition or subtraction of the LIWT and LHT, and need a small on-chip memory.     

Source Symbol Purging-Based Distributed Conditional Arithmetic Coding

A distributed arithmetic coding algorithm based on source symbol purging and using the context model is proposed to solve the asymmetric Slepian–Wolf problem. The proposed scheme is to make better use of both the correlation between adjacent symbols in the source sequence and the correlation between the corresponding symbols of the source and the side information sequences to improve the coding performance of the source. Since the encoder purges a part of symbols from the source sequence, a shorter codeword length can be obtained. Those purged symbols are still used as the context of the subsequent symbols to be encoded. An improved calculation method for the posterior probability is also proposed based on the purging feature, such that the decoder can utilize the correlation within the source sequence to improve the decoding performance. In addition, this scheme achieves better error performance at the decoder by adding a forbidden symbol in the encoding process. The simulation results show that the encoding complexity and the minimum code rate required for lossless decoding are lower than that of the traditional distributed arithmetic coding. When the internal correlation strength of the source is strong, compared with other DSC schemes, the proposed scheme exhibits a better decoding performance under the same code rate.     

字典学习稀疏表示的高光谱图像异常检测

针对稀疏表示高光谱检测算法性能受背景字典影响较大的问题,充分利用高光谱图像空间信息和光谱主成分信息,提出了一种基于字典学习的稀疏表示异常检测算法。首先利用主成分分析提取高光谱数据的主特征,建立目标主成分空间,并证明了在主成分空间进行字典学习稀疏重构的可行性;然后在主成分空间内构造基于K-SVD算法的训练字典,改善了背景字典性能;采用正交匹配算法重构主成分分量,利用主成分分析反变换得到待检测像元重构光谱,增强了高光谱图像的局部异常特性;最后,基于重构误差异常特性实现高光谱图像异常检测。仿真结果证明了该方法的有效性。     

Unsupervised Anomaly Detection with Distillated Teacher-Student Network Ensemble

We address the problem of unsupervised anomaly detection for multivariate data. Traditional machine learning based anomaly detection algorithms rely on specific assumptions of normal patterns and fail to model complex feature interactions and relations. Recently, existing deep learning based methods are promising for extracting representations from complex features. These methods train an auxiliary task, e.g., reconstruction and prediction, on normal samples. They further assume that anomalies fail to perform well on the auxiliary task since they are never trained during the model optimization. However, the assumption does not always hold in practice. Deep models may also perform the auxiliary task well on anomalous samples, leading to the failure detection of anomalies. To effectively detect anomalies for multivariate data, this paper introduces a teacher-student distillation based framework Distillated Teacher-Student Network Ensemble (DTSNE). The paradigm of the teacher-student distillation is able to deal with high-dimensional complex features. In addition, an ensemble of student networks provides a better capability to avoid generalizing the auxiliary task performance on anomalous samples. To validate the effectiveness of our model, we conduct extensive experiments on real-world datasets. Experimental results show superior performance of DTSNE over competing methods. Analysis and discussion towards the behavior of our model are also provided in the experiment section.     

Automatic metal parts inspection: Use of thermographic images and anomaly detection algorithms

A fully-automatic approach based on the use of induction thermography and detection algorithms is proposed to inspect industrial metallic parts containing different surface and sub-surface anomalies such as open cracks, open and closed notches with different sizes and depths. A practical experimental setup is developed, where lock-in and pulsed thermography (LT and PT, respectively) techniques are used to establish a dataset of thermal images for three different mockups. Data cubes are constructed by stacking up the temporal sequence of thermogram images. After the reduction of the data space dimension by means of denoising and dimensionality reduction methods; anomaly detection algorithms are applied on the reduced data cubes. The dimensions of the reduced data spaces are automatically calculated with arbitrary criterion. The results show that, when reduced data cubes are used, the anomaly detection algorithms originally developed for hyperspectral data, the well-known Reed and Xiaoli Yu detector (RX) and the regularized adaptive RX (RARX), give good detection performances for both surface and sub-surface defects in a non-supervised way.     

