Vehicle license plate super-resolution using soft learning prior

This paper proposes a new algorithm to perform single-frame image super-resolution (SR) of vehicle license plate (VLP) using soft learning prior. Conventional single-frame SR/interpolation methods such as bi-cubic interpolation often experience over-smoothing near the edges and textured regions. Therefore, learning-based methods have been proposed to handle these shortcomings by incorporating a learning term so that the reconstructed high-resolution images can be guided towards these models. However, existing learning-based methods employ a binary hard-decision approach to determine whether the prior models are fully relevant or totally irrelevant. This approach, however, is inconsistent with many practical applications as the degree of relevance for the prior models may vary. In view of this, this paper proposes a new framework that adopts a soft learning approach in license plate super-resolution. The method integrates image SR with optical character recognition (OCR) to perform VLP SR. The importance of the prior models is estimated through relevance scores obtained from the OCR. These are then incorporated as a soft learning term into a new regularized cost function. Experimental results show that the proposed method is effective in handling license plate SR in both simulated and real experiments.     

Single-frame super resolution of remote-sensing images by convolutional neural networks

Super resolution (SR) refers to generation of a high-resolution (HR) image from a decimated, blurred, low-resolution (LR) image set, which can be either a single-frame or multi-frame that contains a collection of images acquired from slightly different views of the same observation area. In this study, two convolutional neural network (CNN)-based deep learning techniques are adapted in single-frame SR to increase the resolution of remote sensing (RS) images by a factor of 2, 3, and 4. In order to both preserve the colour information and speed up the algorithm, first an intensity hue saturation (IHS) transform is utilized and the SR techniques are only applied to the intensity channel of the images. Colour information is then restored with an inverse IHS transformation. We demonstrate the results of the proposed method on RS images acquired from Satellites Pour l’Observation de la Terre (SPOT) or Earth-observing satellites and Pleiades satellites with different spatial resolution. First synthetic LR images are created by downsampling, then structural similarity (SSIM) Index, peak signal-to-noise ratio (PSNR), Spectral Angle Mapper (SAM) and Erreur Relative Globale Adimensionnelle de Synthese (ERGAS) values are calculated for a quantitative evaluation of the methods. Finally, the method, with better performance results, is tested within a real scenario, that is, with original LR images as the input. The obtained HR images demonstrated visible qualitative enhancements.     

Remote-sensing image fusion using sparse representation with sub-dictionaries

Remote-sensing image fusion aims to obtain a multispectral (MS) image with a high spatial resolution, which integrates spatial information from the panchromatic (Pan) image and with spectral information from the MS image. Sparse representation (SR) has been recently used in remote-sensing image fusion method, and can obtain superior results to many traditional methods. However, the main obstacle is that the dictionary is generated from high resolution MS images (HRMS), which are difficult to acquire. In this article, a new SR-based remote-sensing image fusion method with sub-dictionaries is proposed. The image fusion problem is transformed into a restoration problem under the observation model with the sparsity constraint, so the fused HRMS image can then be reconstructed by a trained dictionary. The proposed dictionary for image fusion is composed of several sub-dictionaries, each of which is constructed from a source Pan image and its corresponding MS images. Therefore, the dictionary can be constructed without other HRMS images. The fusion results from QuickBird and IKONOS remote-sensing images demonstrate that the proposed method gives higher spatial resolution and less spectral distortion compared with other widely used and the state-of-the-art remote-sensing image fusion methods.     

An incremental-learning neural network for the classification of remote-sensing images

A novel classifier for the analysis of remote-sensing images is proposed. Such a classifier is based on Radial Basis Function (RBF) neural networks and relies on an incremental-learning technique. This technique allows the periodical acquisition of new information whenever a new training set becomes available, while preserving the knowledge learnt by the network on previous training sets. In addition, in each retraining phase, the network architecture is automatically updated so that new classes may be considered. These characteristics make the proposed neural classifier a promising tool for several remote-sensing applications.     

