Ayodeji Oyewale1 and Chris Hughes2, 1School of Computing, Science and Engineering, University of Salford, Salford, Manchester and 2The Crescent, Salford, Manchester, United Kingdom
A growing number of applications that generate massive streams of data need intelligent data processing and online analysis. Data & Knowledge Engineering (DKE) has been known to stimulate the exchange of ideas and interaction between these two related fields of interest. DKE makes it possible to understand, apply and assess knowledge and skills required for the development and application data mining systems. With present technology, companies are able to collect vast amounts of data with relative ease. With no hesitation, many companies now have more data than they can handle. A vital portion of this data entails large unstructured data sets which amount up to 90 percent of an organization’s data. With data quantities growing steadily, the explosion of data is putting a strain on infrastructures as diverse companies having to increase their data center capacity with more servers and storages. This study conceptualized handling enormous data as a stream mining problem that applies to continuous data stream and proposes an ensemble of unsupervised learning methods for efficiently detecting anomalies in stream data.
Stream data, Steam Mining, Compact data structurres, FP Tree, Path Adjustment Method
N. Rada, L. E. Mendoza, E. G. Florez ,TelecommunicationsEngineering, Biomedical engineering, Mechanical Engineering, Research Group in Mechanical Engineering, Universityof Pamplona, Colombia
This article presents a robust compression method known as compression sensitivity (CS). CS, allows to reconstruct scat-tered signals with very few samples unlike the Shannon-Nyquist theorem. In this article the discrete cosine transform and the wavelet transform were used to find most adequate sparse space. Angiographic images were used, which were reconstructed using algorithms such as Large-scale Sparse Reconstruction (SPGL) and Gradient Projection for Sparse Reconstruction (GPRS). In this work, it was demonstrated that using the wavelet-cosine transformed transpose allowed achieving a more satisfactory sparse space than those obtained by other research. Finally, it was demonstrated that CS works in a relevant way for compressing angiographic images and the maximum percentage of error in the reconstruction was 3.56% for SPGL.
Compressive Sensing, sparse signal, images, reconstruction, SPGL1, SPSR.
Hyun Woo Jung, Hankuk Academy of Foreign Studies, Yongin, South Korea
Deep learning has facilitated major advancements in various fields including image detection. This paper is an exploratory study on improving the performance of Convolutional Neural Network (CNN) models in environments with limited computing resources, such as the Raspberry Pi. A pretrained state-of-art algorithm for doing near-real time object detection in videos, YOLO (“You-Only-Look-Once”) CNN model, was selected for evaluating strategies for optimizng the runtime performance. Various performance analysis tools provided by the Linux kernel were used to measure CPU time and memory footprint. Our results show that loop parallelization, static compilation of weights, and flattening of convolution layers reduce the total runtime by 85% and reduce memory footprint by 53% on a Raspberry Pi 3 device. These findings suggest that the methodological improvements proposed in this work can reduce the computational overload of running CNN models on devices with limited computing resources.
Deep Learning, Convolutional Neural Networks, Raspberry Pi, real-time object detection
Niloufar Salehi Dastjerdi and M. Omair Ahmad, Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada
Image descriptors play an important role in any computer vision system e.g. object recognition and tracking. Effective representation of an image is challenging due to significant appearance changes, viewpoint shifts, lighting variations and varied object poses. These challenges have led to the development of several features and their representations. Spatiogram and region covariance are two excellent image descriptors which are widely used in the field of computer vision. Spatiogram is a generalization of the histogram and contains some moments upon the coordinates of the pixels corresponding to each bin. Spatiogram captures richer appearance information as it computes not only information about the range of the function like histograms, also information about the (spatial) domain. However, there is a drawback that multi modal spatial patterns cannot be well modelled. Region covariance descriptor provides a compact and natural way of fusing different visual features inside a region of interest. However, it is based on a global distribution of pixel features inside a region and loses the local structure. In this paper, we aim to overcome the existing drawbacks of these descriptors. To this, we propose r-spatiogram and then a new hybrid descriptor is presented which is combination of r-spatiogram and traditional region covariance descriptors. The results show that our descriptors have the discriminative capability improved in comparison with other descriptors.