不确定海洋环境中基于蒙特卡罗优化的稳健检测方法

常规的检测算法在实际不确定海洋环境中会遇到环境失配的问题,进而导致检测性能下降。本文结合贝叶斯原理和广义似然比方法,基于蒙特卡罗优化技术,提出了一种不确定海洋环境中信号检测方法。该检测器将环境先验信息应用到广义似然比检验中,在保证有效检测的基础上,降低了计算复杂度。同时给出了精确模型匹配检测器、最优贝叶斯信号检测器、平均模型匹配检测器和能量检测器作为对比的检测算法。计算机仿真和SWellEx-96海上实测数据处理结果表明,本文提出的信号检测器检测取得了优于平均模型匹配检测器和能量检测器的性能,其计算效率也有明显提高。      

一种容差模拟电路的分级故障诊断方法

In this paper, a tolerance analog circuit fault diagnosis method based on hierarchical fault dictionary is proposed. During the simulation before test, firstly, the Worse-Case Analysis is used to get the normal characteristics output interval of the circuit under test and the output interval is saved as the first class fault dictionary, which will be used to fault detection; secondly, node-voltage sensitivity sequence is used as fault characteristics to build the second class fault dictionary for locating fault component; thirdly, based on simulation before test according to dividing the component parameters into seven segments, the third class fault dictionary is built to identify the parameter interval of components. In the fault diagnosis stage, based on the established three-class fault dictionary, fault detection, fault locating and component parameter interval identification can be realized respectively according to practical application. The proposed method can improve the efficiency of diagnosis after test and the solution will be a meaningful reference for practical applications. Finally, the simulation experiment demonstrates the effectiveness of the proposed method.     

Optimization of One-Way Quantum Computation Measurement Patterns

In one-way quantum computation (1WQC), an initial highly entangled state, called a graph state, is used to perform universal quantum computations by a sequence of adaptive single-qubit measurements and post-measurement Pauli-X and Pauli-Z corrections. 1WQC computation can be represented by a measurement pattern (or simply a pattern). The entanglement operations in a pattern can be shown by a graph which together with the identified set of its input and output qubits is called the geometry of the pattern. Since a pattern is based on quantum measurements, which are fundamentally nondeterministic evolutions, there must be conditions over geometries to guarantee determinism. These conditions are formalized by the notions of flow and generalized flow (gflow). Previously, three optimization methods have been proposed to optimize 1WQC patterns which can be performed using the measurement calculus formalism by rewriting rules. However, the serial implementation of these rules is time consuming due to executing many ineffective commutation rules. To overcome this problem, in this paper, a new scheme is proposed to perform the optimization techniques simultaneously on patterns with flow and only gflow based on their geometries. Furthermore, the proposed scheme obtains the maximally delayed gflow order for geometries with flow. It is shown that the time complexity of the proposed approach is improved over the previous ones.     

不确定海洋环境中基于主成分量的稳健检测器

传统的信号检测算法在不确定的海洋环境中性能出现下降。基于贝叶斯原理的最优检测算法可以实现对不确定海洋环境中信号的有效检测,但是其突出问题是计算量较大。本文提出了一种基于主成分量分析的稳健信号检测器,该检测器利用贝叶斯原理将环境先验信息引入到检测算法中,同时使用主成分量分析方法来降低运算量,实现了对信号的快速有效检测。分别使用标准失配海洋模型和海上实测数据进行了计算机仿真和实验验证,结果表明:(1)基于主成分量的稳健信号检测器检测性能达到最优贝叶斯检测器的效果。(2)本文方法在线运算速度是贝叶斯最优检测器的5^一8倍。(3)环境先验信息失配的情况下,扩大海洋环境参数模型的不确定度范围有助于提高检测性能。      

Heterogeneous recurrence representation and quantification of dynamic transitions in continuous nonlinear processes

Many real-world systems are evolving over time and exhibit dynamical behaviors. In order to cope with system complexity, sensing devices are commonly deployed to monitor system dynamics. Online sensing brings the proliferation of big data that are nonlinear and nonstationary. Although there is rich information on nonlinear dynamics, significant challenges remain in realizing the full potential of sensing data for system control. This paper presents a new approach of heterogeneous recurrence analysis for online monitoring and anomaly detection in nonlinear dynamic processes. A partition scheme, named as Q-tree indexing, is firstly introduced to delineate local recurrence regions in the multi-dimensional continuous state space. Further, we design a new fractal representation of state transitions among recurrence regions, and then develop new measures to quantify heterogeneous recurrence patterns. Finally, we develop a multivariate detection method for on-line monitoring and predictive control of process recurrences. Case studies show that the proposed approach not only captures heterogeneous recurrence patterns in the transformed space, but also provides effective online control charts to monitor and detect dynamical transitions in the underlying nonlinear processes.     

