<p>Shape-Based matching (SBM) is a known method for 2D object recognition that is rather robust against illumination variations, noise, clutter and partial occlusion.</p><p>The objects to be recognized can be translated, rotated and scaled.</p><p>The translation of an object is determined by evaluating a similarity measure for all possible positions (similar to cross correlation).</p><p>The similarity measure is based on dot products between normalized gradient directions in edges.</p><p>Rotation and scale is determined by evaluating all possible combinations, spanning a huge search space.</p><p>A resolution pyramid is used to form a heuristic for the search that then gains real-time performance.</p><p>For SBM, a model consisting of normalized edge gradient directions, are constructed for all possible combinations of rotation and scale.</p><p>We have avoided this by using (bilinear) interpolation in the search gradient map, which greatly reduces the amount of storage required.</p><p>SBM is highly parallelizable by nature and with our suggested improvements it becomes much suited for running on a GPU.</p><p>This have been implemented and tested, and the results clearly outperform those of our reference CPU implementation (with magnitudes of hundreds).</p><p>It is also very scalable and easily benefits from future devices without effort.</p><p>An extensive evaluation material and tools for evaluating object recognition algorithms have been developed and the implementation is evaluated and compared to two commercial 2D object recognition solutions.</p><p>The results show that the method is very powerful when dealing with the distortions listed above and competes well with its opponents.</p>

ABSTRACT EVALUATION OF MULTI TARGET TRACKING ALGORITHMS IN THE PRESENCE OF CLUTTER G&uuml / ner, Onur M.S., Department of Electrical and Electronics Engineering Supervisor: Prof. Dr. Mustafa Kuzuoglu August 2005, 88 Pages This thesis describes the theoretical bases, implementation and testing of a multi target tracking approach in radar applications. The main concern in this thesis is the evaluation of the performance of tracking algorithms in the presence of false alarms due to clutter. Multi target tracking algorithms are composed of three main parts: track initiation, data association and estimation. Two methods are proposed for track initiation in this work. First one is the track score function followed by a threshold comparison and the second one is the 2/2 &amp / M/N method which is based on the number of detections. For data association problem, several algorithms are developed according to the environment and number of tracks that are of interest. The simplest method for data association is the nearest-neighbor data association technique. In addition, the methods that use multiple hypotheses like probabilistic data association and joint probabilistic data association are introduced and investigated. Moreover, in the observation to track assignment, gating is an important issue since it reduces the complexity of the computations. Generally, ellipsoidal gates are used for this purpose. For estimation, Kalman filters are used for state prediction and measurement update. In filtering, target kinematics is an important point for the modeling. Therefore, Kalman filters based on different target kinematic models are run in parallel and the outputs of filters are combined to yield a single solution. This method is developed for maneuvering targets and is called interactive multiple modeling (IMM). All these algorithms are integrated to form a multi target tracker that works in the presence (or absence) of clutter. Track score function, joint probabilistic data association (JPDAF) and interactive multiple model filtering are used for this purpose. Keywords: clutter, false alarms, track initiation, data association, gating, target kinematics, IMM, JPDAF

During processing of data received by an airborne radar one of the issues is that the typical signal echo from the ground produces a large perturbation. Due to this perturbation it can be difficult to detect targets with low velocity or a low signal-to-noise ratio. Therefore, a filtering process is needed to separate the large perturbation from the target signal. The traditional method include a tapered Fourier transform that operates in parallel with a MTI filter to suppress the main spectral peak in order to produce a smoother spectral output. The difference between a typical signal echo produced from an object in the environment and the signal echo from the ground can be of a magnitude corresponding to more than a 60 dB difference. This thesis presents research of how the multitaper approach can be utilized in concurrence with the minimum variance estimation technique, to produce a spectral estimation that strives for a more effective clutter suppression. A simulation model of the ground clutter was constructed and also a number of simulations for the multitaper, minimum variance estimation technique was made. Compared to the traditional method defined in this thesis, there was a slight improvement of the improvement factor when using the multitaper approach. An analysis of how variations of the multitaper parameters influence the results with respect to minimum detectable velocity and improvement factor have been carried out. The analysis showed that a large number of time samples, a large number of tapers and a narrow bandwidth provided the best result. The analysis is based on a full factorial simulation that provides insight of how to choose the DPSS parameters if the method is to be implemented in a real radar system.

