Find Research Output

Research Output
  • All
  • Scholar Profiles
  • Research Units
  • Research Output
Department Publication Year Content Type Data Sources


School of Advanced Technology
Clear all

1.Vehicle Logo Recognition and Attributes Prediction by Multi-task Learning with CNN

Author:Xia, YZ;Feng, J;Zhang, BL


Abstract:Vehicle Logo Recognition(VLR) has been an important study field in intelligent Transportation system (ITS). This paper proposes to recognize vehicle logo and predict logo attributes by combining Convolutional Neural Network (CNN) with Multi-Task Learning(MTL). In order to accelerate convergence of multi-task model, an adaptive weight training strategy is employed. To verify the algorithm, the Xiamen University Vehicle logo recognition dataset is extended into a larger vehicle logo dataset including 15 brands, 6 visual attributes and 3 no-visual attributes. The experiment results indicate that the proposed multi-task CNN model perform well for both of logo classification and attribution prediction with overall accuracy 98.14 %%.

2.E-marketing services and e-marketing performance: the roles of innovation, knowledge complexity and environmental turbulence in influencing the relationship

Author:Chong, WK;Bian, D;Zhang, N


Abstract:The business-to-business electronic marketplace (e-marketplace) is becoming critical for small- and medium-sized enterprises (SMEs). However, which e-marketing services determine a firm's e-marketing performance and how innovation, knowledge complexity and environmental turbulence influence the relationship between e-marketing services and e-marketing performance are under-researched topics in the field. We first empirically tested 176 SMEs from China to evaluate which e-marketing services are significantly related to e-marketing performance and how these services collectively influence the performance. Then, we used an NK model to examine how innovation, knowledge complexity and environmental turbulence mediate/moderate the relationship. The results show that five e-marketing services (e-CRM, e-SCM, e-competitiveness, IS/IT integration and information transparency) can greatly influence e-marketing performance; innovation positively mediates the relationship between e-marketing services and performance; and knowledge complexity and environmental turbulence positively moderate the relationship.

3.Towards a ranking approach for sensor services using primitive cognitive network process

Author:Yuen, KKF;Wang, W


Abstract:The Abstracting sensor capabilities as services is a promising approach to support interoperability and scalability in Internet of Things based applications; however, ranking and selecting these dynamic and large number of real-world services in real-time is a challenging problem, which involves decision making in the presence of multiple criteria. This paper demonstrates an application of the Primitive Cognitive Network Process (P-CNP) to the sensor service ranking with respect to multiple Quality-of-Service criteria.

4.The Primitive Cognitive Network Process in healthcare and medical decision making: Comparisons with the Analytic Hierarchy Process

Author:Yuen, KKF


Abstract:Analytic Hierarchy Process (AHP) is increasingly applied to healthcare and medical research and applications. However, knowledge representation of pairwise reciprocal matrix is still dubious. This research discusses the related drawbacks, and recommends pairwise opposite matrix as the ideal alternative. Pairwise opposite matrix is the key foundation of Primitive Cognitive Network Process (P-CNP), which revises the AHP approach with practical changes. A medical decision treatment evaluation using AHP is revised by P-CNP with a step-by-step tutorial. Comparisons with AHP have been discussed. The proposed method could be a promising decision tool to replace AHP to share information among patients or/and doctors, and to evaluate therapies, medical treatments, health care technologies, medical resources, and healthcare policies. (C) 2013 Elsevier B.V. All rights reserved.

5.Usable Authentication Mechanisms for Mobile Devices: An Exploration of 3D Graphical Passwords

Author:Yu, Z;Olade, I;Liang, HN;Fleming, C


Abstract:Current authentication systems in mobile devices such as smart phones have many shortcomings. Users tend to use simple textual passwords such as PINs, which are easily cracked by intruders. Meanwhile, graphical passwords suffer from shoulder surfing attack In this paper, a new authentication system using 3D graphical passwords, will be proposed and tested to offer more security for mobile devices. This authentication system allows users to interact with the 3D objects in a 3D virtual environment and these actions are tracked in the virtual environment and used to create unique passwords. Based on the previous studies of the 3D password scheme, this paper developed a simple testing program that enables users to create their own 3D password easily. At the end of the paper, some improvements of the program and this authentication system are discussed.

