基于弱匹配概率典型相关性分析的图像自动标注

针对弱匹配多模态数据的相关性建模问题,提出了一种弱匹配概率典型相关性分析模型（semi-paired probabilistic CCA,简称SemiPCCA）.SemiPCCA模型关注于各模态内部的全局结构,模型参数的估计受到了未匹配样本的影响,而未匹配样本则揭示了各模态样本空间的全局结构.在人工弱匹配多模态数据集上的实验结果表明,SemiPCCA可以有效地解决传统CCA（canonical correlation analysis）和PCCA（probabilistic CCA）在匹配样本不足的情况下出现的过拟合问题,取得了较好的效果.提出了一种基于SemiPCCA的图像自动标注方法.该方法基于关联建模的思想,同时使用标注图像及其关键词和未标注图像学习视觉模态和文本模态之间的关联,从而能够更准确地对未知图像进行标注.     

A printed circuit board capacitive sensor for air bubble inside fluidic flow detection

This paper presents a three-electrode capacitive fluidic sensor for detecting an air bubble inside a fluidic channel such as blood vessels, oil or medical liquid channels. The capacitor is designed and fabricated based on a printed circuit board (PCB). The electrodes are fabricated by using copper via structure through top to bottom surface of the PCB. A plastic pipe is layout through the capacitive sensor and perpendicular to the PCB surface. Capacitance of sensor changes when an air bubble inside fluidic flow cross the sensor. The capacitance change can be monitored by using a differential capacitive amplifier, a lock-in amplifier, filter and an NI acquisition card. Signal is processed and calculated on a computer. Air bubble inside the liquid flow are detected by monitor the unbalance signal between the three electrode potential voltages. Output voltage depends on the volume of the air bubble due to dielectric change between capacitor’s electrodes. Output voltage is up to 53 mV when an 2.28 mm³ air bubble crosses the sensing channel. Air bubble velocity can be estimated based on the output pulse signal. This proposed fluidic sensor can be used for void fraction detection in medical devices and systems; fluidic characterization; and water–gas, oil–water and oil–water–gas multiphase flows in petroleum technology. That structure also can apply to the micro-size for detecting in microfluidic to monitor and control changes in microfluidic channels.     

Canonical Correlation Analysis of Video Volume Tensors for Action Categorization and Detection

This paper addresses a spatiotemporal pattern recognition problem. The main purpose of this study is to find a right representation and matching of action video volumes for categorization. A novel method is proposed to measure video-to-video volume similarity by extending Canonical Correlation Analysis (CCA), a principled tool to inspect linear relations between two sets of vectors, to that of two multiway data arrays (or tensors). The proposed method analyzes video volumes as inputs avoiding the difficult problem of explicit motion estimation required in traditional methods and provides a way of spatiotemporal pattern matching that is robust to intraclass variations of actions. The proposed matching is demonstrated for action classification by a simple Nearest Neighbor classifier. We, moreover, propose an automatic action detection method, which performs 3D window search over an input video with action exemplars. The search is speeded up by dynamic learning of subspaces in the proposed CCA. Experiments on a public action data set (KTH) and a self-recorded hand gesture data showed that the proposed method is significantly better than various state-of-the-art methods with respect to accuracy. Our method has low time complexity and does not require any major tuning parameters.     

Mean field annealing using compound Gauss-Markov random fields foredge detection and image estimation

The authors consider the problem of edge detection and image estimation in nonstationary images corrupted by additive Gaussian noise. The noise-free image is represented using the compound Gauss-Markov random field developed by F.C. Jeng and J.W. Woods (1990), and the problem of image estimation and edge detection is posed as a maximum a posteriori estimation problem. Since the a posteriori probability function is nonconvex, computationally intensive stochastic relaxation algorithms are normally required. A deterministic relaxation method based on mean field annealing with a compound Gauss-Markov random (CGMRF) field model is proposed. The authors present a set of iterative equations for the mean values of the intensity and both horizontal and vertical line processes with or without taking into account some interaction between them. The relationship between this technique and two other methods is considered. Edge detection and image estimation results on several noisy images are included.     

