The traditional tracking approach of forming detections and then associating these detections together is known to perform poorly at low signal-to-noise ratios (SNR). Track-before-detect (TBD) approaches, where the sensor data is used directly as opposed to forming detections, has been shown to perform better than traditional approaches at low SNRs. One recently introduced TBD algorithm is the Quanta Tracking Algorithm that is formed by applying maximum likelihood estimation to the histogram probabilistic multi-target tracker (HPMHT). Quanta has shown promising performance for low SNR scenarios, but the body of literature is small and the evaluations that have been done so far do not consider several practical aspects of using the algorithm in real applications and are difficult to compare to other algorithms due to the SNR definitions used. This paper seeks to address these deficiencies in the literature. A re-derivation of Quanta that corrects some issues present in the original derivation while also integrating extensions from the HPMHT literature will also be presented. These extensions are shown to make Quanta able to correct for errors in the assumed size targets as well as improve estimating the SNR of fluctuating targets. / Thesis / Master of Applied Science (MASc)

No description available.

<p>The domain below the surface in maritime security is hard to monitor with conventional methods, due to the often very noisy environment. In conventional methods the measurements are thresholded in order to distinguish potential targets. This is not always a feasible way of treating measurements. In this thesis a system based on raw measurements, that are not thresholded, is presented in order to track and classify divers with an active sonar. With this system it is possible to detect and track weak targets, even with a signal to noise ratio that often goes below 0 dB.</p><p>The system in this thesis can be divided into three parts: the processing of measurements, the association of measurements to targets and the classification of targets. The processing of measurements is based on a particle filter using Track Before Detect (TBD). Two algorithms for association of measurements, Joint Probabilistic Data Association (JPDA) and Highest Probability Data Association (HPDA), have been implemented. The classification of targets is done using an assumed novel approach. The system is evaluated by doing simulations with approximately 8 hours of recorded data, where divers are present at nine different times. The simulations are done a number of times to catch The classification rate is high and the false alarm rate is low.</p> / <p>Undervattensdomänen är svår att övervaka i marina säkerhetssystem med sedvanliga metoder, på grund av den brusiga miljön. I traditionella metoder trösklas mätningarna för att urskilja potentiella mål. Detta är inte alltid ett godtagbart sätt att behandla mätningar på. I den här rapporten presenteras ett system baserat på behandling av rå mätdata, som inte trösklas, för att spåra och klassificera dykare med en aktiv sonar. Med detta system är det möjligt att detektera och spåra svaga mål, trots att signal till brus förhållandet ofta går under 0 dB.</p><p>Systemet i den här rapporten kan delas upp i tre delar: behandling av mätningar, association av mätningar till mål samt klassificering av mål. Behandlingen av mätningarna görs med ett partikelfilter som använder Track Before Detect (TBD). Två algoritmer för associering av mätningar, Joint Probabilistic Data Association (JPDA) och Highest Probability Data Association (HPDA), har implementerats. Klassificeringen av mål görs med en egenutvecklad metod som inte har hittats i existerande dokumentation. Systemet utvärderas genom att simuleringar görs på ungefär 8 timmar inspelad data, där dykare är närvarande vid nio olika tillfällen. Simuleringarna görs ett antal gånger för att fånga upp stokastiska beteenden. Andelen lyckade klassificeringar är hög och andelen falsklarm är låg.</p>

The domain below the surface in maritime security is hard to monitor with conventional methods, due to the often very noisy environment. In conventional methods the measurements are thresholded in order to distinguish potential targets. This is not always a feasible way of treating measurements. In this thesis a system based on raw measurements, that are not thresholded, is presented in order to track and classify divers with an active sonar. With this system it is possible to detect and track weak targets, even with a signal to noise ratio that often goes below 0 dB. The system in this thesis can be divided into three parts: the processing of measurements, the association of measurements to targets and the classification of targets. The processing of measurements is based on a particle filter using Track Before Detect (TBD). Two algorithms for association of measurements, Joint Probabilistic Data Association (JPDA) and Highest Probability Data Association (HPDA), have been implemented. The classification of targets is done using an assumed novel approach. The system is evaluated by doing simulations with approximately 8 hours of recorded data, where divers are present at nine different times. The simulations are done a number of times to catch The classification rate is high and the false alarm rate is low. / Undervattensdomänen är svår att övervaka i marina säkerhetssystem med sedvanliga metoder, på grund av den brusiga miljön. I traditionella metoder trösklas mätningarna för att urskilja potentiella mål. Detta är inte alltid ett godtagbart sätt att behandla mätningar på. I den här rapporten presenteras ett system baserat på behandling av rå mätdata, som inte trösklas, för att spåra och klassificera dykare med en aktiv sonar. Med detta system är det möjligt att detektera och spåra svaga mål, trots att signal till brus förhållandet ofta går under 0 dB. Systemet i den här rapporten kan delas upp i tre delar: behandling av mätningar, association av mätningar till mål samt klassificering av mål. Behandlingen av mätningarna görs med ett partikelfilter som använder Track Before Detect (TBD). Två algoritmer för associering av mätningar, Joint Probabilistic Data Association (JPDA) och Highest Probability Data Association (HPDA), har implementerats. Klassificeringen av mål görs med en egenutvecklad metod som inte har hittats i existerande dokumentation. Systemet utvärderas genom att simuleringar görs på ungefär 8 timmar inspelad data, där dykare är närvarande vid nio olika tillfällen. Simuleringarna görs ett antal gånger för att fånga upp stokastiska beteenden. Andelen lyckade klassificeringar är hög och andelen falsklarm är låg.