6.Application of Hough Transform Feature Extraction to Reduce Angular Vibration in Images Captured from Moving Objects

Author:Afolabi, D;Man, KL;Liang, HN;Zhang, N;Lim, EG;Wan, KY


Abstract:This paper details an ongoing research aimed at developing computational approach to reducing/eliminating vibration and light glare in images captured by digital cameras especially when the scene contains moving objects or the camera is mounted on a moving vehicle / flying drone. The algorithms developed are focused at real-time image acquisition where the enhanced/corrected images are need almost immediately after they are captured. The results show that these methods of reducing the stated problems are effective and it can be further developed for various applications.


Author:Stankovic, N


Abstract:We report on our experience with two pedagogies over a two year period, using one approach in each academic year, but the learning outcome remained the same. The first approach, in the first year, was backed up by Ian Sommerville's textbook, and the second by Bruegge and Dutoit's. The students came from three different curricula, and the class size in the second year was three times that in the first year. The difference in the outcome and in the meeting of the learning outcome has been strongly in favour of the second pedagogy. In this paper we compare the outcomes, provide a number of examples that explain the issues that the students had to deal with, and how the change in the pedagogy has improved the quality of the laboratory work without changing the project framework itself.

8.MidSens'09 - International Workshop on Middleware Tools, Services and Run-Time Support for Sensor Networks, Co-located with the 10th ACM/IFIP/USENIX International Middleware Conference Preface

Author:Michiels, Sam ; Hughes, Danny

Source:MidSens'09 - International Workshop on Middleware Tools, Services and Run-Time Support for Sensor Networks, Co-located with the 10th ACM/IFIP/USENIX International Middleware Conference,2009,Vol.

9.A Method for Rapid Measurement of Intracellular Lipid Accumulation

Author:Xiong Yiwei;Zhang Junlong;Yang Tian;Zhang Bailing;Jiang Lin


Abstract:Objective: To propose a fast and reliable method for the determination of intracellular lipid accumulation. Method: Lipid accumulation in HepG2 cells were induced by oleic acid. Lipid droplets were stained with BODIPY 493/503 and measured with fluorescent image processing using Matlab or ImageJ. The images results were compared to those of Adipophilin,a widely served marker for lipid accumulation,which were determined from protein and mRNA levels by Western blotting and qPCR. Result: By processing fluorescent images using BODIPY 493/503 staining for lipid droplets and comparing data outputs with Matlab and ImageJ, a method using Matlab as image processing platform was established for the determination of intracellular lipid accumulation ; a 1. 41 - fold increase in oleic acid - induced lipid accumulation was detected by fluorescent image processing, which were in accordance with the markedly enhancement in Adipophilin protein levels (2. 07 - fold) and mRNA levels (1. 53 - fold). Conclusion: The method combining fluorescent staining and image processing shall help analyze the intracellular lipid accumulation both rapidly and accurately.

10.Modeling and Verification of NCL Circuits Using PAT

Author:Ma, JM;Man, KL;Lim, EG;Zhang, N;Lei, CU;Guan, SU;Jeong, TT;Seon, JK

Source:CEIS 2011,2011,Vol.15

Abstract:NULL Conventional Logic (NCL) is a Delay-Insensitive (DI) clockless paradigm and is suitable for implementing asynchronous circuits. Efficient methods of analysis are required to specify and verify such DI systems. Based on Delay Insensitive sequential Process (DISP) specification, this paper demonstrates the application of formal methods by applying Process Analysis Toolkit (PAT) to model and verify the behavior of NCL circuits. A few useful constructs are successfully modeled and verified by using PAT. The flexibility and simplicity of the coding, simulation and verification shows that PAT is effective and applicable for NCL circuit design and verification. (C) 2011 Published by Elsevier Ltd. Selection and/or peer-review under responsibility of [CEIS 2011]

11.One-class kernel subspace ensemble for medical image classification (vol 2014, 17, 2014)

Author:Zhang, YG;Zhang, BL;Coenen, F;Xiao, JM;Lu, WJ




Source:Economic Review,2011,Vol.


