Unsupervised visual object categorization

The number of digital images has been increasing exponentially in the last few years. People have problems managing their image collections and finding a specific image. An automatic image categorization system could help them to manage images and find specific images. In this thesis, an unsupervised visual object categorization system was implemented to categorize a set of unknown images. The system is unsupervised, and hence, it does not need known images to train the system which needs to be manually obtained. Therefore, the number of possible categories and images can be huge. The system implemented in the thesis extracts local features from the images. These local features are used to build a codebook. The local features and the codebook are then used to generate a feature vector for an image. Images are categorized based on the feature vectors. The system is able to categorize any given set of images based on the visual appearance of the images. Images that have similar image regions are grouped together in the same category. Thus, for example, images which contain cars are assigned to the same cluster. The unsupervised visual object categorization system can be used in many situations, e.g., in an Internet search engine. The system can categorize images for a user, and the user can then easily find a specific type of image.

Comparison of Local Feature Detectors and Descriptors for Visual Object Categorization

Local features are used in many computer vision tasks including visual object categorization, content-based image retrieval and object recognition to mention a few. Local features are points, blobs or regions in images that are extracted using a local feature detector. To make use of extracted local features the localized interest points are described using a local feature descriptor. A descriptor histogram vector is a compact representation of an image and can be used for searching and matching images in databases. In this thesis the performance of local feature detectors and descriptors is evaluated for object class detection task. Features are extracted from image samples belonging to several object classes. Matching features are then searched using random image pairs of a same class. The goal of this thesis is to find out what are the best detector and descriptor methods for such task in terms of detector repeatability and descriptor matching rate.

Bayesian classification of forest ecological type from satellite data

The main objective of this study was todo a statistical analysis of ecological type from optical satellite data, using Tipping's sparse Bayesian algorithm. This thesis uses "the Relevence Vector Machine" algorithm in ecological classification betweenforestland and wetland. Further this bi-classification technique was used to do classification of many other different species of trees and produces hierarchical classification of entire subclasses given as a target class. Also, we carried out an attempt to use airborne image of same forest area. Combining it with image analysis, using different image processing operation, we tried to extract good features and later used them to perform classification of forestland and wetland.

Supervised localimage feature detection

This thesis is about detection of local image features. The research topic belongs to the wider area of object detection, which is a machine vision and pattern recognition problem where an object must be detected (located) in an image. State-of-the-art object detection methods often divide the problem into separate interest point detection and local image description steps, but in this thesis a different technique is used, leading to higher quality image features which enable more precise localization. Instead of using interest point detection the landmark positions are marked manually. Therefore, the quality of the image features is not limited by the interest point detection phase and the learning of image features is simplified. The approach combines both interest point detection and local description into one phase for detection. Computational efficiency of the descriptor is therefore important, leaving out many of the commonly used descriptors as unsuitably heavy. Multiresolution Gabor features has been the main descriptor in this thesis and improving their efficiency is a significant part. Actual image features are formed from descriptors by using a classifierwhich can then recognize similar looking patches in new images. The main classifier is based on Gaussian mixture models. Classifiers are used in one-class classifier configuration where there are only positive training samples without explicit background class. The local image feature detection method has been tested with two freely available face detection databases and a proprietary license plate database. The localization performance was very good in these experiments. Other applications applying the same under-lying techniques are also presented, including object categorization and fault detection.

Local and Global Feature Extraction for Invariant Object Recognition

Perceiving the world visually is a basic act for humans, but for computers it is still an unsolved problem. The variability present innatural environments is an obstacle for effective computer vision. The goal of invariant object recognition is to recognise objects in a digital image despite variations in, for example, pose, lighting or occlusion. In this study, invariant object recognition is considered from the viewpoint of feature extraction. Thedifferences between local and global features are studied with emphasis on Hough transform and Gabor filtering based feature extraction. The methods are examined with respect to four capabilities: generality, invariance, stability, and efficiency. Invariant features are presented using both Hough transform and Gabor filtering. A modified Hough transform technique is also presented where the distortion tolerance is increased by incorporating local information. In addition, methods for decreasing the computational costs of the Hough transform employing parallel processing and local information are introduced.

