基于改进Gabor算法的遮挡人脸智能识别方法

为提高有遮挡人脸的识别精度,提出基于改进Gabor算法的遮挡人脸智能识别方法。首先,对人脸图像动态范围压缩,并选择反锐化掩模滤波算法展开图像增强处理;其次,利用Gabor滤波器对信息保留较完整、亮度较高的半边脸进行特征提取;最后将提取到的Gabor特征输入到极限学习机中完成遮挡人脸的智能识别。实验结果表明,所提方法对处理遮挡人脸图像具有良好的效果,且其对人脸图像识别具有精准度高、识别时间短等优点。

基于Gabor滤波器的事件流特征增强及事件相机对象识别

基于Gabor滤波器的事件驱动卷积是仿生分层脉冲神经网络中常用的事件相机对象特征提取方法。为提高该类网络事件相机对象特征提取能力,提出基于Gabor滤波器的事件流特征增强算法,并应用于奖励调节STDP规则的脉冲神经网络事件相机对象识别系统。算法首先将事件流按时间窗口划分为事件流片段,然后提取各时间窗口内的事件流片段特征,同时增强事件数量较多的时间窗口内特征。并基于奖励调节STDP规则帮助网络学习诊断性特征。采用增强算法的网络在MNIST-DVS数据集上的分类精度优于未采用增强算法的网络,并且对于较短的事件流输入也有很好的分类能力。该事件流特征增强算法能够提高基于Gabor滤波器的事件驱动卷积对事件相机对象的特征提取能力。

基于Gabor卷积和Transformer的课堂表情识别研究

为了针对复杂环境变化无法精准识别学生表情问题,本文设计了一个基于Gabor卷积和Transformer的表情识别模型Gabor-Vision-Transformer(GVT).将Gabor卷积和Transformer的思想相结合,设计了一个特征提取块GVT-Block.首先通过Gabor卷积提取富含丰富纹理和边缘信息的面部局部特征,再通过Transformer提取全局数据之间的长距离信息,从而更好地学习面部关键特征,显著提高模型的分类效果.GVT模型在RAF-DB和FER2013Plus数据集上的准确率分别为88.56%和87.38%,并与多个模型进行对比实验和分析,验证了本模型效果的优越性.

基于LTP编码的分数阶Gabor高光谱遥感图像分类

为有效提取高光谱遥感图像的空谱结构特征、增强特征鉴别性及提高分类精度,提出基于局部三元模式编码的分数阶Gabor高光谱遥感图像分类方法。首先,用分数阶三维Gabor滤波器进行局部特征的有效提取;其次,对Gabor相位特征进行局部三元模式编码,以提高特征的可鉴别性;然后,通过随机森林算法对Gabor相位特征进行分类以获得置信度立方体;最后,融合多组基于Gabor的置信立方体,提取具有互补性、强表达性的纹理特征。选择3个训练样本分别在Indian Pines、Salinas、Trento数据集上验证,总体分类精度分别达到63.50%、81.78%、86.89%。实验结果表明,所提的方法具有更好的分类性能。

多模态遥感图像模板匹配Log-Gabor滤波方法

针对多模态遥感图像匹配难的问题,本文提出了一种基于Log-Gabor滤波的高精度匹配方法。该方法采用由粗到细的多层级密集匹配框架,无须进行特征点检测,避开了多模态图像特征点检测重复率低的问题,能够提取大量高精度匹配点对。本文方法主要分为两步:首先,利用多尺度多角度Log-Gabor滤波器构建对图像间非线性辐射差异稳健的特征金字塔;然后,利用粗尺度的底层特征图进行密集模板匹配,提取大量粗粒度的特征匹配点对,在此基础上再利用特征金字塔,实现粗匹配点自下而上的逐层优化,完成高精度特征匹配点对的提取。同时,针对模板匹配滑窗运算效率不高的问题,提出了一种密集模板匹配的快速实现方式,有效减少了密集模板匹配的运算时间。本文使用多组不同模态的遥感图像进行试验,结果表明,本文方法能够克服图像间非线性辐射差异的影响,在正确匹配数目、匹配准确率与匹配精度上均优于现有多模态图像特征匹配方法。

