Iris representation and recognition of humans based on characteristic eye features have taken a significant place in security and identification systems in the past two decades. The first steps in this field were made in the early 90's of the last century by John Daugman who designed the integrodifferential operator for detecting an iris and a pupil in an image. This paper will analyze the process of human identification based on iris features, as well as the impact of the Daugman operator parameters on its performance.
Trustworthy corpora are necessary for training and meaningful evaluation of algorithms which use annotations. These standard collections are called Gold Standard Corpora (GSC). However the construction of GSC is a laborious and time-consuming process and size, quality and most of all availability of task-specific GSC directly influence the development of machine learning based natural language processing algorithms. This paper provides an introduction to gold standard corpus construction in the context of natural language processing and gives an overview of alternative approaches.
—Tables offer an often used way to represent information for the human reader. But as long as those tables are not annotated with semantic information they are meaningless to machines. In this work a methodology is proposed to annotate the headers of table columns with semantic types by creating a ranking of possible column headers based on the column cells. In the performed experiments on 10 independent columns a mean average precision value of about 0.7 was achieved. Moreover experiments showed that on average only 15 cells have to be considered to gain good results. Therefore performance does not depend on the number of rows of the table.
A stencil computation determines the values of points in a grid of some dimensionality by repeatedly evaluating a given function of a grid point and its neighbors. The parallelization and optimization of stencil computations are subject of ongoing research. The most prevalent approach is the subdivision of the iteration domain into smaller pieces, called tiles. We give an overview of a method to increase the performance of one such tiling algorithm further by employing a dynamic schedule for tile processing, improving both load balance and cache efficiency. A set of onedimensional stencil benchmarks exhibits a performance increase of up to 20% in comparison to the Pochoir stencil compiler.
Feature detectors and descriptors seek to find descriptive keypoints in an image that serve as a basis for further usages such as matching, tracking and so forth. We investigate several state-of-the-art feature descriptor and detectors in surgical applications. The presented descriptors and detectors include SIFT, SURF, FAST and FREAK. Furthermore, we present an overview of common techniques to evaluate feature descriptors and detectors. Combinations of these were evaluated on surgical images such as retina and Cholecystectomy images using homographies based on random warpings. In our evaluation, we found that the best combination of feature descriptor and detector was FAST to detect interest points and SURF to describe them in terms of repeatability and computational time.
— Stereolithography (STL) is an additive manufacturing process which employs an ultraviolet curable photopolymer resin to form an object layer by layer. Due to the complicated multi-physical nature of the process, the major challenge that arises is the accuracy of the finished parts. The aim of the student-research project was to gain insight into the main governing effects of STL and to model this process within a high order finite element framework.
This work presents ongoing research on a distributed cluster-based spectrum mobility protocol with the aim to reduce the impact of PU activity on SU communication and enable higher throughput and better QoS in CRNs.
The development of REST APIs requires a great deal of time because developers have to address several issues to ensure a suitable design and the desired usability. This paper uses the model-driven approach to generate proper and functional REST APIs while focusing on the core aspects of REST and the consequences for the generation process. We present a method, which allows to design a REST API from an abstract point of view. This user-defined model allows to generate source code for a REST framework such as Jersey. The implemented framework includes the generation of specified resources, methods, and a persistence layer. This paper presents a REST meta-model and a user model, which enables the user to specify the desired API. After the initialization of the generation process, the framework provides a ready-to-use REST API.
Search processes exert influence on everyday life they support information search and help us to understand complex topics. To facilitate these processes we introduce three different interactive visualization algorithms which allow the comparison of subsequent searches during a search process. Specifically, the visualizations help us to identify the expected search results. Furthermore they show us common search results during the adaptation of the search keywords. On the one hand the visual representation of the search process gives an overview, on the other hand it enables the recognition of similar search results as well as subject areas within the result set. Our design builds upon the visualization techniques Circle Packing, Force-Directed Graph and Chord Diagram to show hierarchies and dependencies in result sets based on the evaluation of different query facets and the related data types. Our work is the topic of the author’s master thesis at the Chair of Media Computer Science at the University of Passau. Its current state is presented.
