La detection, le suivi de cible et la reconnaissance de primitives visuelles constituent des problèmes fondamentaux de la vision robotique. Ces problématiques sont réputés difficiles et sources de défis. Malgré les progrès en puissance de calcul des machines, le gain en résolution et en fréquence des capteurs, l’état-de-l’art de la vision robotique peine à atteindre des performances en coût d’énergie et en robustesse qu’offre la vision biologique. L’apparition des nouveaux capteurs, appelés "rétines de silicium” tel que le DVS (Dynamic Vision Sensor) et l’ATIS (Asynchronous Time-based Imaging Sensor) reproduisant certaines fonctionnalités des rétines biologiques, ouvre la voie à de nouveaux paradigmes pour décrire et modéliser la perception visuelle, ainsi que pour traiter l’information visuelle qui en résulte. Les tâches de suivi et de reconnaissance de formes requièrent toujours la caractérisation et la mise en correspondance de primitives visuelles. La détection de ces dernières et leur description nécessitent des approches fondamentalement différentes de celles employées en vision robotique traditionnelle. Cette thèse développe et formalise de nouvelles méthodes de détection et de caractérisation de primitives spatio-temporel des signaux acquis par les rétines de silicium (plus communément appelés capteurs “event-based”). Une structure théorique pour les tâches de détection, de suivi, de reconnaissance et de classification de primitives est proposée. Elle est ensuite validée par des données issues de ces capteurs “event-based”,ainsi que par des bases données standard du domaine de la reconnaissance de formes, convertit au préalable à un format compatible avec la representation “événement”. Les résultats présentés dans cette thèse démontrent les potentiels et l’efficacité des systèmes "event-based”. Ce travail fournit une analyse approfondie de différentes méthodes de reconnaissance de forme et de classification “event-based". Cette thèse propose ensuite deux solutions basées sur les primitives. Deux mécanismes d’apprentissage, un purement événementiel et un autre, itératif, sont développés puis évalués pour leur capacité de classification et de robustesse. Les résultats démontrent la validité de la classification “event-based” et souligne l’importance de la dynamique de la scène dans les tâches primordiales de définitions des primitives et de leur détection et caractétisation. / One of the fundamental tasks underlying much of computer vision is the detection, tracking and recognition of visual features. It is an inherently difficult and challenging problem, and despite the advances in computational power, pixel resolution, and frame rates, even the state-of-the-art methods fall far short of the robustness, reliability and energy consumption of biological vision systems. Silicon retinas, such as the Dynamic Vision Sensor (DVS) and Asynchronous Time-based Imaging Sensor (ATIS), attempt to replicate some of the benefits of biological retinas and provide a vastly different paradigm in which to sense and process the visual world. Tasks such as tracking and object recognition still require the identification and matching of local visual features, but the detection, extraction and recognition of features requires a fundamentally different approach, and the methods that are commonly applied to conventional imaging are not directly applicable. This thesis explores methods to detect features in the spatio-temporal information from event-based vision sensors. The nature of features in such data is explored, and methods to determine and detect features are demonstrated. A framework for detecting, tracking, recognising and classifying features is developed and validated using real-world data and event-based variations of existing computer vision datasets and benchmarks. The results presented in this thesis demonstrate the potential and efficacy of event-based systems. This work provides an in-depth analysis of different event-based methods for object recognition and classification and introduces two feature-based methods. Two learning systems, one event-based and the other iterative, were used to explore the nature and classification ability of these methods. The results demonstrate the viability of event-based classification and the importance and role of motion in event-based feature detection.