基于多参数Gabor滤波器融合和高斯差分的生丝疵点分割

针对生丝疵点图像背景与疵点差别极小的特点,提出一种基于多方向、多频率Gabor滤波器融合和高斯差分滤波的疵点分割算法。根据疵点纹理特性,首先采用极大值的方法对特定频率多方向疵点Gabor滤波图像进行融合,并基于均值法对不同频率Gabor滤波器图像进行融合,然后使用高斯差分滤波进一步凸显疵点,最后采用全局阈值将疵点与背景分割开来。测试结果表明:该研究算法召回率达到了93.2%,能够准确分割出生丝疵点。

基于分数阶空谱联合Gabor特征的高光谱遥感图像分类

为充分考虑高光谱遥感图像的空谱结构特征,降低数据冗余,获取更具识别性的特征,提高分类精度。提出一种基于分数阶Gabor的高光谱图像分类方法,在分数域实现对局部信号的多分辨分析,以增强对高光谱图像的表征能力。首先,通过设置多阶正弦波构建多组分数阶Gabor滤波器,获得有效的特征表达。其次,对Gabor相位特征进行象限位编码,并通过汉明距离计算码距,降低计算复杂度。最后,融合不同阶的Gabor相位特征从而得到互补的纹理信息,以获取更高的分类性能。基于Trento真实数据集,选择3个分类样本进行训练,总体分类精度达到87.15%,Kappa系数为0.83,实验结果验证了该方法在小样本训练情况下的有效性,对比其他算法,提高了分类精度。

Detection of Residual Yarn in Bobbin Based on Odd Partial Gabor Filter and Multi-Color Space Hierarchical Clustering

In an automatic bobbin management system that simultaneously detects bobbin color and residual yarn,a composite texture segmentation and recognition operation based on an odd partial Gabor filter and multi-color space hierarchical clustering are proposed.Firstly,the parameter-optimized odd partial Gabor filter is used to distinguish bobbin and yarn texture,to explore Garbor parameters for yarn bobbins,and to accurately discriminate frequency characteristics of yarns and texture.Secondly,multi-color clustering segmentation using color spaces such as red,green,blue(RGB)and CIELUV(LUV)solves the problems of over-segmentation and segmentation errors,which are caused by the difficulty of accurately representing the complex and variable color information of yarns in a single-color space and the low contrast between the target and background.Finally,the segmented bobbin is combined with the odd partial Gabor’s edge recognition operator to further distinguish bobbin texture from yarn texture and locate the position and size of the residual yarn.Experimental results show that the method is robust in identifying complex texture,damaged and dyed bobbins,and multi-color yarns.Residual yarn identification can distinguish texture features and residual yarns well and it can be transferred to the detection and differentiation of complex texture,which is significantly better than traditional methods.

一种基于Filter Faster R-CNN的数字PCR液滴检测技术

目的研究液滴数字聚合酶链式反应(ddPCR)液滴检测技术,去除图像中灰尘、气泡、芯片表面的划痕以及微小凹陷等因素产生的异常点对结果的影响,实现高通量、稳定和准确的ddPCR液滴的自动检测。方法提出Filter Faster R-CNN ddPCR液滴检测模型。使用Faster R-CNN生成液滴预测框,之后使用异常点过滤模块(Filter)去除阳性液滴预测框中的异常点。以诺如病毒片段的质粒为模板进行ddPCR实验,建立一个ddPCR数据集,用于模型的训练(2462例,约占78.56%)和测试(672例,约占21.44%)。对异常点过滤模块的3个过滤支路在验证集上进行消融实验,通过与其他ddPCR液滴检测模型进行比较的对比实验以及进行ddPCR的绝对定量实验。结果在少尘和多尘的环境中,Filter Faster R-CNN阳性液滴准确率为98.23%和88.35%,综合指标F1分数分别达到了99.15%和99.14%,高于其他相比较的模型。独立样本T检验的结果证明,相比未添加过滤模块的网络,添加过滤模块后能够显著提示模型在多尘环境中的阳性准确率。在ddPCR绝对定量实验中,将商业化流式检测设备的结果作为标准浓度,绘制了回归线。结果显示,回归线斜率为1.0005,截距为-0.025,决定系数达到了0.9997,二者结果高度一致。结论本文提出了一种基于Filter Faster R-CNN的ddPCR液滴检测技术,为在多种环境条件下的ddPCR实验提供了鲁棒的液滴检测方法。