Modelling additional costs caused by a new product feature

Työn tavoitteena oli mallintaa uuden tuoteominaisuuden aiheuttamat lisäkustannukset ja suunnitella päätöksenteon työkalu Timberjack Oy:n kuormatraktorivalmistuksen johtoryhmälle. Tarkoituksena oli luoda karkean tason malli, joka sopisi eri tyyppisten tuoteominaisuuksien kustannuksien selvittämiseen. Uuden tuoteominaisuuden vaikutusta yrityksen eri toimintoihin selvitettiin haastatteluin. Haastattelukierroksen tukena käytettiin kysymyslomaketta. Haastattelujen tavoitteena oli selvittää prosessit, toiminnot ja resurssit, jotka ovat välttämättömiä uuden tuoteominaisuuden tuotantoon saattamisessa ja tuotannossa. Malli suunniteltiin haastattelujen ja tietojärjestelmästä hankitun tiedon pohjalta. Mallin rungon muodostivat ne prosessit ja toiminnot, joihin uudella tuoteominaisuudella on vaikutusta. Huomioon otettiin sellaiset resurssit, joita uusi tuoteominaisuus kuluttaa joko välittömästi, tai välillisesti. Tarkasteluun sisällytettiin ainoastaan lisäkustannukset. Uuden tuoteominaisuuden toteuttamisesta riippumattomat, joka tapauksessa toteutuvat yleiskustannukset jätettiin huomioimatta. Malli on yleistys uuden tuoteominaisuuden aiheuttamista lisäkustannuksista, koska tarkoituksena on, että se sopii eri tyyppisten tuoteominaisuuksien aiheuttamien kustannusten selvittämiseen. Lisäksi malli soveltuu muiden pienehköjen tuotemuutosten kustannusten kartoittamiseen.

Feature selection using Fuzzy Entropy measures with Yu's Similarity measure

In this study, feature selection in classification based problems is highlighted. The role of feature selection methods is to select important features by discarding redundant and irrelevant features in the data set, we investigated this case by using fuzzy entropy measures. We developed fuzzy entropy based feature selection method using Yu's similarity and test this using similarity classifier. As the similarity classifier we used Yu's similarity, we tested our similarity on the real world data set which is dermatological data set. By performing feature selection based on fuzzy entropy measures before classification on our data set the empirical results were very promising, the highest classification accuracy of 98.83% was achieved when testing our similarity measure to the data set. The achieved results were then compared with some other results previously obtained using different similarity classifiers, the obtained results show better accuracy than the one achieved before. The used methods helped to reduce the dimensionality of the used data set, to speed up the computation time of a learning algorithm and therefore have simplified the classification task

Loss Modelling of Three-Level Inverters controlled with Space Vector Modulation Technique

The objective of this master’s thesis is to investigate the loss behavior of three-level ANPC inverter and compare it with conventional NPC inverter. The both inverters are controlled with mature space vector modulation strategy. In order to provide the comparison both accurate and detailed enough NPC and ANPC simulation models should be obtained. The similar control model of SVM is utilized for both NPC and ANPC inverter models. The principles of control algorithms, the structure and description of models are clarified. The power loss calculation model is based on practical calculation approaches with certain assumptions. The comparison between NPC and ANPC topologies is presented based on results obtained for each semiconductor device, their switching and conduction losses and efficiency of the inverters. Alternative switching states of ANPC topology allow distributing losses among the switches more evenly, than in NPC inverter. Obviously, the losses of a switching device depend on its position in the topology. Losses distribution among the components in ANPC topology allows reducing the stress on certain switches, thus losses are equally distributed among the semiconductors, however the efficiency of the inverters is the same. As a new contribution to earlier studies, the obtained models of SVM control, NPC and ANPC inverters have been built. Thus, this thesis can be used in further more complicated modelling of full-power converters for modern multi-megawatt wind energy conversion systems.