基于深度学习的心脏磁共振影像超分辨率前沿进展

心脏为人体血液流动提供动力,是人体血液循环系统的重要组成部分。受人口老龄化影响,心脏病诊疗已成为重大公共健康话题。非侵入式活体心脏成像对心脏疾病的检测、诊断与治疗意义重大。然而,受活体心跳影响,成像扫描时间与心脏影像分辨率成为难以调和的矛盾。为缓和这一矛盾,基于快速扫描获得的低分辨率影像重建出心脏高分辨率影像的超分辨率（super-resolution,SR）重建技术成为研究热点。深度学习技术在医学影像处理领域中展现出强大生命力,基于深度学习的SR技术因其强大的学习能力与数据驱动性,在心脏影像SR重建领域中表现出明显优于传统方法的性能。目前领域内前沿成果较多,但缺少对领域现状进行总结、对未来发展进行展望的综述性文献。因此,本文对领域内现状进行梳理总结,挑选出代表性方法,分析方法特性,总结文献中心脏影像数据来源与规模,给出常用的评价指标,以及模型得出的性能评价结论。分析发现,基于深度学习的心脏SR重建技术取得了较大进展,但在运动伪影抑制、模型简化程度与时间性能方面仍有进步空间。此外,现有模型基本完全依靠网络强大的表达能力,鲜有临床先验知识的引入。最后,模型间性能对比相对较少,且领域内缺少代表性的可用于评价不同心脏SR重建模型性能的数据集。基于深度学习的心脏影像SR技术仍有较大发展空间。     

基于TLSHOG特征新方法的动态手势识别

为了提高实际复杂场景的人机交互中动态手势识别的准确性和实时性,提出了一种时序局部敏感直方图（Temporal Locality Sensitive Histograms of Oriented Gradients,TLSHOG）特征新方法,用于描述手势运动的时序变化和空间姿态,实现了快速而精确的动态手势识别。采用普通网络摄像头获取手部的二维图像序列作为训练样本,然后构造单帧图像特征描述手部的空间姿态,并结合时间金字塔（Temporal Pyramid,TP）来描述手势运动轨迹的时空特征,运用多维支持向量机（Support Vector Machine,SVM）算法进行模型训练,对测试样本中的多种手势进行精确的分类。实验结果表明,该方法准确度高,实时性好,对于复杂背景干扰、光照强度变化有较强的鲁棒性。     

Performance Analytic Models and Analyses for Workflow Architectures

The design and implementation of a workflow management system is typically a large and complex task. Decisions need to be made about the hardware and software platforms, the data structures, the algorithms, and network interconnection of various modules utilized by various users and administrators. These decisions are further complicated by requirements such as flexibility, robustness, modifiability, availability, performance, and usability. As the size of workflow systems increases, organizations are finding that the standard server/client architectures, and off-the-shelf solutions are not adequate. We can further see that in the future, very large-scale workflow systems (VLSW) will become more complex, and more prevalent. Thus, one further requirement is an emphasis of this document: scalability. For the purposes of our scalable workflow investigations, we describe a framework, a taxonomy, a model, and a methodology to investigate the performance of various workflow architectures as the size of the system (number of workcases) grows very large.First, this paper presents a novel workflow architectural framework and taxonomy. We survey some example current workflow products and research prototype systems, illustrating some of the taxonomical categories. In fact, most current workflow architectures fall into only one of the many categories of this taxonomy: the centralized server/client category. The paper next explains a performance analysis methodology useful for exploring this taxonomy. The methodology deploys a layered queuing model, and performs mathematical analysis on this model using a modified MOL (method of layers) combined with a linearization algorithm. Finally, the paper utilizes this methodology to compare and contrast the various architectural categories, providing interesting results about performance as the number of workcases increases. Our analytic results suggest that (a) for VLSW performance determination, software architecture is as important as hardware architecture, and (b) alternatives to the client server architecture provide significantly better scalability.     

基于小波变换和非局部平均的超分辨率图像重建

针对小波域超分辨率方法中重建图像存在的模糊效应,提出一种结合离散小波变换(DWT)、平稳小波变换(SWT)和非局部平均(NLM)的单帧图像重建方法DSNLM。算法首先对低分辨率图像同时进行DWT和SWT,得到四个子带图像;然后结合对应高频子带图像,直接将原始低频图像作为低频子带,各子带利用NLM滤波处理,得到待重建高分辨率图像的各子带图像;最后,通过离散小波逆变换(IDWT)得到最终的重建高分辨率图像。实验结果和重建视觉效果表明,所提方法与已有的超分辨率方法相比更优,在峰值信噪比(PSNR)、均方差(MSE)和结构相似性度量(SSIM)的评价指标上有显著的提高,对图像去噪、去模糊有效。     

Application of the wavelet transform for feature extraction in the analysis of hyperspectral laser-induced fluorescence data

Remote-sensing approaches for environmental protection and exploration have evolved rapidly in the last decade. Among the new operational tools, hyperspectral Fluorescent LiDAR System (FLS®) lidar has demonstrated a high sensitivity and the ability to function in complex environments for real-time, robust oil-spill monitoring on airborne or ship-borne analytical platforms. The capabilities of such analytical platforms include real-time analysis of laser-induced fluorescence (LIF) data. Although numerous examples of the application of signal theory to the analysis of hyperspectral data appear in the remote-sensing literature, the conventional data analysis strategies are not well adapted to the practical issues of the LIF applications. The aim of this article is to provide a new approach for LIF lidar analytical platforms, which is focused on the specifics of hyperspectral LIF data. The approach is based on structural data analysis and interpretation, through which more detailed spectral matching is performed. This article is based on a simulated experiment in which the spectra of actual seawater and well-known types of petroleum products were combined to demonstrate the wavelet-transform-based analysis of LIF data. The final part of the article demonstrates the application of the wavelet transform to the structural analysis of LIF data from field experiments for the detection and identification of oil products in difficult environmental conditions.     