Aircraft classification with a low resolution infrared sensor

Existing computer simulations of aircraft infrared signature (IRS) do not account for dispersion induced by uncertainty on input parameters, such as aircraft aspect angles and meteorological conditions. As a result, they are of little use to quantify the detection performance of IR optronic systems: in this case, the scenario encompasses a lot of possible situations that must indeed be considered, but cannot be individually simulated. In this paper, we focus on low resolution infrared sensors and we propose a methodological approach for predicting simulated IRS dispersion of an aircraft, and performing a classification of different aircraft on the resulting set of low resolution infrared images. It is based on a quasi-Monte Carlo survey of the code output dispersion, and on a maximum likelihood classification taking advantage of Bayesian dense deformable template models estimation. This method is illustrated in a typical scenario, i.e., a daylight air-to-ground full-frontal attack by a generic combat aircraft flying at low altitude, over a database of 30,000 simulated aircraft images. Assuming a spatially white noise background model, classification performance is very promising, and appears to be more accurate than more classical state of the art techniques (such as kernel-based support vector classifiers).     

Exposing Region Splicing Forgeries with Blind Local Noise Estimation

Region splicing is a simple and common digital image tampering operation, where a chosen region from one image is composited into another image with the aim to modify the original image’s content. In this paper, we describe an effective method to expose region splicing by revealing inconsistencies in local noise levels, based on the fact that images of different origins may have different noise characteristics introduced by the sensors or post-processing steps. The basis of our region splicing detection method is a new blind noise estimation algorithm, which exploits a particular regular property of the kurtosis of nature images in band-pass domains and the relationship between noise characteristics and kurtosis. The estimation of noise statistics is formulated as an optimization problem with closed-form solution, and is further extended to an efficient estimation method of local noise statistics. We demonstrate the efficacy of our blind global and local noise estimation methods on natural images, and evaluate the performances and robustness of the region splicing detection method on forged images.     

基于旋转中心点估计的遥感目标精确检测算法

由于遥感图像背景复杂、目标密集分布以及目标尺度、形状差异巨大,给检测带来挑战.当前基于R-CNN的两阶段算法在水平框(HBB)检测上取得了良好效果,然而在定向框(OBB)检测上效果有限.基于点估计的HBB目标检测框架,提出用于定向遥感目标检测的旋转中心点估计网络(RCNet),大幅提升一阶段anchor-free算法在倾斜目标检测上的性能,同时保持较高的检测速度.RCNet通过添加一个用于方向预测的分支,实现旋转中心点估计.提出新的角度表示方式,解决回归角度参数loss不连续以及宽高交换导致训练过程不稳定的问题.所提方法在DOTA数据集上取得66.68 mAP的检测精度以及29.4 fps的检测速度,实现了最佳的速度和精度平衡.     

A new algorithm for forest height estimation based on the varied extinction random volume over ground (VERVoG) model using PolInSAR data

ABSTRACT

This paper examines a simple geometrical method for forest height estimation using single-baseline single frequency polarimetric synthetic aperture radar interferometry (PolInSAR) data. The suggested method estimates the forest biophysical parameters based on the varied extinction random volume over ground (VERVoG) model with top layer extinction greater than zero. We approach the problem using a geometrical method without the need for any auxiliary data or prior information. The biophysical parameters, i.e. top layer extinction value, forest height and extinction gradient are estimated in two separate stages. In this framework, the offset value of the extinction is estimated in an independent procedure as a function of a geometrical index based on the signal penetration in the volume layer. As a result, two remaining biophysical parameters can be calculated in a geometrical way based on the observed volume coherence. The proposed algorithm was evaluated using the L-band PolInSAR data of the European Space Agency (ESA) BioSAR 2007 campaign. A pair of experimental SAR (ESAR) images was acquired over the Remningstorp test site in southern Sweden. The selected images were employed for the performance analysis of the proposed approach in the forest height estimation application based on the VERVoG model. The experimental result shows that the proposed inversion method based on the VERVoG model with top layer extinction greater than zero estimates the volume height with an average root mean square error (RMSE) of 2.08 m against light detection and ranging (LiDAR) heights. It presents a significant improvement of forest height accuracy, i.e. 4.1 m compared to the constant extinction RVoG model result, which ignores the forest heterogeneity in the vertical direction.     

