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.

Radio Frequency Measurement Method for Detecting Bearing Currents in Induction Motors

Induction motors are widely used in industry, and they are generally considered very reliable. They often have a critical role in industrial processes, and their failure can lead to significant losses as a result of shutdown times. Typical failures of induction motors can be classified into stator, rotor, and bearing failures. One of the reasons for a bearing damage and eventually a bearing failure is bearing currents. Bearing currents in induction motors can be divided into two main categories; classical bearing currents and inverter-induced bearing currents. A bearing damage caused by bearing currents results, for instance, from electrical discharges that take place through the lubricant film between the raceways of the inner and the outer ring and the rolling elements of a bearing. This phenomenon can be considered similar to the one of electrical discharge machining, where material is removed by a series of rapidly recurring electrical arcing discharges between an electrode and a workpiece. This thesis concentrates on bearing currents with a special reference to bearing current detection in induction motors. A bearing current detection method based on radio frequency impulse reception and detection is studied. The thesis describes how a motor can work as a “spark gap” transmitter and discusses a discharge in a bearing as a source of radio frequency impulse. It is shown that a discharge, occurring due to bearing currents, can be detected at a distance of several meters from the motor. The issues of interference, detection, and location techniques are discussed. The applicability of the method is shown with a series of measurements with a specially constructed test motor and an unmodified frequency-converter-driven motor. The radio frequency method studied provides a nonintrusive method to detect harmful bearing currents in the drive system. If bearing current mitigation techniques are applied, their effectiveness can be immediately verified with the proposed method. The method also gives a tool to estimate the harmfulness of the bearing currents by making it possible to detect and locate individual discharges inside the bearings of electric motors.

Inactive endosialidase-based detection of bacterial and oncofetal polysialic acid

Polysialic acid is a carbohydrate polymer which consist of N-acetylneuraminic acid units joined by alpha2,8-linkages. It is developmentally regulated and has an important role during normal neuronal development. In adults, it participates in complex neurological processes, such as memory, neural plasticity, tumor cell growth and metastasis. Polysialic acid also constitutes the capsule of some meningitis and sepsis-causing bacteria, such as Escherichia coli K1, group B meningococci, Mannheimia haemolytica A2 and Moraxella nonliquefaciens. Polysialic acid is poorly immunogenic; therefore high affinity antibodies against it are difficult to prepare, thus specific and fast detection methods are needed. Endosialidase is an enzyme derived from the E. coli K1 bacteriophage, which specifically recognizes and degrades polysialic acid. In this study, a novel detection method for polysialic acid was developed based on a fusion protein of inactive endosialidase and the green fluorescent protein. It utilizes the ability of the mutant, inactive endosialidase to bind but not cleave polysialic acid. Sequencing of the endosialidase gene revealed that amino acid substitutions near the active site of the enzyme differentiate the active and inactive forms of the enzyme. The fusion protein was applied for the detection of polysialic acid in bacteria and neuroblastoma. The results indicate that the fusion protein is a fast, sensitive and specific reagent for the detection of polysialic acid. The use of an inactive enzyme as a specific molecular tool for the detection of its substrate represents an approach which could potentially find wide applicability in the specific detection of diverse macromolecules.

Print-capture PCR for detection of tomato begomoviruses from plants and whiteflies

Print-capture (PC) Polymerase chain reaction (PCR) was evaluated as a novel detection method of plant viruses. Tomato (Lycopersicon esculentum) plants infected with begomovirus (fam. Geminiviridae, gen. Begomovirus) and viruliferous whiteflies were used to study the efficiency of the method. Print-capturing steps were carried out using non-charged nylon membrane or filter paper as the solid support for DNA printings. Amplified DNA fragments of expected size were consistently obtained by PCR from infected plants grown in a greenhouse, after direct application of printed materials to the PCR mix. However, virus detection from a single whitefly and from field-grown tomato samples required a high temperature treatment of printed material prior to PCR amplification. Comparison of nylon membrane and filter paper as the solid support revealed the higher efficiency of the nylon membrane. The application of print-capture PCR reduces the chances of false-positive amplification by reducing manipulation steps during preparation of the target DNA. This method maintains all the advantages of PCR diagnosis, such as the high sensitivity and no requirement of radioactive reagents.

Local Features in Image and Video Processing - Object Class Matching and Video Shot Detection

The usage of digital content, such as video clips and images, has increased dramatically during the last decade. Local image features have been applied increasingly in various image and video retrieval applications. This thesis evaluates local features and applies them to image and video processing tasks. The results of the study show that 1) the performance of different local feature detector and descriptor methods vary significantly in object class matching, 2) local features can be applied in image alignment with superior results against the state-of-the-art, 3) the local feature based shot boundary detection method produces promising results, and 4) the local feature based hierarchical video summarization method shows promising new new research direction. In conclusion, this thesis presents the local features as a powerful tool in many applications and the imminent future work should concentrate on improving the quality of the local features.