Faced with imminent spectrum scarcity largely due to inflexible licensed band arrangements, cognitive radio (CR) has been proposed to facilitate higher spectrum utilization by allowing cognitive users (CUs) to access the licensed bands without causing harmful interference to primary users (PUs). To achieve this without the aid of PUs, the CUs have to perform spectrum sensing reliably detecting the presence or absence of PU signals. Without reliable spectrum sensing, the discovery of spectrum opportunities will be inefficient, resulting in limited utilization enhancement. This dissertation examines three major techniques for spectrum sensing, which are matched filter, energy detection, and cyclostationary feature detection. After evaluating the advantages and disadvantages of these techniques, we narrow down our research to a focus on cyclostationary feature detection (CFD). Our first contribution is to boost performance of an existing and prevailing CFD method. This boost is achieved by our proposed optimal and sub-optimal schemes for identifying best hypothesis test points. The optimal scheme incorporates prior knowledge of the PU signals into test point selection, while the sub-optimal scheme circumvents the need for this knowledge. The results show that our proposed can significantly outperform other existing schemes. Secondly, in view of multi-antenna deployment in CR networks, we generalize the CFD method to include the multi-antenna case. This requires effort to justify the joint asymptotic normality of vector-valued statistics and show the consistency of covariance estimates. Meanwhile, to effectively integrate the received multi-antenna signals, a novel cyclostationary feature based channel estimation is devised to obtain channel side information. The simulation results demonstrate that the errors of channel estimates can diminish sharply by increasing the sample size or the average signal-to-noise ratio. In addition, no research has been found that analytically assessed CFD performance over fading channels. We make a contribution to such analysis by providing tight bounds on the average detection probability over Nakagami fading channels and tight approximations of diversity reception performance subject to independent and identically distributed Rayleigh fading. For successful coexistence with the primary system, interference management in cognitive radio networks plays a prominent part. Normally certain average or peak transmission power constraints have to be placed on the CR system. Depending on available channel side information and fading types (fast or slow fading) experienced by the PU receiver, we derive the corresponding constraints that should be imposed. These constraints indicate that the second moment of interference channel gain is an important parameter for CUs allocating transmission power. Hence, we develop a cooperative estimation procedure which provides robust estimate of this parameter based on geolocation information. With less aid from the primary system, the success of this procedure relies on statistically correlated channel measurements from cooperative CUs. The robustness of our proposed procedure to the uncertainty of geolocation information is analytically presented. Simulation results show that this procedure can lead to better mean-square error performance than other existing estimates, and the effects of using inaccurate geolocation information diminish steadily with the increasing number of cooperative cognitive users.

Visualizing flow structures according to the users’ interests provides insight to scientists and engineers. In previous work, a flow structure based on streamline predicates, that examine, whether a streamline has a given property, was defined. Evaluating all streamlines results in characteristic sets grouping all streamlines with similar behavior with respect to a given predicate. Since there are infinitely many streamlines, the algorithm chooses a finite subset for the computation of an approximated discrete version of the characteristic sets. However, even the construction of characteristic sets based on a finite set of streamlines tends to be computationally expensive. Based on a thorough analysis of all processing steps, we present and compare different acceleration approaches. The techniques are based on simplifications that result in characteristic set boundaries deviating from the correct but computational expensive boundaries. We developmeasures for objective comparison of the introduced errors. An adaptive refinement approach turns out to be the best compromise between computation time and quality.

The goal of this paper is to transfer convolution, correlation and Fourier transform to second order tensor fields. Convolution of two tensor fields is defined using matrix multiplication. Convolution of a tensor field with a scalar mask can thus be described by multiplying the scalars with the real unit matrix. The Fourier transform of tensor fields defined in this paper corresponds to Fourier transform of each of the tensor components in the field. It is shown that for this convolution and Fourier transform, the well known convolution theorem holds and optimization in speed can be achieved by using Fast Fourier transform algorithms. Furthermore, pattern matching on tensor fields based on this convolution is described.

A substantial fraction of the automotive assembly comprises formed sheet metal parts. To reduce vehicle weight and improve fuel economy, total sheet metal mass should be minimized without compromising the structural integrity of the vehicle. Excessive deformation contributes to tearing or buckling of the metal, and therefore a forming limit is investigated experimentally to determine the extent to which each particular material can be safely strained. To assess sheet metal formability, this thesis proposes a novel framework for sheet metal surface strain measurement using a scalable dot-grid pattern. Aluminum sheet metal samples are marked with a regular grid of dot-features and imaged with a close-range monocular vision system. After forming, the sheet metal samples are imaged once again to examine the deformation of the surface pattern, and thereby resolve the material strain. Grid-features are localized with sub-pixel accuracy, and then topologically mapped using a novel algorithm for deformation-invariant grid registration. Experimental results collected from a laboratory setup demonstrate consistent robustness under practical imaging conditions. Accuracy, repeatability, and timing statistics are reported for several state-of-the-art feature detectors. / Thesis / Master of Applied Science (MASc)

The aims of this research work are to develop a feature detection, description, and matching system for low-resource applications. This work was motivated by the need for a vision sensor to assist the flight of a quad-rotor UAV. This application presented a real-world challenge of autonomous drift stabilization using vision sensors. The initial solution implemented a basic feature detector and matching system on an FPGA. The research then pursued ways to improve the vision system. Research began with color feature detection, and the Color Difference of Gaussians feature detector was developed. CDoG provides better results than gray scale DoG and does not require any additional processing than gray scale if implemented in a parallel architecture. The CDoG Scale-Invariant Feature Transform modification was developed which provided color feature detection and description to the gray scale SIFT descriptor. To demonstrate the benefits of color information, the CDSIFT algorithm was applied to a real application: library book inventory. While color provides added benefit to the CDSIFT descriptor, CDSIFT descriptors are still computationally intractable for a low-resource hardware implementation. Because of these shortcomings, this research focused on developing a new feature descriptor. The BAsis Sparse-coding Inspired Similarity (BASIS) descriptor was developed with low-resource systems in mind. BASIS utilizes sparse coding to provide a generic description of feature characterstics. The BASIS descriptor provided improved accuracy over SIFT, and similar accuracy to SURF on the task of aerial UAV frame-to-frame feature matching. However, basis dictionaries are non-orthogonal and can contain redundant information. In addition to a feature descriptor, an FPGA-based feature correlation (or matching) system needed to be developed. TreeBASIS was developed to answer this need and address the redundancy issues of BASIS. TreeBASIS utilizes a vocabulary tree to drastically reduce descriptor computation time and descriptor size. TreeBASIS also obtains a higher level of accuracy than SIFT, SURF, and BASIS on the UAV aerial imagery task. Both BASIS and TreeBASIS were implemented in VHDL and are well suited for low-resource FPGA applications. TreeBASIS provides a complete feature detection, description, and correlation system-on-a-chip for low-resource FPGA vision systems.