Simulation of MPI applications with time‐independent traces

Analyzing and understanding the performance behavior of parallel applications on parallel computing platforms is a long‐standing concern in the High Performance Computing community. When the targeted platforms are not available, simulation is a reasonable approach to obtain objective performance indicators and explore various hypothetical scenarios. In the context of applications implemented with the Message Passing Interface, two simulation methods have been proposed, on‐line simulation and off‐line simulation, both with their own drawbacks and advantages. In this work, we present an off‐line simulation framework, that is, one that simulates the execution of an application based on event traces obtained from an actual execution. The main novelty of this work, when compared to previously proposed off‐line simulators, is that traces that drive the simulation can be acquired on large, distributed, heterogeneous, and non‐dedicated platforms. As a result, the scalability of trace acquisition is increased, which is achieved by enforcing that traces contain no time‐related information. Moreover, our framework is based on a state‐of‐the‐art scalable, fast, and validated simulation kernel. We introduce the notion of performing off‐line simulation from time‐independent traces, propose and evaluate several trace acquisition strategies, describe our simulation framework, and assess its quality in terms of trace acquisition scalability, simulation accuracy, and simulation time. Copyright © 2014 John Wiley & Sons, Ltd.     

基于智能网卡实现的TCP会话数据采集平台

很多网络数据分析系统需要实时采集TCP会话数据,随着网络带宽的快速增长,基于通用网卡和软件协议栈实现的数据采集平台,用于TCP会话管理和数据传输的开销越来越大,已经成为整个数据分析系统整体性能的瓶颈。为了有效提升系统性能,本文描述了一种基于智能网卡实现的TCP会话数据的采集平台。实验表明,使用该数据采集平台后,网络数据分析系统的性能可以大幅提升。     

一种新的星机双基地SAR空间同步方法

由于卫星与飞机在飞行速度上的巨大差异,收发平台波束足迹的空间同步是星机双基地合成孔径雷达(Spaceborne/Airborne Hybrid Bistatic Synthetic Aperture Radar,SA\|BSAR)的一大技术难点。已经有学者提出了一种基于波束指向控制的同步方法,但该方法存在着成像时间短、算法复杂、方位分辨率低等缺点。提出了一种基于宽波束照射的新方法,对新方法的可行性进行了充分的量化论证和仿真验证。与已有方法相比,所能获得的场景长度略小,但成像时间更长,方位分辨率更高。最后,分析了两种方法的优缺点,明确了各自的适用场合。     

Face Super-resolution Reconstruction and Recognition Using Non-local Similarity Dictionary Learning Based Algorithm

One of the challenges of face recognition in surveillance is the low resolution of face region. Therefore many superresolution (SR) face reconstruction methods are proposed to produce a high-resolution face image from one or a set of low-resolution face images. However, existing dictionary learning based algorithms are sensitive to noise and very time-consuming. In this paper, we define and prove the multi-scale linear combination consistency. In order to improve the performance of SR, we propose a novel SR face reconstruction method based on nonlocal similarity and multi-scale linear combination consistency (NLS-MLC). We further proposed a new recognition approach for very low resolution face images based on resolution scale invariant feature (RSIF). A series of experiments are conducted on two public face image databases to test feasibility of our proposed methods. Experimental results show that the proposed SR method is more robust and computationally effective in face hallucination, and the recognition accuracy of RSIF is higher than some state-of-art algorithms.      

A toolbox for unsupervised change detection analysis

The analysis of multi-temporal remote-sensing images is one of the main applications in Earth’s observation and monitoring. In this paper, we present a Matlab toolbox for change detection analysis of optical multi-temporal remote-sensing data in which unsupervised approaches, iterative principal component analysis (ITPCA), and iteratively reweighted multivariate alteration detection (IR-MAD) are implemented and optimized. The optimization is represented by the implementation of novel pre- and post-processing strategies that aim to mitigate the side effects introduced by different acquisition conditions affecting change detection analysis. Special modules have been designed in order to decrease the required memory when large data sets are processed.     