The Design and Implementation of a Power Management Feature to a Configuration Management System

Green IT is a term that covers various tasks and concepts that are related to reducing the environmental impact of IT. At enterprise level, Green IT has significant potential to generate sustainable cost savings: the total amount of devices is growing and electricity prices are rising. The lifecycle of a computer can be made more environmentally sustainable using Green IT, e.g. by using energy efficient components and by implementing device power management. The challenge using power management at enterprise level is how to measure and follow-up the impact of power management policies? During the thesis a power management feature was developed to a configuration management system. The feature can be used to automatically power down and power on PCs using a pre-defined schedule and to estimate the total power usage of devices. Measurements indicate that using the feature the device power consumption can be monitored quite precisely and the power consumption can be reduced, which generates electricity cost savings and reduces the environmental impact of IT.

Accuracy Analysis of Frequency Converter Estimates in Diagnostic Applications

Transportation of fluids is one of the most common and energy intensive processes in the industrial and HVAC sectors. Pumping systems are frequently subject to engineering malpractice when dimensioned, which can lead to poor operational efficiency. Moreover, pump monitoring requires dedicated measuring equipment, which imply costly investments. Inefficient pump operation and improper maintenance can increase energy costs substantially and even lead to pump failure. A centrifugal pump is commonly driven by an induction motor. Driving the induction motor with a frequency converter can diminish energy consumption in pump drives and provide better control of a process. In addition, induction machine signals can also be estimated by modern frequency converters, dispensing with the use of sensors. If the estimates are accurate enough, a pump can be modelled and integrated into the frequency converter control scheme. This can open the possibility of joint motor and pump monitoring and diagnostics, thereby allowing the detection of reliability-reducing operating states that can lead to additional maintenance costs. The goal of this work is to study the accuracy of rotational speed, torque and shaft power estimates calculated by a frequency converter. Laboratory tests were performed in order to observe estimate behaviour in both steady-state and transient operation. An induction machine driven by a vector-controlled frequency converter, coupled with another induction machine acting as load was used in the tests. The estimated quantities were obtained through the frequency converter’s Trend Recorder software. A high-precision, HBM T12 torque-speed transducer was used to measure the actual values of the aforementioned variables. The effect of the flux optimization energy saving feature on the estimate quality was also studied. A processing function was developed in MATLAB for comparison of the obtained data. The obtained results confirm the suitability of this particular converter to provide accurate enough estimates for pumping applications.

Scaling up stepwise feature introduction to construction of large software systems

Developing software is a difficult and error-prone activity. Furthermore, the complexity of modern computer applications is significant. Hence,an organised approach to software construction is crucial. Stepwise Feature Introduction – created by R.-J. Back – is a development paradigm, in which software is constructed by adding functionality in small increments. The resulting code has an organised, layered structure and can be easily reused. Moreover, the interaction with the users of the software and the correctness concerns are essential elements of the development process, contributing to high quality and functionality of the final product. The paradigm of Stepwise Feature Introduction has been successfully applied in an academic environment, to a number of small-scale developments. The thesis examines the paradigm and its suitability to construction of large and complex software systems by focusing on the development of two software systems of significant complexity. Throughout the thesis we propose a number of improvements and modifications that should be applied to the paradigm when developing or reengineering large and complex software systems. The discussion in the thesis covers various aspects of software development that relate to Stepwise Feature Introduction. More specifically, we evaluate the paradigm based on the common practices of object-oriented programming and design and agile development methodologies. We also outline the strategy to testing systems built with the paradigm of Stepwise Feature Introduction.

Support vector machines in spectral data classification

Työssä käydään läpi tukivektorikoneiden teoreettista pohjaa sekä tutkitaan eri parametrien vaikutusta spektridatan luokitteluun.