abstract: Tracking a time-varying number of targets is a challenging dynamic state estimation problem whose complexity is intensified under low signal-to-noise ratio (SNR) or high clutter conditions. This is important, for example, when tracking multiple, closely spaced targets moving in the same direction such as a convoy of low observable vehicles moving through a forest or multiple targets moving in a crisscross pattern. The SNR in these applications is usually low as the reflected signals from the targets are weak or the noise level is very high. An effective approach for detecting and tracking a single target under low SNR conditions is the track-before-detect filter (TBDF) that uses unthresholded measurements. However, the TBDF has only been used to track a small fixed number of targets at low SNR. This work proposes a new multiple target TBDF approach to track a dynamically varying number of targets under the recursive Bayesian framework. For a given maximum number of targets, the state estimates are obtained by estimating the joint multiple target posterior probability density function under all possible target existence combinations. The estimation of the corresponding target existence combination probabilities and the target existence probabilities are also derived. A feasible sequential Monte Carlo (SMC) based implementation algorithm is proposed. The approximation accuracy of the SMC method with a reduced number of particles is improved by an efficient proposal density function that partitions the multiple target space into a single target space. The proposed multiple target TBDF method is extended to track targets in sea clutter using highly time-varying radar measurements. A generalized likelihood function for closely spaced multiple targets in compound Gaussian sea clutter is derived together with the maximum likelihood estimate of the model parameters using an iterative fixed point algorithm. The TBDF performance is improved by proposing a computationally feasible method to estimate the space-time covariance matrix of rapidly-varying sea clutter. The method applies the Kronecker product approximation to the covariance matrix and uses particle filtering to solve the resulting dynamic state space model formulation. / Dissertation/Thesis / Doctoral Dissertation Electrical Engineering 2015

The thesis performs an analysis on visuospatial complexity of dynamic scenes, and morespecifically driving scenes in the propose of gaining a knowledge on human visual perception of the visual information present in a typical driving scene. The analysis and measurement of visual complexity is performed by utilizing two different measure modelsfor measuring visual clutter, Feature congestion clutter measure [1] and Subband entropyclutter measure[1] introduced by Rosenholtz, a cognitive science and research. The thesisrepresent the performance of the computational models on a data set consisting of sixepisodes that simulate driving scenes with different settings and combination of visualfeatures. The results of evaluating the measure models are used to introduce a formulafor measuring visual complexity of annotated images by extracting valuable informationfrom the annotated data set using Scalabel[2], an annotation web- based open source tool.

Visuospatial complexity refers to the level of detail or intricacy present within a scene, takinginto account both spatial and visual properties of the dynamic scene or the place (e.g.moving images, everyday driving, video games and other immersive media). There havebeen several studies on measuring visual complexity from various viewpoints, e.g. marketing,psychology, computer vision and cognitive science. This research project aims atanalysing and evaluating different models and tools that have been developed to measurelow-level features of visuospatial complexity such as Structural Similarity Index measurement,Feature Congestion measurement of clutter and Subband Entropy measurement ofclutter. We use two datasets, one focusing on (reflectional) symmetry in static images,and another that consists of real-world driving videos. The results of the evaluation showdifferent correlations between the implemented models such that the nature of the sceneplays a significant role.

No description available.

Thesis (MScEng (Electrical and Electronic Engineering))--University of Stellenbosch, 2009. / ENGLISH ABSTRACT: Sonars are used extensively for underwater sensing and recent advances in forward-looking imaging sonar have made this type of sonar an appropriate choice for use on Autonomous Underwater Vehicles. The images received from these sonar do however, tend to be noisy and when used in shallow water contain strong bottom reflections that obscure returns from actual targets. The focus of this work was the investigation and development of post-processing techniques to enable the successful use of the sonar images for automated navigation. The use of standard image processing techniques for noise reduction and background estimation, were evaluated on sonar images with varying amounts of noise, as well as on a set of images taken from an AUV in a harbour. The use of multiple background removal and noise reduction techniques on a single image was also investigated. To this end a performance measure was developed, based on the dynamic range found in the image and the uniformity of returned targets. This provided a means to quantitatively compare sets of post-processing techniques and identify the “optimal” processing. The resultant images showed great improvement in the visibility of target areas and the proposed techniques can significantly improve the chances of correct target extraction. / AFRIKAANSE OPSOMMING: Sonars word algemeen gebruik as onderwater sensors. Onlangse ontwikkelings in vooruit-kykende sonars, maak hierdie tipe sonar ’n goeie keuse vir die gebruik op ’n Outomatiese Onderwater Voertuig. Die beelde wat ontvang word vanaf hierdie sonar neig om egter raserig te wees, en wanneer dit in vlak water gebruik word toon dit sterk bodemrefleksies, wat die weerkaatsings van regte teikens verduister. Die fokus van die werk was die ondersoek en ontwikkeling van naverwerkings tegnieke, wat die sonar beelde bruikbaar maak vir outomatiese navigasie. Die gebruik van standaard beeldverwerkingstegnieke vir ruis-onderdrukking en agtergrond beraming, is geëvalueer aan die hand van sonar beelde met verskillende hoeveelhede ruis, asook aan die hand van ’n stel beelde wat in ’n hawe geneem is. Verdere ondersoek is ingestel na die gebruik van meer as een agtergrond beramings en ruis onderdrukking tegniek op ’n enkele beeld. Hierdie het gelei tot die ontwikkeling van ’n maatstaf vir werkverrigting van toegepaste tegnieke. Hierdie maatstaf gee ’n kwantitatiewe waardering van die verbetering op die oorspronklike beeld, en is gebaseer op die verbetering in dinamiese bereik in die beeld en die uniformiteit van die teiken se weerkaatsing. Hierdie maatstaf is gebruik vir die vergelyking van verskeie tegnieke, en identifisering van die “optimale” verwerking. Die verwerkte beelde het ’n groot verbetering getoon in die sigbaarheid van teikens, en die voorgestelde tegnieke kan ’n betekenisvolle bedrae lewer tot die suksesvolle identifisering van obstruksies.