Feature Descriptor, Spatiogram, Region Covariance
Niloufar Salehi Dastjerdi and M. Omair Ahmad, Department of Electrical and Computer Engineering, Concordia University, Montreal, Quebec, Canada
Object tracking, in general, is a promising technology that can be utilized in a wide variety of applications. It is a challenging problem and its difficulties in tracking objects may fail when confronted with challenging scenarios such as similar background color, occlusion, illumination variation, or background clutter. A number of ongoing challenges still remain and an improvement on accuracy can be obtained with additional processing of information. Hence, utilizing depth information can potentially be exploited to boost the performance of traditional object tracking algorithms. Therefore, a large trend in this paper is to integrate depth data with other features in tracking to improve the performance of tracking algorithm and disambiguate occlusions and overcome other challenges such as illumination artifacts. For this, we use the basic idea of many trackers which consists of three main components of the reference model, i.e., object modeling, object detection and localization, and model updating. However, there are major improvements in our system. Our forth component, occlusion handling, utilizes the depth spatiogram of target and occluder to localize the target and occluder. The proposed research develops an efficient and robust way to keep tracking the object throughout video sequences in the presence of significant appearance variations and severe occlusions. The proposed method is evaluated on the Princeton RGBD tracking dataset and the obtained results demonstrate the effectiveness of the proposed method.
Visual Tracking, Depth Spatiogram, Multi-feature Data, Occlusion Handling
Ranjana S. Zinjore1 and Rakesh J. Ramteke2,1Department of Computer Science, G.G. Khadse College, Muktainagar and 2School of Computer Sciences, KBC North Maharashtra University, Jalgaon
Optical Character Recognition has got a special significance in Multi-lingual, Multi-Script country like India, where a single document may contain words in two or more languages/scripts. There is a need to digitize such type of documents for easy communication and storage. It is also useful in applications like processing of handwritten messages on social media and processing of handwritten criminal records for judicial purpose. This paper reveals the approach used in the digitization of handwritten bilingual documents consist of Marathi and English languages. In this approach three phases are used. The first phase focuses on preprocessing of handwritten bilingual document and solution of merged line segmentation. An algorithm Two _Fold_ Word _Segmentation is developed to extract words from lines. A fusion of two feature extraction methods is used for script identification. Second phase focuses on recognition of script identified words. For recognition of words two different feature extraction methods are used. The first method is based on combination of structural and statistical features and second method is based on Histogram of Oriented Gradient Method. K-Nearest Neighbor classifier gives good recognition accuracy for second feature extraction method than that of first method. Finally in third phase digitization and transliteration of recognized words is performed. A graphical user interface is designed for conversion of transliterated text into speech which is useful in the society for blind and visually impaired people to read a book consisting of bilingual text.
Digitization, Transliteration, Script Identification Histogram of Oriented Gradient, K-Nearest Neighbor
Roxana Flores-Quispe and Yuber Velazco-Paredes, Deparment of Computer Science, Universidad Nacional de San Agustin, Arequipa, Peru
This paper proposes a method based on Multitexton Histogram (MTH) descriptor to classificate eight different human parasite eggs: Ascaris, Uncinarias, Trichuris, Hymenolepis Nana, Dyphillobothrium Pacificum, Taenia-Solium, Fasciola Hepatica and Enterobius-Vermicularis identifying textons of irregular shapes in their microscopic images. This proposed method includes two stages. In the first a feature extraction mechanism integrates the advantages of co-occurrence matrix and histograms to identify irregular morphological structures in the biological images throughs textons of irregular shape. In the second stage the Support Vector Machine (SVM) is used to classificate the different human parasite eggs. The results were obtaining using a dataset with 2053 human parasite eggs images achieving a success rate of 96,82% in the classification.
Human Parasite Eggs, Multitexton Histogram descriptor, Textons.
Mate Kisantal1, Zbigniew Wojna1,2, Jakub Murawski2,3, Jacek Naruniec3, Kyunghyun Cho4, 1Tensor ight, Inc., 2University College London, 3Warsaw University of Technology and 4New York University
In the recent years, object detection has experienced impressive progress. Despite these improvements, there is still a significant gap in the performance between the detection of small and large objects. We analyze the current state-of-the-art model, Mask-RCNN, on a challenging dataset, MS COCO. We show that the overlap between small ground-truth objects and the predicted anchors is much lower than the expected IoU threshold. We conjecture this is due to two factors; (1) only a few images are containing small objects, and (2) small objects do not appear enough even within each image containing them.We thus propose to oversample those images with small objects and augment each of those images by copy-pasting small objects many times. It allows us to trade off the quality of the detector on large objects with that on small objects. We evaluate different pasting augmentation strategies, and ultimately, we achieve 9.7% relative improvement on the instance segmentation and 7.1% on the object detection of small objects, compared to the current state of the art method on MS COCO.