Effectiveness Of Feature Detection Operators On The Performance Of Iris Biometric Recognition System

Iris Recognition is a highly efficient biometric identification system with great possibilities for future in the security systems area.Its robustness and unobtrusiveness, as opposed tomost of the currently deployed systems, make it a good candidate to replace most of thesecurity systems around. By making use of the distinctiveness of iris patterns, iris recognition systems obtain a unique mapping for each person. Identification of this person is possible by applying appropriate matching algorithm.In this paper, Daugman’s Rubber Sheet model is employed for irisnormalization and unwrapping, descriptive statistical analysis of different feature detection operators is performed, features extracted is encoded using Haar wavelets and for classification hammingdistance as a matching algorithm is used. The system was tested on the UBIRIS database. The edge detection algorithm, Canny, is found to be the best one to extract most of the iris texture. The success rate of feature detection using canny is 81%, False Accept Rate is 9% and False Reject Rate is 10%.

An Interval Intelligent-based Approach for Fault Detection and Modelling

Not considered in the analytical model of the plant, uncertainties always dramatically decrease the performance of the fault detection task in the practice. To cope better with this prevalent problem, in this paper we develop a methodology using Modal Interval Analysis which takes into account those uncertainties in the plant model. A fault detection method is developed based on this model which is quite robust to uncertainty and results in no false alarm. As soon as a fault is detected, an ANFIS model is trained in online to capture the major behavior of the occurred fault which can be used for fault accommodation. The simulation results understandably demonstrate the capability of the proposed method for accomplishing both tasks appropriately

Extraction of tidal channel networks from airborne scanning laser altimetry and aerial photography

The study of the morphodynamics of tidal channel networks is important because of their role in tidal propagation and the evolution of salt-marshes and tidal flats. Channel dimensions range from tens of metres wide and metres deep near the low water mark to only 20-30cm wide and 20cm deep for the smallest channels on the marshes. The conventional method of measuring the networks is cumbersome, involving manual digitising of aerial photographs. This paper describes a semi-automatic knowledge-based network extraction method that is being implemented to work using airborne scanning laser altimetry (and later aerial photography). The channels exhibit a width variation of several orders of magnitude, making an approach based on multi-scale line detection difficult. The processing therefore uses multi-scale edge detection to detect channel edges, then associates adjacent anti-parallel edges together to form channels using a distance-with-destination transform. Breaks in the networks are repaired by extending channel ends in the direction of their ends to join with nearby channels, using domain knowledge that flow paths should proceed downhill and that any network fragment should be joined to a nearby fragment so as to connect eventually to the open sea.

Uses of airborne scanning laser altimetry for river flood modelling and coastal geomorphology

Two ongoing projects at ESSC that involve the development of new techniques for extracting information from airborne LiDAR data and combining this information with environmental models will be discussed. The first project in conjunction with Bristol University is aiming to improve 2-D river flood flow models by using remote sensing to provide distributed data for model calibration and validation. Airborne LiDAR can provide such models with a dense and accurate floodplain topography together with vegetation heights for parameterisation of model friction. The vegetation height data can be used to specify a friction factor at each node of a model’s finite element mesh. A LiDAR range image segmenter has been developed which converts a LiDAR image into separate raster maps of surface topography and vegetation height for use in the model. Satellite and airborne SAR data have been used to measure flood extent remotely in order to validate the modelled flood extent. Methods have also been developed for improving the models by decomposing the model’s finite element mesh to reflect floodplain features such as hedges and trees having different frictional properties to their surroundings. Originally developed for rural floodplains, the segmenter is currently being extended to provide DEMs and friction parameter maps for urban floods, by fusing the LiDAR data with digital map data. The second project is concerned with the extraction of tidal channel networks from LiDAR. These networks are important features of the inter-tidal zone, and play a key role in tidal propagation and in the evolution of salt-marshes and tidal flats. The study of their morphology is currently an active area of research, and a number of theories related to networks have been developed which require validation using dense and extensive observations of network forms and cross-sections. The conventional method of measuring networks is cumbersome and subjective, involving manual digitisation of aerial photographs in conjunction with field measurement of channel depths and widths for selected parts of the network. A semi-automatic technique has been developed to extract networks from LiDAR data of the inter-tidal zone. A multi-level knowledge-based approach has been implemented, whereby low level algorithms first extract channel fragments based mainly on image properties then a high level processing stage improves the network using domain knowledge. The approach adopted at low level uses multi-scale edge detection to detect channel edges, then associates adjacent anti-parallel edges together to form channels. The higher level processing includes a channel repair mechanism.