软件缺陷集成预测模型研究

利用单一分类器构造的缺陷预测模型已经遇到了性能瓶颈, 而集成分类器相比单一分类器往往具有显著的性能优势。以构造高效的集成缺陷预测模型为出发点, 比较了七种不同类型集成分类器的算法和特点。在14个基准数据集上的实验显示, 部分集成预测模型的性能优于基于朴素贝叶斯的单一预测模型。其中, 基于投票的集成分类框架具有最优的预测性能以及统计学意义上的性能优势显著性, 随机森林算法次之。Stacking集成框架也具有较强的泛化能力。     

基于非负邻域嵌入和非局部正则化的单帧图像超分辨率重建算法

单帧图像超分辨率重建是指利用一幅低分辨率图像,通过相应的算法来获取一幅高分辨率图像的技术。提出了一种基于 非负邻域嵌入和 非局部正则化 的单帧图像超分辨率重建算法,以弥补传统邻域嵌入算法的不足。在训练阶段,首先对低分辨率图像预放大2倍,以保证在放大倍数较大时,高、低分辨率图像块之间的邻域关系也能得到较好的保持;在重建阶段,使用非负邻域嵌入来有效地解决近邻数的选取问题;最后利用图像块的非局部相似性构造非局部正则项对重建结果进行修正。实验结果表明,相对于传统算法,本方法的重建结果纹理丰富、边缘清晰。     

Multi Hive Artificial Bee Colony Programming for high dimensional symbolic regression with feature selection

Feature selection is a process that provides model extraction by specifying necessary or related features and improves generalization. The Artificial Bee Colony (ABC) algorithm is one of the most popular optimization algorithms inspired on swarm intelligence developed by simulating the search behavior of honey bees. Artificial Bee Colony Programming (ABCP) is a recently proposed high level automatic programming technique for a Symbolic Regression (SR) problem based on the ABC algorithm. In this paper, a new feature selection method based on ABCP is proposed, Multi Hive ABCP (MHABCP) for high-dimensional SR problems. The learning ability and generalization performance of the proposed MHABCP is investigated using synthetic and real high-dimensional SR datasets and is compared with basic ABCP and GP automatic programming methods. Experimental results show that MHABCP has better performance choosing relevant features in high dimensional SR problems and generalization than other methods.     

A symmetric rank-one method based on extra updating techniques for unconstrained optimization

In this paper, we present a new symmetric rank-one (SR1) method for the solution of unconstrained optimization problems. The proposed method involves an algorithm in which the usual SR1 Hessian is updated a number of times in a way to be specified in some iterations, to improve the performance of the Hessian approximation.In particular, we discuss how to consider a criterion for indicating at each iteration whether it is necessary to employ extra updates. However it is well known that there are some theoretical difficulties when applying the SR1 update. Even for a current positive definite Hessian approximation, it is possible that the SR1 update may not be defined or the SR1 update may not preserve positive definiteness at some iterations. We then employ a restarting procedure that guarantees that updated matrices will be well-defined while preserving positive definiteness of updates. Numerical results support these theoretical considerations. They show that the implementation of the SR1 method using extra updating techniques improves the performance of the SR1 method substantially for a number of test problems from the literature.     

Assessing spectral indices to estimate the fraction of photosynthetically active radiation absorbed by the vegetation canopy

ABSTRACT

The fraction of absorbed photosynthetically active radiation (FPAR) by the vegetation canopy (FPAR_canopy) is an important parameter for vegetation productivity estimation using remote-sensing data. FPAR_canopy is widely estimated using many different spectral vegetation indices (VIs), especially the simple ratio vegetation index (SR) and normalized difference vegetation index (NDVI). However, there have been few studies into which VIs are most suitable for this estimation or into their sensitivities to the leaf area index and the observation geometry of remote-sensing data, which are very important for the accurate estimation of FPAR_canopy based on the plant growth stage and satellite imagery. In this study, nine main VIs calculated from field-measured spectra were evaluated and it was found that the SR and NDVI underestimated and overestimated FPAR_canopy, respectively. It was also found that the enhanced vegetation index produced lesser errors and a higher agreement than other broadband VIs used to estimate FPAR_canopy. Among all the selected VIs, the photochemical reflectance index (PRI) turned out to have the lowest root mean square error of 0.17. The SR produced the highest errors (about 0.37) and lowest index of agreement (about 0.50) compared to the measured values of FPAR_canopy. Except for carotenoid reflectance index (CRI), FPAR_canopy estimated by VIs are evidently sensitive to the leaf area index (LAI), especially for FPAR_canopy (SR), which are also most sensitive to solar zenith angles (SZA). SR, CRI, PRI, and EVI have remarked variations with view zenith angles. Our study shows that FPAR_canopy can be simply and accurately estimated using the most suitable VIs – i.e. EVI and PRI – with broadband and hyperspectral remote-sensing data, respectively, and that the nadir reflectance or nadir bidirectional reflectance distribution function adjusted reflectance should be used to calculate these VIs.