The advancement of computational fluid dynamics to simulate highly complex fluid flow situations have allowed for simulations that require weeks of computation using expensive high performance clusters. These simulations often generate terabytes of data and hinder the design process by greatly increasing the post-processing time. This research discusses a method to extract shock waves and separation and attachment lines as the simulation is calculating and as a post-processing step. Software agents governed by subjective logic were used to make decisions about extracted features in converging and converged data sets. Two different extraction algorithms were incorporated for shock waves and separation and attachment lines and were tested on four different simulations. A supersonic ramp simulation showed two shock waves at 10% of convergence, but did not reach their final spatial locations until 85% convergence. A similar separation and attachment line analysis was performed on a cylinder in a cross flow simulation. The cylinder separation and attachment lines were within 5% of their final spatial locations at 10% convergence, and at 85% convergence, much of the cylinder and trailing separation and attachment lines showed probability expectation values of approximately 0.90 - 1.00. An Onera M6 wing simulation was used to investigate the belief tuples of the two separate shock waves at full convergence. Probability expectation values of approximately 0.90 - 1.00 were displayed within the two shock waves because they are strong shock waves and because they met the physical requirements of shock waves. A separation and attachment line belief tuple analysis was also performed on a delta wing simulation. The forward portions of these lines showed probability expectation values of approximately 0.90 - 1.00, but dropped to approximately 0.60 - 0.75 as a consequence of their respective vortices breaking down and losing their strength. Similar to shock waves, high probability expectation values meant the separation and attachment lines were strong and physically met separation and attachment line physics. The subjective logic process presented in this research was able to determine which shock waves and separation and attachment lines were most probable, making it easier to view and further investigate these important features.

Mass spectrometry facilitates cutting edge advancements in many fields. Although instrumentation has advanced dramatically in the last 100 years, data processing algorithms have not kept pace. Without sensitive and accurate signal segmentation algorithms, the utility of mass spectrometry is limited. In this dissertation, we provide an overview and analysis of mass spectrometry data processing. A tutorial to ease the learning curve for those outside the field is provided. We draw attention to the lack of critical evaluation in the field and describe the resulting effects, including a glut of algorithm contributions of questionable novel contribution. To facilitate increased critical evaluation, we show the importance of a modular paradigm for mass spectrometry data processing through highlighting the impact of data processing algorithm choice upon experimental results. Our novel controlled vocabulary is presented with the aim of facilitating literature reviews for comparisons. We propose a novel nomenclature and mathematical characterization of mass spectrometry data. We present several novel algorithms for mass spectrometry data segmentation that outperform existing standard approaches. We end with an overview of future research which will continue to advance the state of the art in mass spectrometry data processing.

No / In this paper we present a new method for accurate real-time facial feature detection. Our method is based on local feature detection and enhancement. Previous work in this area, such as that of Viola and Jones, require looking at the face as a whole. Consequently, such approaches have increased chances of reporting negative hits. Furthermore, such algorithms require greater processing power and hence they are especially not attractive for real-time applications. Through our recent work, we have devised a method to identify the face from real-time images and divide it into regions of interest (ROI). Firstly, based on a face detection algorithm, we identify the face and divide it into four main regions. Then, we undertake a local search within those ROI, looking for specific facial features. This enables us to locate the desired facial features more efficiently and accurately. We have tested our approach using the Cohn-Kanade’s Extended Facial Expression (CK+) database. The results show that applying the ROI has a relatively low false positive rate as well as provides a marked gain in the overall computational efficiency. In particular, we show that our method has a 4-fold increase in accuracy when compared to existing algorithms for facial feature detection.

No / The purpose of this chapter is to discuss and present a computational framework for detecting and analysing facial expressions efficiently. The approach here is to identify the face and estimate regions of facial features of interest using the optical flow algorithm. Once the regions and their dynamics are computed a rule based system can be utilised for classification. Using this framework, we show how it is possible to accurately identify and classify facial expressions to match with FACS coding and to infer the underlying basic emotions in real time.