Underwater four-quadrant dual-beam circumferential scanning laser fuze using nonlinear adaptive backscatter filter based on pauseable SAF-LMS algorithm

The phenomenon of a target echo peak overlapping with the backscattered echo peak significantly undermines the detection range and precision of underwater laser fuzes.To overcome this issue,we propose a four-quadrant dual-beam circumferential scanning laser fuze to distinguish various interference signals and provide more real-time data for the backscatter filtering algorithm.This enhances the algorithm loading capability of the fuze.In order to address the problem of insufficient filtering capacity in existing linear backscatter filtering algorithms,we develop a nonlinear backscattering adaptive filter based on the spline adaptive filter least mean square(SAF-LMS)algorithm.We also designed an algorithm pause module to retain the original trend of the target echo peak,improving the time discrimination accuracy and anti-interference capability of the fuze.Finally,experiments are conducted with varying signal-to-noise ratios of the original underwater target echo signals.The experimental results show that the average signal-to-noise ratio before and after filtering can be improved by more than31 d B,with an increase of up to 76%in extreme detection distance.

Estimating the subsolar magnetopause position from soft X-ray images using a low-pass image filter

The Lunar Environment heliospheric X-ray Imager(LEXI)and Solar wind Magnetosphere Ionosphere Link Explorer(SMILE)missions will image the Earth’s dayside magneto pause and cusps in soft X-rays after their respective launches in the near future,to specify glo bal magnetic reconnection modes for varying solar wind conditions.To suppo rt the success of these scientific missions,it is critical to develop techniques that extract the magnetopause locations from the observed soft X-ray images.In this research,we introduce a new geometric equation that calculates the subsolar magnetopause position(RS)from a satellite position,the look direction of the instrument,and the angle at which the X-ray emission is maximized.Two assumptions are used in this method:(1)The look direction where soft X-ray emissions are maximized lies tangent to the magnetopause,and(2)the magnetopause surface near the subsolar point is almost spherical and thus RSis nea rly equal to the radius of the magneto pause curvature.We create synthetic soft X-ray images by using the Open Geospace General Circulation Model(OpenGGCM)global magnetohydrodynamic model,the galactic background,the instrument point spread function,and Poisson noise.We then apply the fast Fourier transform and Gaussian low-pass filte rs to the synthetic images to re move noise and obtain accurate look angles for the soft X-ray pea ks.From the filte red images,we calculate RS and its accuracy for different LEXI locations,look directions,and solar wind densities by using the OpenGGCM subsolar magnetopause location as ground truth.Our method estimates RS with an accuracy of<0.3 RE when the solar wind density exceeds>10 cm-3.The accuracy improves for greater solar wind densities and during southward interplanetary magnetic fields.The method ca ptures the magnetopause motion during southwa rd interplaneta ry magnetic field turnings.Consequently,the technique will enable quantitative analysis of the magnetopause motion and help reveal the dayside reconnection modes for dynamic solar wind conditions.This technique will suppo rt the LEXI and SMILE missions in achieving their scientific o bjectives.

基于Gabor变换的电网调度自动化设备运行状态监控技术

提出基于Gabor变换的电网调度自动化设备运行状态监控技术。该方法从电网调度自动化设备的大规模数据库技术和自动化电网调度技术入手,分别采集2项技术在电网调度自动化设备运行中存储和调度的状态信息,并将状态信息与Gabor变换结合,消除状态信息携带的干扰噪声,通过将优化后的状态信息输入以模糊神经网络与电网调度自动化设备健康评估体系为基础建立的设备运行状态监控模型中,实现电网调度自动化设备运行状态监控。实验结果表明,所提方法能够迅速监控电网调度自动化设备运行故障状态,监控功率与实际功率一致,监控精确度高。

An improved particle filter indoor fusion positioning approach based on Wi-Fi/PDR/geomagnetic field