Steganografia ja kuvien digitaalinen vesileimaus

Steganografian tarkoituksena on salaisen viestin piilottaminen muun informaation sekaan. Tutkielmassa perehdytään kirjallisuuden pohjalta steganografiaan ja kuvien digitaaliseen vesileimaamiseen. Tutkielmaan kuuluu myös kokeellinen osuus. Siinä esitellään vesileimattujen kuvien tunnistamiseen kehitetty testausjärjestelmä ja testiajojen tulokset. Testiajoissa kuvasarjoja on vesileimattu valituilla vesileimausmenetelmillä parametreja vaihdellen. Tunnistettaville kuville tehdään piirreirrotus. Erotellut piirteet annetaan parametreina luokittimelle, joka tekee lopullisen tunnistamispäätöksen. Tutkimuksessa saatiin toteutettua toimiva ohjelmisto vesileiman lisäämiseen ja vesileimattujen kuvien tunnistamiseen kuvajoukosta. Tulosten perusteella, sopivalla piirreirrottimella ja tukivektorikoneluokittimella päästään yli 95 prosentin tunnistamistarkkuuteen.

Analytic evaluation of three-phase short circuit demagnetization and hysteresis loss risk in rotor-surface-magnet Permanent-Magnet Synchronous Machines

Permanent magnet synchronous machines (PMSM) have become widely used in applications because of high efficiency compared to synchronous machines with exciting winding or to induction motors. This feature of PMSM is achieved through the using the permanent magnets (PM) as the main excitation source. The magnetic properties of the PM have significant influence on all the PMSM characteristics. Recent observations of the PM material properties when used in rotating machines revealed that in all PMSMs the magnets do not necessarily operate in the second quadrant of the demagnetization curve which makes the magnets prone to hysteresis losses. Moreover, still no good analytical approach has not been derived for the magnetic flux density distribution along the PM during the different short circuits faults. The main task of this thesis is to derive simple analytical tool which can predict magnetic flux density distribution along the rotor-surface mounted PM in two cases: during normal operating mode and in the worst moment of time from the PM’s point of view of the three phase symmetrical short circuit. The surface mounted PMSMs were selected because of their prevalence and relatively simple construction. The proposed model is based on the combination of two theories: the theory of the magnetic circuit and space vector theory. The comparison of the results in case of the normal operating mode obtained from finite element software with the results calculated with the proposed model shows good accuracy of model in the parts of the PM which are most of all prone to hysteresis losses. The comparison of the results for three phase symmetrical short circuit revealed significant inaccuracy of the proposed model compared with results from finite element software. The analysis of the inaccuracy reasons was provided. The impact on the model of the Carter factor theory and assumption that air have permeability of the PM were analyzed. The propositions for the further model development are presented.

Scalable Feature Selection Applications for Genome-Wide Association Studies of Complex Diseases

Personalized medicine will revolutionize our capabilities to combat disease. Working toward this goal, a fundamental task is the deciphering of geneticvariants that are predictive of complex diseases. Modern studies, in the formof genome-wide association studies (GWAS) have afforded researchers with the opportunity to reveal new genotype-phenotype relationships through the extensive scanning of genetic variants. These studies typically contain over half a million genetic features for thousands of individuals. Examining this with methods other than univariate statistics is a challenging task requiring advanced algorithms that are scalable to the genome-wide level. In the future, next-generation sequencing studies (NGS) will contain an even larger number of common and rare variants. Machine learning-based feature selection algorithms have been shown to have the ability to effectively create predictive models for various genotype-phenotype relationships. This work explores the problem of selecting genetic variant subsets that are the most predictive of complex disease phenotypes through various feature selection methodologies, including filter, wrapper and embedded algorithms. The examined machine learning algorithms were demonstrated to not only be effective at predicting the disease phenotypes, but also doing so efficiently through the use of computational shortcuts. While much of the work was able to be run on high-end desktops, some work was further extended so that it could be implemented on parallel computers helping to assure that they will also scale to the NGS data sets. Further, these studies analyzed the relationships between various feature selection methods and demonstrated the need for careful testing when selecting an algorithm. It was shown that there is no universally optimal algorithm for variant selection in GWAS, but rather methodologies need to be selected based on the desired outcome, such as the number of features to be included in the prediction model. It was also demonstrated that without proper model validation, for example using nested cross-validation, the models can result in overly-optimistic prediction accuracies and decreased generalization ability. It is through the implementation and application of machine learning methods that one can extract predictive genotype–phenotype relationships and biological insights from genetic data sets.