<p>In this thesis, the problem of estimating the clutter spatial intensity function for the multitarget tracking algorithms has been considered. In many scenarios, after the signal detection process, measurement points provided by the sensor (e.g., sonar, infrared sensor, radar) are not distributed uniformly in the surveillance region as assumed by most tracking algorithms. On the other hand, in order to obtain accurate results, the multitarget tracking algorithm requires information about clutter’s spatial intensity. Thus, non-homogeneous clutter spatial intensity has to be estimated from the measurement set and the tracking filter’s output. Also, in order to take advantage of existing tracking algorithms, it is desirable for the clutter estimation method to be integrated into the tracker itself. In this thesis, the clutter is modeled by a non-homogeneous Poisson point (NHPP) process with a spatial intensity function g(z). To calculate the value of the clutter spatial intensity, all we need to do is estimating g(z). First, two new methods for joint spatial clutter intensity estimation and multitarget tracking using the Probability Hypothesis Density (PHD) Filter are presented. Then, based on NHPP process, multitarget multi-Bernoulli processes and set calculus, the approximated Bayesian method is extended to joint the non–homogeneous clutter background estimation and multitarget tracking with standard multitarget tracking algorithms, like the Multiple Hypothesis Tracking (MHT) and the Joint Integrated Probabilistic Data Association (JIPDA) tracker. Finally, a kernel density method is proposed for the clutter spatial intensity estimation problem. Simulation results illustrate the performance of the above algorithms, both in terms of the false track number and the true track initialization speed. All proposed algorithms show the ability to improve the performance of the multitarget tracker in the presence of slowly time varying non–homogeneous clutter background.</p> / Doctor of Philosophy (PhD)

This dissertation addresses two fundamental problems in computer vision—namely, multitarget tracking and event recognition in videos. These problems are challenging because uncertainty may arise from a host of sources, including motion blur, occlusions, and dynamic cluttered backgrounds. We show that these challenges can be successfully addressed by using a multiscale, volumetric video representation, and taking into account various constraints between events offered by domain knowledge. The dissertation presents our two alternative approaches to multitarget tracking. The first approach seeks to transitively link object detections across consecutive video frames by finding the maximum independent set of a graph of all object detections. Two maximum-independent-set algorithms are specified, and their convergence properties theoretically analyzed. The second approach hierarchically partitions the space-time volume of a video into tracks of objects, producing a segmentation graph of that video. The resulting tracks encode rich contextual cues between salient video parts in space and time, and thus facilitate event recognition, and segmentation in space and time. We also describe our two alternative approaches to event recognition. The first approach seeks to learn a structural probabilistic model of an event class from training videos represented by hierarchical segmentation graphs. The graph model is then used for inference of event occurrences in new videos. Learning and inference algorithms are formulated within the same framework, and their convergence rates theoretically analyzed. The second approach to event recognition uses probabilistic first-order logic for reasoning over continuous time intervals. We specify the syntax, learning, and inference algorithms of this probabilistic event logic. Qualitative and quantitative results on benchmark video datasets are also presented. The results demonstrate that our approaches provide consistent video interpretation with respect to acquired domain knowledge. We outperform most of the state-of-the-art approaches on benchmark datasets. We also present our new basketball dataset that complements existing benchmarks with new challenges. / Graduation date: 2011 / Access restricted to the OSU Community at author's request from May 12, 2011 - May 12, 2012