基于新型光谱相似度量核的高光谱异常检测算法

提出一种新型光谱相似度量核函数,并应用于高光谱异常检测.由于高斯径向基核函数是基于光谱向量间欧式距离的度量,其对于光谱向量间距离变化的适应性较强,而对于因光照强度变化,阴影和遮挡等引起的同种地物光谱变化的适应性较弱,使得基于高斯径向基核的高光谱异常检测算法性能下降.为解决该问题,从光谱曲线形状描述出发,基于光谱相似度量提出了光谱相似度量核函数.通过理论分析和真实高光谱数据异常检测实验检验,实验结果说明相对于高斯径向基核函数,光谱相似度量核函数具有一定的优越性,能改善基于核方法的高光谱异常检测算法的性能.     

基于3D卷积自编解码器和低秩表示的高光谱异常检测

针对高光谱影像数据维度高、空间和光谱信息利用不足以及局部结构特征表达有限等问题,提出了一种基于3D卷积自编解码器和低秩表示的高光谱异常检测算法。首先,通过3D卷积自编解码器提取高光谱影像的空谱特征,并针对高光谱图像的局部区域强相关性,设计了一种新的损失函数来约束中心像素和周围像素,以提取判别性较强的特征图;然后,针对所提取的特征图,通过基于密度的空间聚类算法构建背景字典,并利用低秩表示分离出异常区域;最后,融合由3D卷积自编解码器得到的重构误差和异常区域检测结果,得到最终检测图并为异常目标关键信息的挖掘提供依据。为了验证所提算法的有效性,在两个真实的机场高光谱数据集上进行飞机等目标检测实验,ROC、AUC量化指标和主观分析等实验结果表明,与其它6种异常检测算法相比,本文算法具有更高的异常目标检测精度。     

An extended segment pattern dictionary for a pattern matching tracking algorithm at BESⅢ

A pattern matching based tracking algorithm, named MdcPatRec, is used for the reconstruction of charged tracks in the drift chamber of the BESⅢ detector. This paper addresses the shortage of segment finding in the MdcPatRec algorithm. An extended segment construction scheme and the corresponding pattern dictionary are presented. Evaluation with Monte-Carlo and experimental data show that the new method can achieve higher efficiency for low transverse momentum tracks.     

一种符号阵列编码结构光三维检测方法

编码结构光检测技术是一种主动视觉方法,利用投射的模式光,通过三角原理获得深度信息。根据伪随机阵列的特性,提出了一种可用于检测场景目标的一次投射模式的结构光三维检测方法。符号的拐点、交叉点比传统的基于伪随机序列及M阵列模式提供了更多的检测点。符号阵列能够提供足够多的码字使所有的子模式获得全局唯一性。基于符号链分解算法及角度变化的轮廓特征,解码方法能够识别绝大多数码字。重建实验表明可以对一定曲面变化的物体进行检测,并能重建其形廓。     

Research on the Fastest Detection Method for Weak Trends under Noise Interference

Trend anomaly detection is the practice of comparing and analyzing current and historical data trends to detect real-time abnormalities in online industrial data-streams. It has the advantages of tracking a concept drift automatically and predicting trend changes in the shortest time, making it important both for algorithmic research and industry. However, industrial data streams contain considerable noise that interferes with detecting weak anomalies. In this paper, the fastest detection algorithm “sliding nesting” is adopted. It is based on calculating the data weight in each window by applying variable weights, while maintaining the method of trend-effective integration accumulation. The new algorithm changes the traditional calculation method of the trend anomaly detection score, which calculates the score in a short window. This algorithm, SNWFD–DS, can detect weak trend abnormalities in the presence of noise interference. Compared with other methods, it has significant advantages. An on-site oil drilling data test shows that this method can significantly reduce delays compared with other methods and can improve the detection accuracy of weak trend anomalies under noise interference.