13.Driving Posture Recognition by a Hierarchal Classification System with Multiple Features

Author:Yan, C;Zhang, BL;Coenen, F


Abstract:This paper presents a novel system for vision-based driving posture recognition. The driving posture dataset was prepared by a side-mounted camera looking at a driver's left profile. After pre-processing for illumination variations, eight action classes of constitutive components of the driving activities were segmented, including normal driving, operating a cell phone, eating and smoking. A global grid-based representation for the action sequence was emphasized, which featured two consecutive steps. Step 1 generates a motion descriptive shape based on a motion frequency image(MFI), and step 2 applies the pyramid histogram of oriented gradients (PHOG) for more discriminating characterization. A three level hierarchal classification system is designed to overcome the difficulties of some overlapping classes. Four commonly applied classifiers, including k-nearest neighbor(KNN), random forest (RF), support vector machine(SVM) and multiple layer perceptron (MLP), are evaluated in each level. The overall classification accuracy is over 87.2%% for the eight classes of driving actions by the proposed classification system.

14.An exploration of techniques for off-screen content interaction in mobile devices

Author:Shen, T;Liang, HN;Liu, D;Man, KL


Abstract:This paper is an attempt to explore new ways for allowing interaction with mobile devices, such as smartphones and tablets. Mobile devices have become very popular in a relatively short time. Their popularity is mainly due to the portability and until very recently their ease-of-interaction through the touch-enabled screen when compared to older devices. Their portability comes with a trade-off: the small screen through which users can interact with their content. The small size screen can limit significantly the amount of content they can interact with, and also the manner in which this interaction is carried out. In this research, we explore the ways of extending the interaction space of these devices. We assess if, how well, and what mechanisms are possible for enabling users to interact with content located off the screen. In this paper, we report our approach and some early results of experiments we have conducted.

15.Subcellular phenotype images classification by MLP ensembles with random linear oracle

Author:Zhang, Bai-Ling ; Han, Guoxia

Source:5th International Conference on Bioinformatics and Biomedical Engineering, iCBBE 2011,2011,Vol.

Abstract:Subcellular localization is a key functional characteristic of proteins. An automatic, reliable and efficient prediction system for protein subcellular localization can be used for establishing knowledge of the spatial distribution of proteins within living cells and permits to screen systems for drug discovery or for early diagnosis of a disease. In this paper, we investigate an approach based on augmented image features by incorporating curvelet transform and neural network (MLP) ensemble for classification. A simple Random Subspace (RS) ensemble offers satisfactory performance, which contains a set of base MLP classifiers trained with subsets of attributes randomly drawn from the combined features of curvelet coefficients and original Subcellular Location Features (SLF). An MLP ensemble with Random Linear Oracle (RLO) can further improve the performance by replacing a base classifier with a "miniensemble", which consists of a pair of base classifiers and a fixed, randomly created oracle that selects between them. With the benchmarking 2D HeLa images, our experiments show the effectiveness of the proposed approach. The RS-MLP ensemble offers the classification rate 95%% while the RS-RLO ensemble gives 95.7%% accuracy, which compares sharply with the previously published benchmarking result 84%%. © 2011 IEEE.

16.Classified Vector Quantisation and population decoding for pattern recognition

Author:Bailing Zhang;Sheng-Uei Guan

Source:International journal of artificial intelligence and soft computing: IJAISC,2009,Vol.

Abstract:Learning Vector Quantisation (LVQ) is a method of applying the Vector Quantisation (VQ) to generate references for Nearest Neighbour (NN) classification. Though successful in many occasions, LVQ suffers from several shortcomings, especially the reference vectors are prone to diverge. In this paper, we propose a Classified Vector Quantisation (CVQ) to establish VQ for classification. By CVQ, each data category is represented by its own codebook, which can be implemented by some learning algorithms. In classification process, each codebook offers a generalised NN. The examples of handwritten digit recognition and offline signature verification are used to demonstrate the efficiency of the proposed scheme.

17.E-commerce systems for software agents: Challenges and opportunities

Author:Tadjouddine,Emmanuel M.

Source:E-Business Issues, Challenges and Opportunities for SMEs: Driving Competitiveness,2010,Vol.