A framework for joint estimation of age,gender and ethnicity on a large database

Human age, gender and ethnicity are valuable demographic characteristics. They are also important soft biometric traits useful for human identification or verification. We present a framework that can estimate the three traits jointly. The joint estimation framework could deal with the mutual influence of age, gender, and ethnicity implicitly. Under this joint estimation framework, we explore different methods for simultaneous estimation of age, gender, and ethnicity. The canonical correlation analysis (CCA) based methods, and partial least squares (PLS) models are explored under our joint estimation framework. Both the linear and nonlinear methods are investigated to measure the performance. We also validate some extensions of these methods, such as the least squares formulations of the CCA methods. We found some consistent ranking of these methods under our joint estimation framework. More importantly, we found that the CCA based methods can derive an extremely low dimensionality in estimating age, gender and ethnicity. An analysis of this property is given based on the rank theory. The experiments are conducted on a very large database containing more than 55,000 face images.     

A two-stage registration algorithm for oil spill aerial image by invariants-based similarity and improved ICP

Image registration is a crucial progress in detecting oil spilled on the sea and is also important for estimating the volume of the oil spill, especially when one image cannot cover the entire polluted region. In this article, a new algorithm is proposed to register geometrically distorted aerial images of oil spill accurately and automatically. There are two stages in this algorithm: coarse registration and fine registration. Invariants-based similarity and relative space distance are applied to coarse matching. Then improved iterative closest point (ICP) algorithm is used for registering images finely, which is the combination of ICP and a method of solving assignment problem to deal with mismatches. The performance of the proposed algorithm is evaluated by registering oil spill ultraviolet (UV) and infrared (IR) images, respectively. Compared with traditional ICP and other algorithms, the efficiency and accuracy of the proposed algorithm are highly improved.     

结合ASM及特征融合策略的Tagged MR左心室分割

提出了一种基于ASM框架的Tagged MR图像左心室分割方法。即从基于典型相关分析的特征融合角度对LM滤波器组提取的Tagged MR图像左心室纹理特征用典型相关分析进行优化组合,再用SVM构造分类器,通过分类器来确定边缘点,驱动ASM模型边界变形得到分割结果。通过典型相关分析的特征融合可以降低分类错误率,提高分类性能;用分类器代替经典ASM模型的基于轮廓灰度的匹配法来确定边缘点具有较强的鲁棒性。该方法在不同时刻不同断层Tagged MR图像上进行了验证,实验结果表明该方法具有较高的准确度和较强的鲁棒性。     

Onboard target detection in hyperspectral image based on deep learning with FPGA implementation

Onboard target detection of Hyperspectral Imagery (HSI) is widely adopted in the field of remote sensing. It requires high detection accuracy and low computational complexity for processing a large volume of HSI data. In this study, a Locally Preserving Discriminative Broad Learning (LPDBL) was introduced for target detection due to its simple, excellent generalization ability, and its competitive performance. The detection was done through spatial-spectral information, band selection, and estimation of the covariance matrix. The fisher discriminant method was used to reduce the dimension of HSI data. Weights was adjusted through manifold regularization in order to enhance the detection ability of the proposed method. To study the performance of the proposed LPDBL, experiment was conducted on two different datasets of HSI. The results revealed that the proposed method performed better and suitable for target detection. The LPDBL was implemented on Virtex-7 Field Programmable Gate Array (FPGA) board. Furthermore, the LPDBL technique was practically validated by two different techniques such as a broad learning system (BLS) and Automatic Target Detection in HSI (ATD-HSI). The result obtained from the FPGA was very close to the actual target position.     

基于多视角学习策略的手部姿态估计

手部姿态估计在人机交互、手功能评估、虚拟现实和增强现实等应用中发挥着重要作用, 为此本文提出了一种新的手部姿态估计方法, 以解决手部区域在大多数图像中占比较小和已有单视图关键点检测算法无法应对遮挡情况的问题. 所提方法首先通过引入Bayesian卷积网络的语义分割模型提取手部目标区域, 在此基础上针对手部定位结果, 利用所提基于注意力机制和级联引导策略的新模型以获得较为准确的手部二维关键点检测结果.然后提出了一种利用立体视觉算法计算关键点深度信息的深度网络, 并在深度估计中提供视角自学习的功能. 该方式以三角测量为基础, 利用RANSAC算法对测量结果进行校准. 最后经过多任务学习和重投影训练对手部关键点的3D检测结果进行优化, 最终提取手部关键点的三维姿态信息. 实验结果表明: 相比于已有的一些代表性人手区域检测算法, 本文方法在人手区域检测上的平均检测精度和运算时间上有一定的改善. 此外, 从本文所提姿态估计方法与已有其他方法的平均端点误差(EPE_mean)和PCK曲线下方面积(AUC)这些指标的对比结果来看, 本文方法的关键点检测性能更优, 因而能获得更好的手部姿态估计结果.     