Evaluation of probabilistic occupancy map people detection for surveillance systems

In this paper, we evaluate the Probabilistic Occupancy Map (POM) pedestrian detection algorithm on the PETS 2009 benchmark dataset. POM is a multi-camera generative detection method, which estimates ground plane occupancy from multiple background subtraction views. Occupancy probabilities are iteratively estimated by fitting a synthetic model of the background subtraction to the binary foreground motion. Furthermore, we test the integration of this algorithm into a larger framework designed for understanding human activities in real environments. We demonstrate accurate detection and localization on the PETS dataset, despite suboptimal calibration and foreground motion segmentation input.

Motion Detection for Video Surveillance

This thesis is related to the broad subject of automatic motion detection and analysis in videosurveillance image sequence. Besides, proposing the new unique solution, some of the previousalgorithms are evaluated, where some of the approaches are noticeably complementary sometimes.In real time surveillance, detecting and tracking multiple objects and monitoring their activities inboth outdoor and indoor environment are challenging task for the video surveillance system. Inpresence of a good number of real time problems limits scope for this work since the beginning. Theproblems are namely, illumination changes, moving background and shadow detection.An improved background subtraction method has been followed by foreground segmentation, dataevaluation, shadow detection in the scene and finally the motion detection method. The algorithm isapplied on to a number of practical problems to observe whether it leads us to the expected solution.Several experiments are done under different challenging problem environment. Test result showsthat under most of the problematic environment, the proposed algorithm shows the better qualityresult.

Two-colour photocurrent detection technique for coherent control of a single InGaAs/GaAs quantum dot

We present a two-colour photocurrent detection method for coherent control of a single InGaAs/GaAs self-assembled quantum dot. A pulse shaping technique provides a high degree of control over picosecond optical pulses. Rabi rotations on the exciton to biexciton transition are presented, and fine structure beating is detected via time-resolved measurements. (c) 2009 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

Label-free DNA detection based on modified conducting polypyrrole films at microelectrodes

A label-free electrochemical detection method for DNA hybridization based on electrostatic modulation of the ion-exchange kinetics of a polypyrrole film deposited at microelectrodes is reported. Synthetic single-stranded 27-mer oligonucleotides (probe) have been immobilized at 2,5-bis(2-thienyl)-N-(3-phosphorylpropyl)pyrrole film formed by electropolymerization on the previously formed polypyrrole layer. The 27- or 18-mer target oligonucleotides were monitored via the electrochemically driven anion exchange of the inner polypyrrole film. The performance of the miniaturized DNA biosensor system was studied in respect to selectivity, sensitivity, reproducibility, and regeneration of the sensor. Control experiments were performed with a noncomplementary target of 27-mer DNA and 12 base-pair mismatched 18-mer sequences, respectively, and did not show any unspecific binding. Under optimized experimental conditions, the label-free electrochemical biosensor enabled the detection limits of 0.16 and 3.5 fmol for the 18- and 2 7-mer DNA strand, respectively. Furthermore, we demonstrate reusability of the electrochemical DNA biosensor after successful recovery of up to 100% of the original signal by regenerating the DNA label-free electrode with 50 mM HCl at room temperature.

Determination of buprofezin, pyridaben, and tebufenpyrad residues by gas chromatography mass-selective detection in clementine citrus

A gas chromatography-mass-selective (GC-MS) detection method to determine buprofezin, pyridaben, and tebufenpyrad on the pulp, peel, and whole fruit of clementines is described. The extraction/partition procedure was performed in one step and no cleanup was necessary with the GC-MS in the SIM-mode pesticide determination. Recovery ranged from 75 to 124% with coefficients of variance ranging between 1 and 13%. The limit of determination was 0.01 mg/kg for all pesticides. The field trials showed a similar degradative behavior for all active ingredients (AI), with a great residue decrease during the first week and stability in the second. Just after treatment buprofezin and tebufenpyrad showed lower residues than the maximum residue limit (MRL) fixed in Italy, while pyridaben was below the MRL after a week.

Dynamic programming approach for road centerline extraction from digital images

This paper presents a dynamic programming approach for semi-automated road extraction from medium-and high-resolution images. This method is a modified version of a pre-existing dynamic programming method for road extraction from low-resolution images. The basic assumption of this pre-existing method is that roads manifest as lines in low-resolution images (pixel footprint> 2 m) and as such can be modeled and extracted as linear features. On the other hand, roads manifest as ribbon features in medium- and high-resolution images (pixel footprint ≤ 2 m) and, as a result, the focus of road extraction becomes the road centerlines. The original method can not accurately extract road centerlines from medium- and high- resolution images. In view of this, we propose a modification of the merit function of the original approach, which is carried out by a constraint function embedding road edge properties. Experimental results demonstrated the modified algorithm's potential in extracting road centerlines from medium- and high-resolution images.