Lei Feng, Key Laboratory of Computational Optical Imaging Technology, Academy of Opto-electronics, Chinese Academy of Sciences, Beijing, China
Wide spectral band, combination of imaging and spectrum and fine spectral detection capability are the outstanding advantages of imaging spectrometer. Rich spectral information combined with spatial image of the object point greatly improves the accuracy of target detection, expands the function of traditional detection technology, and realizes the qualitative analysis of target characteristics. Spectrometer plays an irreplaceable role compared with other technologies. It has been widely used in many military and civilian fields, such as land and ocean remote sensing, remote sensing monitoring of pollutants in atmosphere, soil and water, military target detection, medical spectral imaging diagnosis, scientific experiments and so on. Curved prism spectrometer is widely used because of its high energy and no ghost image. However, curved prism spectrometer is as a non coaxial symmetric system and its aberration theory is complex. Therefore, it is necessary to establish a numerical model and construct an initial structure to provide a good starting point for system optimization. In a practical imaging spectrometer based on prism, there are many aberrations when ray incidented on the surface of each element. In the design, it is very important to establish a mathematical model to analyze these aberrations. Curved prism is a kind of non-coaxial prism which is obtained by processing the front and rear surfaces of triangular prism into two spheres. Its front and rear surfaces are not coaxial with the optical axis, so its characteristics are complex. Firstly, on the basis of the primary aberration theory, the numerical calculation model of curved prism is established, and the optimal object distance of curved prism and the effective incident angle of curved prism are solved according to the principle of minimum aberration. For given system parameters, the coordinates of object points are known, and then the numerical model of curved surface prism spectrometer is established. The vector method is used to solve the incident and output vectors of given light rays. After transmission, the optical path extremum function is established, and the second-order partial differential equation is derived. The surface equation of each element is expanded by higher order Taylor series, that is, each surface is expressed as a functional expression of the incident point and structural parameters. A set of partial differential equations is constructed, and the least square method is used to solve the minimum of the equations, and then the initial structural parameters are calculated.
Mathematical computation, partial differential equations, vector solving, curved prism
Hamidreza Bolhasani, Amir Masoud Rahmani and Farid Kheiri, Department of Computer Engineering, Science and Research branch, Islamic Azad University, Tehran, Iran
Since 1982 that Richard Feynman proposed the idea of quantum computing for the first time, it has become a new field of interest for many physics and computer scientists. Although it’s more than 30 years that this concept has been presented but it’s still considered as unknown and several subjects are open for research. Accordingly, concepts and theoretical reviews may always be useful. In this paper, a brief history and fundamental ideas of quantum computers are introduced with focus on architecture part.
Quantum, Computer, Hardware, Qubit, Gate
Vahab Pournaghshband1 and Peter Reiher2 , 1Computer Science Department, University of San Francisco, San Francisco, USA, 2Computer Science Department, University of California, Los Angeles, Los Angeles,USA
The market is currently sated with mobile medical devices and new technology is continuously emerging. Thus, it is costly, and in some cases impractical, to replace these devices for new ones with greater security. In this paper, we present the implementation of a prototype for Personal Security Device a self-contained, specialized wearable device that augments security to existing mobile medical devices. The main research challenge for, and hence the state of the art of, the proposed hardware design is that the device, to work with legacy devices, must require no changes to either the medical device or its monitoring software. This requirement is essential since we aim to protect already existing devices, as making modifications to the device or its proprietary software often impossible or impractical (e.g., closed source executables and implantable medical devices). Through performance evaluation of this prototype, we confirmed the feasibility of having a special-purpose hardware with limited computational and memory resources to perform necessary security operations.
Wireless medical device security, Man-in-the-middle attack.
