Development of a stable localized visual inspection system for underwater structures

To avert potential crisis from Japan’s aging infrastructure and declining birth rate, the Japanese Government is planning to introduce robotic technology for the inspection of social infrastructure (such as pipes, dams, and bridges). Recording underwater positions is a difficult task for human divers who undertake conventional dam inspections. This study presents the Anchor Diver 5.2 system for efficient and effectual dam inspection. Anchor Diver 5.2 is based on an extended-tether-maneuvered remotely operated vehicle (ROV) equipped with cameras. The ROV is lowered into water by a hoist system from a boat and implements a visual survey of the concrete underwater structure. To improve the visibility of the ROV in murky and cloudy water, a novel concept named Water Loupe is proposed. In addition, a simple boat-fixing method is proposed to provide a stable base on the water surface, and the underwater position of the ROV, which cannot be accessed by global positioning systems, is recorded using a feasible localization method. Finally, the developed system was evaluated in field experiments conducted in the Amagase Dam, Japan, and its merits and problems are discussed.     

情感词典构建综述

文本情感分析是近年来迅速兴起的一个研究课题,具有显著的研究价值和应用价值。情感词典的构建在情感分析任务中发挥着越来越重要的影响力。该文对情感词典构建的研究进展进行了总结。首先重点介绍了情感词典构建的研究现状,将其归纳为四种方法,即基于启发式规则的方法、基于图的方法、基于词对齐模型的方法以及基于表示学习的方法,并对每种方法进行介绍和分析;然后对一些常见的语料库、词典资源以及评测组织进行介绍;最后,对情感词典的构建进行了总结,并对发展趋势进行了展望。
     

基于mean-shift全局立体匹配方法

针对图像全局立体匹配精度高、计算量大的问题,提出基于mean shift图像分割的全局立体匹配方法。首先,通过mean shift算法对图像进行分割,获取图像同质区域数量和区域的标号。在计算匹配代价时,根据像素所属的分割区域,对像素进行筛选,从而提高匹配代价计算速度;其次,在代价聚合前,将mean shift算法获取的同质区域数K值赋值给K-means聚类算法,对像素再次聚类,提高立体匹配精度和速度;最后通过TRW-S置信传播解决能量最小化问题。实验表明,该算法明显提高了匹配的准确性和速度,与单纯的全局匹配算法相比,具有更大的优势。     

PETOOL: MATLAB-based one-way and two-way split-step parabolic equation tool for radiowave propagation over variable terrain

A MATLAB-based one-way and two-way split-step parabolic equation software tool (PETOOL) has been developed with a user-friendly graphical user interface (GUI) for the analysis and visualization of radio-wave propagation over variable terrain and through homogeneous and inhomogeneous atmosphere. The tool has a unique feature over existing one-way parabolic equation (PE)-based codes, because it utilizes the two-way split-step parabolic equation (SSPE) approach with wide-angle propagator, which is a recursive forward–backward algorithm to incorporate both forward and backward waves into the solution in the presence of variable terrain. First, the formulation of the classical one-way SSPE and the relatively-novel two-way SSPE is presented, with particular emphasis on their capabilities and the limitations. Next, the structure and the GUI capabilities of the PETOOL software tool are discussed in detail. The calibration of PETOOL is performed and demonstrated via analytical comparisons and/or representative canonical tests performed against the Geometric Optic (GO) + Uniform Theory of Diffraction (UTD). The tool can be used for research and/or educational purposes to investigate the effects of a variety of user-defined terrain and range-dependent refractivity profiles in electromagnetic wave propagation.

Program summary

Program title: PETOOL (Parabolic Equation Toolbox)Catalogue identifier: AEJS_v1_0Program summary URL:http://cpc.cs.qub.ac.uk/summaries/AEJS_v1_0.htmlProgram obtainable from: CPC Program Library, Queen?s University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 143 349No. of bytes in distributed program, including test data, etc.: 23 280 251Distribution format: tar.gzProgramming language: MATLAB (MathWorks Inc.) 2010a. Partial Differential Toolbox and Curve Fitting Toolbox requiredComputer: PCOperating system: Windows XP and VistaClassification: 10Nature of problem: Simulation of radio-wave propagation over variable terrain on the Earth?s surface, and through homogeneous and inhomogeneous atmosphere.Solution method: The program implements one-way and two-way Split-Step Parabolic Equation (SSPE) algorithm, with wide-angle propagator. The SSPE is, in general, an initial-value problem starting from a reference range (typically from an antenna), and marching out in range by obtaining the field along the vertical direction at each range step, through the use of step-by-step Fourier transformations. The two-way algorithm incorporates the backward-propagating waves into the standard one-way SSPE by utilizing an iterative forward–backward scheme for modeling multipath effects over a staircase-approximated terrain.Unusual features: This is the first software package implementing a recursive forward–backward SSPE algorithm to account for the multipath effects during radio-wave propagation, and enabling the user to easily analyze and visualize the results of the two-way propagation with GUI capabilities.Running time: Problem dependent. Typically, it is about 1.5 ms (for conducting ground) and 4 ms (for lossy ground) per range step for a vertical field profile of vector length 1500, on Intel Core 2 Duo 1.6 GHz with 2 GB RAM under Windows Vista.     