Abstract Localization and tracking of multiple targets is becoming an essential feature of modern communication services and systems. Although necessary in many contexts, such as surveillance and monitoring applications, low-complexity and reliable solutions capable of coping with different degrees of information are not yet available. This thesis deals with different problems that are encountered in localization and tracking applications and aims to establish a broad understanding of multitarget systems ranging from complete active to incomplete passive solutions in dynamic scenarios. Thereby we start by investigating a fully algebraic framework which is proved to be advantageous in dynamic contexts characterized by no a-priori knowledge. Subsequently we extend the approach to improve its robustness versus corrupted observations. Finally we focus on a Bayesian formulation of the passive multitarget tracking (MTT) problem. The Thesis is based on three parts. The first part focuses on a low complexity mathematical representation of the active problem (i.e manifold-based solution). In particular, the spectrum of the matrices used to represent target locations within an algebraic, multidimensional scaling (MDS) based, solution is characterized statistically. In so doing we propose a novel Jacobi-based eigenspace tracking algorithms for Gramian matrices which is shown to be particularly convenient in a multidimensional scaling formulation of the multitarget tracking problem. The second part deals with incomplete-active multitarget scenarios as well as eventual disturbances on the ranging measurements such as bias due to non-line-of-sight conditions. In particular the aforementioned algebraic solution is extended to cope with heterogeneous information and to incorporate eventual knowledge on the confidence of the measurement information. To do so we solve the classical multidimensional scaling (C-MDS) over a novel kernel matrix and show how the intrinsic nature of this formulation allows to deal with heterogeneous information, specifically angle and distance measurements. Finally, the third part focuses on the random finite sets formulation of Bayesian multisensor MTT problem for passive scenarios. In this area a new gating strategy is proposed to lower the computational complexity of the algorithms without compromising their performance. / Tiivistelmä Useiden kohteiden yhtäaikaisesta paikannuksesta ja seurannasta on tulossa olennainen osa nykyaikaisia viestinnän palveluita ja järjestelmiä. Huolimatta siitä, että yhtäaikainen paikannus on erittäin tarpeellinen osa monissa yhteyksissä, kuten valvonnan ja kontrolloinnin sovelluksissa, siihen ei ole olemassa kompleksisuudeltaan alhaista ratkaisua, joka ottaisi huomioon kaiken saatavilla olevan informaation. Väitöskirja käsittelee useiden kohteiden paikannukseen ja seurantaan liittyviä ongelmia, ja se keskittyy antamaan laajan ymmärryksen aktiivisista täydellisistä menetelmistä passiivisiin epätäydellisiin menetelmiin dynaamisissa ympäristöissä. Saavuttaakseen tavoitteen väitöskirjassa esitetään algebrallinen kehys, jonka todistetaan olevan edistyksellinen dynaamisissa ympäristöissä, joissa ei ole ennakkoinformaatiota saatavilla. Seuraavaksi väitöskirja laajentaa esitettyä lähestymistapaa parantamalla sen vakautta vääriä havaintoja vastaan. Lopuksi esitetään bayesialainen formulointi passiiviselle usean kohteen seuranta -ongelmalle (MTT). Väitöskirja on jaettu kolmeen on osaan. Ensimmäinen osa käsittelee aktiivisen ongelman kuvaamista matemaattisesti säilyttäen alhaisen kompleksisuuden. Erityisesti tässä osassa karakterisoidaan tilastollisesti matriisien spektrin käyttäminen kohteiden paikan esittämiseen moniulotteiseen skaalaukseen (MDS) pohjautuvassa menetelmässä. Saavuttaakseen tämän väitöskirja esittää Jacobin ominaisavaruuksiin perustuvan seuranta-algoritmin Gramian matriiseille, joiden osoitetaan olevan erityisen soveltuvia usean kohteen seuraamisongelman kuvaamiseen MDS-menetelmän avulla. Toinen osa käsittelee epätäydellistä aktiivista usean kohteen skenaariota, kuten myös mittausten lopullisia häiriötä, esim. ei-näköyhteyskanavasta johtuvaa harhaa. Edellä mainittu algebrallinen ratkaisu on laajennettu ottamaan huomioon heterogeeninen informaatio sekä tieto mittausdatan luotettavuudesta. Lisäksi tässä osassa esitetään ratkaisu klassiseen moniulotteiseen skaalausongelmaan (C-MDS) esittelemällä uudenlainen ydinmatriisi ja osoitetaan, kuinka tämä mahdollistaa heterogeenisen informaation, tässä tapauksessa kulma-ja etäisyysmittauksien, huomioon ottamisen. Viimeisessä osassa käsitellään äärellisten satunnaisten joukkojen soveltuvuutta bayesialaisen MTT-ongelman ratkaisuun passiivisissa skenaarioissa. Väitöskirja esittää uuden porttistrategian algoritmien kompleksisuuksien pienentämiseksi säilyttäen kuitenkin samalla niiden suorituskyvyn.

Multitarget tracking in the presence of false alarm is a difficult problem to consider. The objective of multitarget tracking is to estimate the number of targets and their states recursively from available observations. At any given time, targets can be born, die and spawn from already existing targets. Sensors can detect these targets with a defined threshold, where normally the observation is influenced by false alarm. Also if the targets are with low signal to noise ratio (SNR) then the targets may not be detected. The Random Finite Set (RFS) filters can be used to solve such multitarget problem efficiently. Specially, one of the best and most widely used RFS based filter is the Probability Hypothesis Density (PHD) filter. The PHD filter approximates the posterior probability density function (PDF) by the first order moment only, where the targets SNR assumed to be much higher. The PHD filter supports targets die, born, spawn and missed-detection by using the well known implementations including Sequential Monte Carlo Probability Hypothesis Density (SMC-PHD) and Gaussian Mixture Probability Hypothesis Density (GM-PHD) methods. The SMC-PHD filter suffers from the well known degeneracy problems while GM-PHD filter may not be suitable for nonlinear and non-Gaussian target tracking problems. It is desirable to have a filter that can provide continuous estimates for any distribution. This is the motivation for the use of B-Splines in this thesis. One of the main focus of the thesis is the B-Spline based PHD (SPHD) filters. The Spline is a well developed theory and been used in academia and industry for more than five decades. The B-Spline can represent any numerical, geometrical and statistical functions and models including the PDF and PHD. The SPHD filter can be applied to linear, nonlinear, Gaussian and non-Gaussian multitarget tracking applications. The SPHD continuity can be maintained by selecting splines with order of three or more, which avoids the degeneracy-related problem. Another important characteristic of the SPHD filter is that the SPHD can be locally controlled, which allow the manipulations of the SPHD and its natural tendency for handling the nonlinear problems. The SPHD filter can be further extended to support maneuvering multitarget tracking, where it can be an alternative to any available PHD filter implementations. The PHD filter does not work well for very low observable (VLO) target tracking problems, where the targets SNR is normally very low. For very low SNR scenarios the PDF must be approximated by higher order moments. Therefore the PHD implementations may not be suitable for the problem considered in this thesis. One of the best estimator to use in VLO target tracking problem is the Maximum-Likelihood Probability Data Association (ML-PDA) algorithm. The standard ML-PDA algorithm is widely used in single target initialization or geolocation problems with high false alarm. The B-Spline is also used in the ML-PDA (SML-PDA) implementations. The SML-PDA algorithm has the capability to determine the global maximum of ML-PDA log-likelihood ratio with high efficiency in terms of state estimates and low computational complexity. For fast passive track initialization, search and rescue operations the SML-PDA algorithm can be used more efficiently compared to the standard ML-PDA algorithm. Also the SML-PDA algorithm with the extension supports the multitarget tracking. / Thesis / Doctor of Philosophy (PhD)