The existing indoor fusion positioning methods based on Pedestrian Dead Reckoning(PDR)and geomagnetic technology have the problems of large initial position error,low sensor accuracy,and geomagnetic mismatch.In this study,a novel indoor fusion positioning approach based on the improved particle filter algorithm by geomagnetic iterative matching is proposed,where Wi-Fi,PDR,and geomagnetic signals are integrated to improve indoor positioning performances.One important contribution is that geomagnetic iterative matching is firstly proposed based on the particle filter algorithm.During the positioning process,an iterative window and a constraint window are introduced to limit the particle generation range and the geomagnetic matching range respectively.The position is corrected several times based on geomagnetic iterative matching in the location correction stage when the pedestrian movement is detected,which made up for the shortage of only one time of geomagnetic correction in the existing particle filter algorithm.In addition,this study also proposes a real-time step detection algorithm based on multi-threshold constraints to judge whether pedestrians are moving,which satisfies the real-time requirement of our fusion positioning approach.Through experimental verification,the average positioning accuracy of the proposed approach reaches 1.59 m,which improves 33.2%compared with the existing particle filter fusion positioning algorithms.

Uncertainties of landslide susceptibility prediction: Influences of random errors in landslide conditioning factors and errors reduction by low pass filter method

In the existing landslide susceptibility prediction(LSP)models,the influences of random errors in landslide conditioning factors on LSP are not considered,instead the original conditioning factors are directly taken as the model inputs,which brings uncertainties to LSP results.This study aims to reveal the influence rules of the different proportional random errors in conditioning factors on the LSP un-certainties,and further explore a method which can effectively reduce the random errors in conditioning factors.The original conditioning factors are firstly used to construct original factors-based LSP models,and then different random errors of 5%,10%,15% and 20%are added to these original factors for con-structing relevant errors-based LSP models.Secondly,low-pass filter-based LSP models are constructed by eliminating the random errors using low-pass filter method.Thirdly,the Ruijin County of China with 370 landslides and 16 conditioning factors are used as study case.Three typical machine learning models,i.e.multilayer perceptron(MLP),support vector machine(SVM)and random forest(RF),are selected as LSP models.Finally,the LSP uncertainties are discussed and results show that:(1)The low-pass filter can effectively reduce the random errors in conditioning factors to decrease the LSP uncertainties.(2)With the proportions of random errors increasing from 5%to 20%,the LSP uncertainty increases continuously.(3)The original factors-based models are feasible for LSP in the absence of more accurate conditioning factors.(4)The influence degrees of two uncertainty issues,machine learning models and different proportions of random errors,on the LSP modeling are large and basically the same.(5)The Shapley values effectively explain the internal mechanism of machine learning model predicting landslide sus-ceptibility.In conclusion,greater proportion of random errors in conditioning factors results in higher LSP uncertainty,and low-pass filter can effectively reduce these random errors.

基于Gabor时频滤波的电磁超声螺栓轴力测量

高强度螺栓轴向预紧力(轴力)的测量在工程应用中具有重要意义。使用电磁超声波对螺栓轴力检测时对超声回波声时测量精度要求较高,传统互相关估计法对超声回波声时估计易因噪声干扰发生估计错误,无法满足轴力测量精度要求。针对互相关法对电磁超声测量信号声时估计存在不准确的问题,提出了Gabor时频滤波法。通过螺栓轴力测量实验采集测量信号,对测量信号进行Gabor变换,在时频域中进行滤波,再对去噪后的信号进行互相关估计测得信号的声时,进而计算出螺栓轴力。实验表明:Gabor时频滤波法能有效地滤除电磁超声信号中的噪声,改善互相关估计的稳定性,提高螺栓轴力测量的准确率。

Nonlinear Filtering With Sample-Based Approximation Under Constrained Communication:Progress, Insights and Trends

The nonlinear filtering problem has enduringly been an active research topic in both academia and industry due to its ever-growing theoretical importance and practical significance.The main objective of nonlinear filtering is to infer the states of a nonlinear dynamical system of interest based on the available noisy measurements. In recent years, the advance of network communication technology has not only popularized the networked systems with apparent advantages in terms of installation,cost and maintenance, but also brought about a series of challenges to the design of nonlinear filtering algorithms, among which the communication constraint has been recognized as a dominating concern. In this context, a great number of investigations have been launched towards the networked nonlinear filtering problem with communication constraints, and many samplebased nonlinear filters have been developed to deal with the highly nonlinear and/or non-Gaussian scenarios. The aim of this paper is to provide a timely survey about the recent advances on the sample-based networked nonlinear filtering problem from the perspective of communication constraints. More specifically, we first review three important families of sample-based filtering methods known as the unscented Kalman filter, particle filter,and maximum correntropy filter. Then, the latest developments are surveyed with stress on the topics regarding incomplete/imperfect information, limited resources and cyber security.Finally, several challenges and open problems are highlighted to shed some lights on the possible trends of future research in this realm.