Efficient region-aware large graph construction towards scalable multi-label propagation

With fast growing number of images on photo-sharing websites such as Flickr and Picasa, it is in urgent need to develop scalable multi-label propagation algorithms for image indexing, management and retrieval. It has been well acknowledged that analysis in semantic region level may greatly improve image annotation performance compared to that in the holistic image level. However, region level approach increases the data scale to several orders of magnitude and proposes new challenges to most existing algorithms. In this work, we present a novel framework to effectively compute pairwise image similarity by accumulating the information of semantic image regions. Firstly, each image is encoded as Bag-of-Regions based on multiple image segmentations. Secondly, all image regions are separated into buckets with efficient locality-sensitive hashing (LSH) method, which guarantees high collision probabilities for similar regions. The k-nearest neighbors of each image and the corresponding similarities can be efficiently approximated with these indexed patches. Lastly, the sparse and region-aware image similarity matrix is fed into the multi-label extension of the entropic graph regularized semi-supervised learning algorithm [1]. In combination they naturally yield the capability of handling large-scale dataset. Extensive experiments on NUS-WIDE (260k images) and COREL-5k datasets validate the effectiveness and efficiency of our proposed framework for region-aware and scalable multi-label propagation.     

Operational causes of bankruptcy propagation in supply chain

With the increasing interdependence among supply chain members, bankruptcy of a supply chain member may be caused by operational decisions of other members. To investigate how bankruptcy occurs and propagates in supply chain networks, we build a multi-agent simulation model for a two-stage supply chain that consists of multiple upstream manufacturers and multiple downstream retailers. Based on the developed simulation model, we study impacts of various operational parameters and decisions, such as horizontal competition among retailers, order allocation strategies of retailers, wholesale price of manufacturers, characteristics of market demand and number of retailers, on bankruptcy propagation. Since many operational decisions of a firm are made under financial constraints, we also investigate the linkage between firm's operational risks and financial decisions (e.g., the maximal risk of cash flow that a member is willing to take, and the up-front payment proportion of retailers in a two-period payment policy). Experimental results reveal that operational interactions between supply chain members and operational decisions made by supply chain members are important causes of bankruptcy propagation, but impacts of these operational parameters and decisions depend on financial decisions. These observations indicate that supply chain members can moderately hedge their operational risk through financial decisions.     

溶剂系统的模式识别

通过三层ＢＰ人工神经网络,对溶剂系统进行模式识别。以经规格化自理伯Ｘｔ、Ｘｅ、Ｘｍ、Ｘｎ、Ｘｄ这５个溶剂性质的选择性参数作为判别指标,５类共１９种不同的溶剂作为训练样本,得到最佳网络参数。然后以十我种有代表性的溶剂作为待测样本进行了模式识别。研究结果表明,人工神经网络用于溶剂系统的模式识别,识别结果与实际一致,本文结果有助于色谱分析及分离技术中溶剂的选择。     

Feedforward networks in financial predictions: the future that modifies the present

The main goal of this paper is to show how relatively minor modifications of well-known algorithms (in particular, back propagation) can dramatically increase the performance of an artificial neural network (ANN) for time series prediction. We denote our proposed sets of modifications as the 'self-momentum', 'Freud' and 'Jung' rules. In our opinion, they provide an example of an alternative approach to the design of learning strategies for ANNs, one that focuses on basic mathematical conceptualization rather than on formalism and demonstration. The complexity of actual prediction problems makes it necessary to experiment with modelling possibilities whose inherent mathematical properties are often not well understood yet. The problem of time series prediction in stock markets is a case in point. It is well known that asset price dynamics in financial markets are difficult to trace, let alone to predict with an operationally interesting degree of accuracy. We therefore take financial prediction as a meaningful test bed for the validation of our techniques. We discuss in some detail both the theoretical underpinnings of the technique and our case study about financial prediction, finding encouraging evidence that supports the theoretical and operational viability of our new ANN specifications. Ours is clearly only a preliminary step. Further developments of ANN architectures with more and more sophisticated 'learning to learn' characteristics are now under study and test.     

关于BP算法推导过程与讨论

本文通过对多层神经网络模型的BP算法的推导,讨论用BP算法对网络进行优化后误差精度、网络结构和学习次数之间的关系,最后指出BP算法的不足之处,提出加快学习过程的一些方法和技巧。     

重复传播网络的改进及其在遥感分类中的应用研究

对重复传播网络在权值初始化、网络学习进行修改,实现在一定警戒条件下竞争层节点的动
态分配,并在遥感分类上进行实验。重复传播网络中学习次数影响了竞争层节点数目和学习精度,
学习精度随学习次数的增加而有限度的增加;警戒参数决定了节点调整的搜索范围,随学习过程进
行,节点搜索范围变小,保证了分类精度和网络的稳定性;两层竞争层结构分类精度要优于单层,其
总精度和Kappa系数分别提高1.1%,0.02。实验结果表明该方法是一可行的遥感分类方法。