Abstract:It is hoped agent mediated e-commerce will be carried out as open systems of agents interoperating between different institutions, where different auction protocols may be in use. The authors argue that in order to put such a scenario to work, agents will need a method to automatically verify the properties of a previously unseen auction protocol. This, in turn poses the problem of automatically verifying desirable properties in order to trust a given auction mechanism. This challenge needs be addressed so that the business scenario of agent mediated e-commerce becomes a reality. In this chapter, the authors discuss salient opportunities for SMEs in addressing the issues of enabling software agents (e.g., PDAs, mobile phones) to connect to auction houses and verify desirable properties that need to hold before engaging any transactions. © 2011, IGI Global.

18.SimpleGAN Stabilizing generative adversarial networks with simple distributions

Author:Zhang, Shufei ; Qian, Zhuang ; Huang, Kaizhu ; Zhang, Rui ; Hussain, Amir

Source:IEEE International Conference on Data Mining Workshops, ICDMW,2019,Vol.2019-November

Abstract:Generative Adversarial Networks (GANs) are powerful generative models, but usually suffer from hard training and poor generation. Due to complex data and generation distributions in high dimensional space, it is difficult to measure the departure of two distributions, which is however vital for training successful GANs. Previous methods try to alleviate this problem by choosing reasonable divergence metrics. Unlike previous methods, in this paper, we propose a novel method called SimpleGAN to tackle this problem transform original complex distributions to simple ones in the low dimensional space while keeping information and then measure the departure of two simple distributions. This novel method offers a new direction to tackle the stability of GANs. Specifically, starting from maximization of the mutual information between variables in the original high dimensional space and low dimensional space, we eventually derive to optimize a much simplified version, i.e. the lower bound of the mutual information. For experiments, we implement our proposed method on different baselines i.e. traditional GAN, WGAN-GP and DCGAN for CIFAR-10 dataset. Our proposed method achieves obvious improvement over these baseline models. © 2019 IEEE.

19.Image Captioning using Adversarial Networks and Reinforcement Learning

Author:Yan, SY;Wu, FY;Smith, JS;Lu, WJ;Zhang, BL


Abstract:Image captioning is a significant task in artificial intelligence which connects computer vision and natural language processing. With the rapid development of deep learning, the sequence to sequence model with attention, has become one of the main approaches for the task of image captioning. Nevertheless, a significant issue exists in the current framework: the exposure bias problem of Maximum Likelihood Estimation (MLE) in the sequence model. To address this problem, we use generative adversarial networks (GANs) for image captioning, which compensates for the exposure bias problem of MLE and also can generate more realistic captions. GANs, however, cannot be directly applied to a discrete task, like language processing, due to the discontinuity of the data. Hence, we use a reinforcement learning (RL) technique to estimate the gradients for the network. Also, to obtain the intermediate rewards during the process of language generation, a Monte Carlo roll-out sampling method is utilized. Experimental results on the COCO dataset validate the improved effect from each ingredient of the proposed model. The overall effectiveness is also evaluated.

20.SecIoT: a security framework for the Internet of Things

Author:Huang, X;Craig, P;Lin, HY;Yan, Z


Abstract:The 5th generation wireless system (5G) will support Internet of Things (IoT) by increasing the interconnectivity of electronic devices to support a variety of new and promising networked applications such as the home of the future, environmental monitoring networks, and infrastructure management systems. The potential benefits of the IoT are as profound as they are diverse. However, the benefits of the IoT come with some significant challenges. Not the least of these is that the increased interconnectivity integral to an IoT network increases its vulnerability to malevolent attacks. There is still no proven methodology for the design of security frameworks with device authentication and access control. This paper attempts to address this problem through the development of a prototype security framework with robust and transparent security protection. This includes an investigation into the security requirements of three different characteristic IoT scenarios (concretely, body IoT, home IoT, and hotel IoT), a design of new authentication mechanisms, and an access control subsystem with fine-grained roles and risk indicators. Our prototype security framework gives us an insight into some of the major difficulties of IoT security as well as providing some feasible solutions. Copyright (C) 2015 John Wiley & Sons, Ltd.
Total 600 results found
Copyright 2006-2020 © Xi'an Jiaotong-Liverpool University 苏ICP备07016150号-1 京公网安备 11010102002019号