A dual adaptive unscented Kalman filter algorithm for SINS-based integrated navigation system

In this study, the problem of measuring noise pollution distribution by the intertial-based integrated navigation system is effectively suppressed. Based on nonlinear inertial navigation error modeling, a nested dual Kalman filter framework structure is developed. It consists of unscented Kalman filter (UKF)master filter and Kalman filter slave filter. This method uses nonlinear UKF for integrated navigation state estimation. At the same time, the exact noise measurement covariance is estimated by the Kalman filter dependency filter. The algorithm based on dual adaptive UKF (Dual-AUKF) has high accuracy and robustness, especially in the case of measurement information interference. Finally, vehicle-mounted and ship-mounted integrated navigation tests are conducted. Compared with traditional UKF and the Sage-Husa adaptive UKF (SH-AUKF), this method has comparable filtering accuracy and better filtering stability. The effectiveness of the proposed algorithm is verified.

Recursive Filtering for Stochastic Systems With Filter-and-Forward Successive Relays

In this paper,the recursive filtering problem is considered for stochastic systems over filter-and-forward successive relay(FFSR)networks.An FFSR is located between the sensor and the remote filter to forward the measurement.In the successive relay,two cooperative relay nodes are adopted to forward the signals alternatively,thereby existing switching characteristics and inter-relay interferences(IRI).Since the filter-and-forward scheme is employed,the signal received by the relay is retransmitted after it passes through a linear filter.The objective of the paper is to concurrently design optimal recursive filters for FFSR and stochastic systems against switching characteristics and IRI of relays.First,a uniform measurement model is proposed by analyzing the transmission mechanism of FFSR.Then,novel filter structures with switching parameters are constructed for both FFSR and stochastic systems.With the help of the inductive method,filtering error covariances are presented in the form of coupled difference equations.Next,the desired filter gain matrices are further obtained by minimizing the trace of filtering error covariances.Moreover,the stability performance of the filtering algorithm is analyzed where the uniform bound is guaranteed on the filtering error covariance.Finally,the effectiveness of the proposed filtering method over FFSR is verified by a three-order resistance-inductance-capacitance circuit system.

Bayesian Filtering for High-Dimensional State-Space Models With State Partition and Error Compensation

In the era of exponential growth of data availability,the architecture of systems has a trend toward high dimensionality,and directly exploiting holistic information for state inference is not always computationally affordable.This paper proposes a novel Bayesian filtering algorithm that considers algorithmic computational cost and estimation accuracy for high-dimensional linear systems.The high-dimensional state vector is divided into several blocks to save computation resources by avoiding the calculation of error covariance with immense dimensions.After that,two sequential states are estimated simultaneously by introducing an auxiliary variable in the new probability space,mitigating the performance degradation caused by state segmentation.Moreover,the computational cost and error covariance of the proposed algorithm are analyzed analytically to show its distinct features compared with several existing methods.Simulation results illustrate that the proposed Bayesian filtering can maintain a higher estimation accuracy with reasonable computational cost when applied to high-dimensional linear systems.

Dynamic Event-Triggered Quadratic Nonfragile Filtering for Non-Gaussian Systems:Tackling Multiplicative Noises and Missing Measurements

This paper focuses on the quadratic nonfragile filtering problem for linear non-Gaussian systems under multiplicative noises,multiple missing measurements as well as the dynamic event-triggered transmission scheme.The multiple missing measurements are characterized through random variables that obey some given probability distributions,and thresholds of the dynamic event-triggered scheme can be adjusted dynamically via an auxiliary variable.Our attention is concentrated on designing a dynamic event-triggered quadratic nonfragile filter in the well-known minimum-variance sense.To this end,the original system is first augmented by stacking its state/measurement vectors together with second-order Kronecker powers,thus the original design issue is reformulated as that of the augmented system.Subsequently,we analyze statistical properties of augmented noises as well as high-order moments of certain random parameters.With the aid of two well-defined matrix difference equations,we not only obtain upper bounds on filtering error covariances,but also minimize those bounds via carefully designing gain parameters.Finally,an example is presented to explain the effectiveness of this newly established quadratic filtering algorithm.