Over the years, large scientific code packages have been successfully developed using FORTRAN. Today, maintaining and extending these codes is a challenging task and often forbids several possibilities. Rewriting legacy code packages is almost prohibitive due to lacking resources and inertia against changing functional codes. One way of tackling these challenges is to integrate scripting languages into existing codes thereby retaining performance-critical kernels in FORTRAN and add an interface to a flexible, dynamically typed scripting language. Amongst the scripting languages, Python stands as a forerunner, especially due to the availability of scientific packages such as NUMPY and SciPy. In this project, an interface to Python is developed for a thermal-hydraulic simulation tool ATHLET, which is mainly used for nuclear reactor safety problems. The FORTRAN to Python interface generator tool (f2py) is used in an innovative way to have access to FORTRAN subroutines and module variables from Python and vice versa. Standard interfaces for three purposes have been developed. The interfaces allow for processing the simulation results during the simulation, calling Python subroutines within a simulation time step and coupling external simulations with ATHLET. The result of the project is an easy-to-use Python Framework to a large FORTRAN code which is very low on overhead, provides standard interfaces and requires very little changes to the original code. This framework allows the user to extend the functionality of ATHLET, by writing new modules in Python and coupling with external simulation codes, without having to deal with the rigours of FORTRAN programming. Index Terms — Scientific computing, Programming environments, Application programming interfaces, Computer languages, Computer applications
— With a rapid growth of information available on the internet there is also growing demand in applications that can process data from different sources accessing information that is only needed for its particular use. Due to the fact that information on the internet is mostly unstructured and thus cannot be processed automatically many tasks which require information extraction are still performed manually. Thereby research of methods for selectively structuring and semantically classifying of unstructured text data is playing an increasingly significant role. In this paper, based on a research project at University of Applied Sciences Augsburg, a method for extracting application-specific information is discussed on a concrete practical example. Beside the presented algorithms this paper intents to provide the reader with knowledge about semantic application development and thus focuses on applied research in this area. Consequently, methods from fields such as information retrieval, Semantic Web and ontologies are brought together and are applied in the example application in order to investigate applicability and relevance in practice.
Flow-Design is a new approach in developing software architectures. Flow-Design focuses on data processing function units and the communication among them. Event-Based Components represent a programming language independent methodology for implementing Flow-Design and can also be considered as a resulting software architecture. The Event- Based Components software architecture is structured by components that communicate via events. Therefore it is compared to the concept of service-oriented architecture and event-driven architecture which aim to create independent components as well. In conclusion Flow-Design and Event-Based Components provide a platform and programming language independent concept to develop software components. The resulting components are reusable, maintainable and testable. Although it can lead to a more complex coding.
Although serial communication as in RS-232 is venerable by today’s standards, it is still found in new designs. Moreover, unlike as in the old days when used to communicate over lossy phone lines, for instance in an automotive environment it is crucial to ensure its reliability and safety. In this paper, an approach for the necessary Bit Error Rate (BER) testing using Pseudo-Random Bit Sequences (PRBS) and borrowing from the IEEE 802.3 Ethernet standard is presented. The result of this research is a Bit Error Rate Tester (BERT), suitable to be included within the embedded devices composing the communication equipment.
—Sensor faults continue to be a major hurdle for systems health management to reach its full potential. At the same time, few recorded instances of sensor faults exist. It is equally difficult to seed particular sensor faults. Therefore, research is underway to better understand the different fault modes seen in sensors and to model the faults. The fault models can then be used in simulated sensor fault scenarios to ensure that algorithms can distinguish between sensor faults and system faults. The paper illustrates the work with data collected from an electro-mechanical actuator in an aerospace setting, equippedout were simulated and injected into the experimental data, with the goal of making these simulations as realistic as feasible. A neural network based classifier was then created and tested on both experimental data and the more challenging randomized data sequences. Additional studies were also conducted to determine sensitivity of detection and disambiguation efficacy to severity of fault conditions.
Recent advances in mobile technologies allow new applications in medicine and healthcare. Such systems cover often biosignal acquisition and processing. A framework can be used to provide basic structures and functionalities of biosignal applications. It should support developers to build complex biosignal systems for their specific application or to develop and test new analysing methods. For this, model-driven software development offers a promising approach. This paper proposes an approach to the model-driven development of a biosignal framework. The impacts to the development process and architecture are pointed out and its benefits are assessed.
This paper introduces a cross-platform low-cost system for 3D object reconstruction using a projection-based laser scanner. It uses contact-free measurement techniques for 3D object reconstruction and fast surface registration using Iterative Closest Point (ICP) [1]. The only hardware requirements are a simple hand-held laser line projector, a calibration rig and a standard camera. The camera is initially calibrated using Zhang’s camera calibration method so that its external and internal parameters are known. The visible intersection with the known background is used to find the 3D pose of the laser plane. This laser plane is used to triangulate new 3D point coordinates of the object’s surface. The point clouds obtained are processed using the "3DTK The 3D Toolkit" [2] which includes an automatic high-accurate registration process and a fast 3D viewer.
The influence of the configuration of the Nexus trace port for on-chip-instruction-trace with regard to the achievable compression ratio is studied in detail. On the base of the EEMBC Automotive Benchmark Software a wide range of configurations and trace modes is analyzed. The impacts of the port width, the count of MSEO-bits, the basic messaging mode and the addition of timestamp information to the compression ratio are examined.
The principal component analysis (also named Karhunen–Loeve transformation) and the factor analysis are both tools of the multivariate statistics, more precisely the exploratory data analysis. They are used e.g. in data mining or machine learning. Although they share the same goal, they reach it with different methods. Over the years, some misunderstandings came up, how these methods differ from each other. The result is, that sometimes if one talks about using the factor analysis, actually the principal component analysis was implemented. Or it is believed, that the factor analysis is an superset of the principal component analysis. These misconceptions could prevent practitioners to chose the appropriate tool for their intended use. This paper clarifies some misconceptions. After some introductory words, principal component analysis and factor analysis are described on its own. This lays the foundation of a comparison between them. The conclusion settles the case.