激光激发下拉曼光谱相对辐射定标模型研究

为了提高变压器油中气体检测性能,提出基于激光激发下拉曼光谱的相对辐射定标模型,采用激光成像技术进行拉曼光谱图像成像处理,提取拉曼光谱图像的边缘轮廓特征量,采用多尺度小波降噪方法进行激光激发下拉曼光谱图像降噪滤波,在邻域内采用颜色梯度分解方法进行拉曼光谱图像区域融合性增强处理,根据激光激发下拉曼光谱图像RGB值与邻域均值实现拉曼光谱相对辐射标定和检测,提高拉曼光谱图像的相对辐射定标能力。仿真结果表明,采用该方法进行激光激发下拉曼光谱相对辐射定标的准确性较高,识别精度较高,图像信息增强性能较好。     

一种基于典型相关分析的主动声纳水下混响抑制算法

混响抑制是提升主动声纳接收机性能的关键技术之一。传统的混响抑制算法主要采用多普勒和时间增益控制,以及预白化处理。这些算法需要对混响建模并进行参数估计,其计算量较大,且易受水声环境变化的影响。为了提升抗混响算法的鲁棒性,提出了一种基于典型相关分析的算法,能够与支持向量机结合实现对混响的预先分类识别。湖上试验结果表明该算法能够显著提升混响背景下主动声纳的检测性能。     

A Novel Fuzzy Approach for State Estimation of Nonlinear Hybrid Systems Using Particle Filtering Method

Reliable state estimation is challenging for nonlinear hybrid systems. Particle filtering has emerged as an appealing approach for online hybrid state estimation. Mode detection in nonlinear hybrid systems is, however, a troublesome issue for the conventional particle filter mainly due to sample impoverishment. The problem is also exacerbated when dynamics that govern healthy or faulty modes are close together. False mode detection consequently leads to erroneous continuous state estimation. This paper proposes a novel fuzzy‐based particle filter to reduce continuous state estimation errors due to failures in mode detection. It is fulfilled by considering a fuzzified contribution of each feasible mode in overall estimation. In addition, two new resampling strategies are presented to tackle the degeneracy problem. A set of simulation test studies are conducted to extract the characteristic features and evaluate the performance of the proposed algorithm compared to observation and transition‐based most likely modes tracking particle filter (OTPF) as one of the most meticulous proposed estimation algorithms. The simulation results demonstrate the superior efficiency of the algorithm in dealing with the considered potential estimation problems.     

Gradient feature extraction for classification-based face detection

Face detection from cluttered images is challenging due to the wide variability of face appearances and the complexity of image backgrounds. This paper proposes a classification-based method for locating frontal faces in cluttered images. To improve the detection performance, we extract gradient direction features from local window images as the input of the underlying two-class classifier. The gradient direction representation provides better discrimination ability than the image intensity, and we show that the combination of gradient directionality and intensity outperforms the gradient feature alone. The underlying classifier is a polynomial neural network (PNN) on a reduced feature subspace learned by principal component analysis (PCA). The incorporation of the residual of subspace projection into the PNN was shown to improve the classification performance. The classifier is trained on samples of face and non-face images to discriminate between the two classes. The superior detection performance of the proposed method is justified in experiments on a large number of images.     

一种用于电容层析成象三相流重建的神经网络算法

在对B-P网络激励函数作出修改的基础上,提出了充分利用ECT系统的测量数据对气/油/水三相流进行成象的方法.该图象重建算法具有概念简单,易于实施,成象速度快,抗噪能力强等优点.     

Box-plot-SA-BP：变压器DGA多参量故障诊断模型

油中溶解气体分析(DGA)方法是一种典型的充油电力设备故障诊断方法,广泛应用于电力变压器故障检测与状态评估,但由于样本数据的可靠性和诊断模型的有效性影响,导致DGA诊断方法准确率较低。文中提出了一种Box-plot-SA-BP模型,首先,采用Box-plot数据检测法去除异常数据以解决数据质量的问题,然后,利用自注意力机制(Self-attention, SA)准确捕捉多参量样本数据间的联系,提取更加稳定可靠的特征,最后设计BP网络多分类模型实现变压器故障诊断。对比实验证明了Box-plot-SA-BP模型的良好性能,具有较高的应用价值。