Shailja Dalmia, Ashwin T S and Ram Mohana Reddy Guddeti, National Institute of Technology Karnataka, Surathkal, Mangalore, Karnataka, India
With the ever-growing variety of information, the retrieval demands of different users are so multifarious that the traditional search engine cannot afford such heterogeneous retrieval results of huge magnitudes. Harnessing the advancements in a user-centered adaptive search engine will aid in groundbreaking retrieval results achieved efficiently for high-quality content. Previous work in this field have made using the excessive server load to achieve good retrieval results but with the limited extended ability and ignoring on demand generated content. To address this gap, we propose a novel model of adaptive search engine and describe how this model is realized in a distributed cluster environment. Using an improved current algorithm of topic-oriented web crawler with User Interface based Information Extraction Technique was able to produce a renewed set of user-centered retrieval results with higher efficiency than all existing methods. The proposed method was found to exceed by 1.5 times and two times for crawler and indexer, respectively than all prevailing methods with improved and highly precise results in extracting semantic information from Deep web.
Search Engine, WWW, Web Content Mining, Inverted Indexing, Hidden Crawler, Distributed Web Crawler, Precision, Deep Web
Ikwu Ruth and Louvieris Panos, Department of Computer Sciences, Brunel University, London
Cyberspace has gradually replaced the physical reality, its role evolving from a simple enabler of daily live processes to a necessity for modern existence. As a result of this convergence of physical and virtual realities, for all processes being critically dependent on networked communications, information representative of our physical, logical and social thoughts are constantly being generated in cyberspace. The interconnection and integration of links between our physical and virtual realities create a new hyperspace as a source of data and information. Additionally, significant studies in cyber analysis have predominantly revolved around a single linear analysis of information from a single source of evidence (The Network). These studies are limited in their ability to understand the dynamics of relationships across the multiple dimensions of cyberspace. This paper introduces a multi-dimensional perspective for data identification in cyberspace. It provides critical discussions for identifying entangled relationships amongst entities across cyberspace.
Cyberspace, Data-streams, Multi-Dimensional Cyberspace
Solomon Cheung1, Yu Sun1 and Fangyan Zhang2, 1Department of Computer Science, California State Polytechnic University, Pomona, CA, 91768 and 2ASML, San Jose, CA, 95131
As an act of disposing waste and maintaining homeostasis, humans have to use the restroom multiple times a day. One item that is consumed in the process is toilet paper; it often runs out easily in the most inconvenient times. One of the most fatal positions to be in is to be stuck without toilet paper. Since humans are not capable of a 100% resupply rate, we should give this task to a computer. The approach we selected was to use a pair of laser sensors to detect whether toilet paper was absent or not. Utilizing an ultrasound sensor, we would be able to detect whether a person was nearby and send a notification to a database. The online app, PaperSafe, takes the information stored and displays it onto a device for quick access. Once a sufficient amount of data is acquired, we can train a machine learning algorithm to predict the next supply date, optimized for the specific scenario.
Amenity, Homeostasis, Machine Learning, Mobile Application
Vijayalakshmi M,Shanthi ThangamM and Bushra H, Department of Information Science and Technology, Anna University, Chennai City, Tamil Nadu, India
The usages of mobile devices are drastically increasing every day with high end support to the users. Due to high end configurations mobile devices such as smart phones, laptops, tablets, etc., computations are complex in these devices. Computation intensive and data intensive are plays a vital role in the mobile devices. The main challenges in the mobile devices are handling the mobile applications in the devices with high computation and high storage. The above mentioned challenges can be overcome by using mobile cloud computing. The limitations while handling the mobile cloud computing is offloading decision making, which part of computation should offload and which should execute in the mobile side. The proposed work provides the solution to the limitations and challenges mentioned earlier by providing agent based offloading decision maker for mobile cloud. The decision maker should decide which computation part is executed in the mobile side and the cloud side. The evaluation shows the mobile applications having high complexity get benefited over other high applications.
Agent based, Mobile cloud, Offloading, Computational device.
Sandile Mhlanga1, Dr Tawanda Blessing Chiyangwa2, Dr Lall Manoj1 and Prof Sunday Ojo1, 1Tshwane University of Technology,South Africa and 2University of South Africa,South Africa
With the rapid growth of Web services in recent years, it is very difficult to choose the suitable web serv