This dissertation considers radar detection and tracking of weak fluctuating targets using dynamic programming (DP) based track-before-detect (TBD). TBD combines target detection and tracking by integrating data over consecutive scans before making a decision on the presence of a target. A novel algorithm is proposed which employs order statistics in dynamic programming based TBD (OS-DP-TBD) to detect weak fluctuating targets. The well-known Swerling type 0, 1 and 3 targets are considered with non-Gaussian distributed clutter and complex Gaussian noise. The clutter is modeled using the Weibull, K and G0 distributions. The proposed algorithm is shown to provide better performance than well-known techniques in the literature. In addition, a novel expanding window multiframe (EW-TBD) technique is presented to improve the detection performance with reasonable computational complexity compared to batch processing. It is shown that EW-TBD has lower complexity than existing multiframe processing techniques. Simulation results are presented which confirm the superiority of the proposed expanding window technique in detecting targets even when they are not present in every scan in the window. Further, the throughput of the proposed technique is higher than with batch processing. Depending on the range and azimuth resolution of the radar system, the target may appear as a point in some radar systems and there will be target energy spillover in other systems. This dissertation considers both extended targets with different energy spillover levels and point targets. Simulation results are presented which confirm the superiority of the proposed algorithm in both cases. / Graduate

L’incidence rasante est un problème spécifique, qui apparaît notamment lorsqu'une antenne est placée sur un mât (télécommunications, défense…) ou sur la côte (surveillance environnementale ou militaire de l'espace maritime). Elle rend la modélisation du problème de diffraction difficile, de par le faible niveau de rétrodiffusion et l’importance de phénomènes complexes comme la diffusion multiple. La question reste importante même si l’écho est très faible, puisqu’il est potentiellement suffisant pour perturber le bon fonctionnement de systèmes antennaires microondes sur un navire. Il porte de plus des informations intéressantes sur l’état de la mer, comme cela a été démontré aux bandes HF et VHF. Un modèle rigoureux de diffraction tridimensionnelle précédemment développé est étendu au calcul des quatre polarisations fondamentales (polarisations Horizontale et Verticale des ondes incidente et réfléchie). Il permet désormais de prendre en compte la conductivité finie de la surface, point crucial dans le cas d'une polarisation incidente Verticale. L’opérateur hyper-singulier impliqué dans l’équation intégrale discrétisée par la méthode des moments est étudié pour évaluer la précision des calculs numériques.Les méthodes approchées de diffraction permettent des calculs numériques beaucoup plus rapides, et sont donc en pratique incontournables. Le modèle rigoureux est donc utilisé, en conjonction avec des données expérimentales, pour servir de référence permettant d’étudier la précision, en incidence rasante et dans le cas de la surface de la mer, de ces méthodes approchées. Nous étudions en particulier la méthode à deux échelles GOSSA, et proposons une correction à son comportement aux angles rasants.Le mouvement de la surface de la mer crée un décalage de fréquence radar dans l’onde rétrodiffusée (effet Doppler), décalage mesuré expérimentalement et que l’algorithme de méthode des moments permet de simuler. Nous étudions par des simulations bidimensionnelles l’évolution du décalage Doppler micro onde avec l’incidence, et l’influence des nonlinéarités de la surface de la mer. Le comportement limite en incidence rasante est précisé, et les contributions respectives des phénomènes électromagnétiques et hydrodynamiques discutées. / The grazing incidence is a specific problem, which appears especially when an antenna is placed on a mast (telecommunications, defence...) or on the coast (environmental or military maritime spatial monitoring). The modelization of the scattering process in such a configuration is difficult, due to low backscattering and to the importance of complex phenomena such as multiple scattering. The issue remains important even if the echo is very low, because it is potentially sufficient to disturb the proper functioning of microwave antenna systems on a ship. Moreover, it carries interesting informations about the sea state, as was demonstrated in HF and VHF bands.A rigorous model of the three-dimensional scattering process, previously developed, is extended to the computation of the scattered fiel in the four fundamental polarizations (Horizontal and Vertical polarization of incident and reflected waves). It is now possible to take into account the finite conductivity of the surface, a crucial point when the incident field is vertically polarized. The hyper- singular operator involved in the integral equation discretized by the method of moments is studied to evaluate the accuracy of numerical calculations.The approximate methods of diffraction allow much faster numerical calculations, and are therefore essential. The rigorous model is used in conjunction with experimental data, as a reference to study the accuracy of such approximate methods, in the case of the sea surface at grazing incidence. We study in particular the two-scale method GOSSA and propose a correction to its behaviour at grazing angles. The motion of the sea surface creates a frequency shift in the radar backscattered wave (Doppler effect). This offset can be measured experimentally, our algorithm allows us to simulate it. We proceed to two-dimensional simulations showing the evolution of the Doppler shift with respect to the grazing angle, and show the influence of the nonlinearities in the sea model. The limit of the mean Doppler shift at very low grazing angles is studied, and the respective contributions of electromagnetic and hydrodynamic phenomena are discussed.