Detection and tracking of signals used in sonar applications in noisy environment is the focus of this thesis. We have concentrated on the low Signal-to-Noise Ratio (SNR) case where the conventional detection methods are not applicable. Furthermore, it is assumed that the duty cycle is relatively low. In the problem that is of concern the carrier frequency, pulse repetition interval (PRI) and the existence of the signal are not known. The unknown character of PRI makes the problem challenging since it means that the signal exists at some unknown intervals. A recursive, Bayesian track-before-detect (TBD) filter using particle filter based methods is proposed to solve the concerned problem. The data used by the particle filter is the magnitude of a complex spectrum in complex Gaussian noise. The existence variable is added in the design of the filter to determine the existence of the signal. The evolution of the signal state is modeled by a linear stochastic process. The filter estimates the signal state including the carrier frequency and PRI. Simulations are done under different scenarios where the carrier frequency, PRI and the existence of the signal varies. The results demonstrate that the algorithm presented in this thesis can detect signals which cannot be detected by conventional methods. Besides detection, the tracking performance of the filter is satisfying.

The Morita-Baylis-Hillman (MBH) reaction is a very versatile synthetic protocol to synthesize various useful compounds containing several functional groups. MBH acetates and carbonates are highly valued compounds as they have good potential to be precursors for organic synthesis reactions due to their ease of modification and synthesis. This thesis utilizes Density Functional Theory (DFT) calculations to understand the mechanism and selectivity of an unexpected tandem conjugate addition-elimination (CA-E) reaction of allylic alkylated Morita-Baylis-Hillman C- adducts. This synthetic protocol was developed by Prof. Zhi-Yong Jiang and co-workers from Henan University, China. The reaction required the use of sub-stoichiometric amounts of an organic or inorganic Brøndst base as a catalyst and was achieved with excellent yields (96%) in neat conditions. TBD gave the highest yield amongst the organocatalysts and Cs2CO3 gave the highest yield amongst all screened bases. A possible mechanistic pathway was proposed and three different energy profiles were modeled using 1,5,7-triaza-bicyclo-[4.4.0]-dec-5-ene (TBD), Cs2CO3 and CO32- as catalysts. All three models were able to explain the experimental observations, revealing both kinetic and thermodynamic factors influencing the selectivity of the CA-E reaction. CO32- model gave the most promising result, revealing a significant energy difference of 17.9 kcal/mol between the transition states of the two differing pathways and an energy difference of 20.9 kcal/mol between the two possible products. Although TBD modeling did not show significant difference in the transition states of the differing pathways, it revealed an unexpected secondary non-covalent electrostatic interaction, involving the electron deficient C atom of the triaza CN3 moiety of the TBD catalyst and the O atom of a neighboring NO2- group in the intermediate. Subsequent modeling using a similar substrate proved the possibility of this non-covalent electrostatic interaction, as there was significant overlap of the orbital cloud present in both the Highest Occupied Molecular Orbital (HOMO) and Lowest Unoccupied Molecular Orbital (LUMO) of the molecule between the C atom of the triaza moiety belonging to the TBD catalyst and the O atom of the nitro group of the substrate. The Mayer bond order was of the C-O interaction was determined to be 0.138.

<p>Integrated tracking and detection, based on unthresholded measurements, also referred to as track before detect (TBD) is a hard nonlinear and non-Gaussian dynamical estimation and detection problem. However, it is a technique that enables the user to track and detect targets that would be extremely hard to track and detect, if possible at all with ''classical'' methods. TBD enables us to be better able to detect and track weak, stealthy or dim targets in noise and clutter and particles filter have shown to be very useful in the implementation of TBD algorithms. </p><p>This Master's thesis has investigated the use of particle filters on radar measurements, in a TBD approach.</p><p>The work has been divided into two major problems, a time efficient implementation and new functional features, as estimating the radar cross section (RCS) and the extension of the target. The later is of great importance when the resolution of the radar is such, that specific features of the target can be distinguished. Results will be illustrated by means of realistic examples.</p>

Integrated tracking and detection, based on unthresholded measurements, also referred to as track before detect (TBD) is a hard nonlinear and non-Gaussian dynamical estimation and detection problem. However, it is a technique that enables the user to track and detect targets that would be extremely hard to track and detect, if possible at all with ''classical'' methods. TBD enables us to be better able to detect and track weak, stealthy or dim targets in noise and clutter and particles filter have shown to be very useful in the implementation of TBD algorithms. This Master's thesis has investigated the use of particle filters on radar measurements, in a TBD approach. The work has been divided into two major problems, a time efficient implementation and new functional features, as estimating the radar cross section (RCS) and the extension of the target. The later is of great importance when the resolution of the radar is such, that specific features of the target can be distinguished. Results will be illustrated by means of realistic examples.

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

The problem of object detection deals with determining whether an instance of a given class of object is present or not. There are robust, supervised learning based algorithms available for object detection in an image. These image object detectors (image-based object detectors) use characteristics learnt from the training samples to find object and non-object regions. The characteristics used are such that the detectors work under a variety of conditions and hence are very robust. Object detection in video can be performed by using such a detector on each frame of the video sequence. This approach checks for presence of an object around each pixel, at different scales. Such a frame-based approach completely ignores the temporal continuity inherent in the video. The detector declares presence of the object independent of what has happened in the past frames. Also, various visual cues such as motion and color, which give hints about the location of the object, are not used. The current work is aimed at building a generic framework for using a supervised learning based image object detector for video that exploits temporal continuity and the presence of various visual cues. We use temporal continuity and visual cues to speed up the detection and improve detection accuracy by considering past detection results. We propose a generic framework, based on Experiential Sampling [1], which considers temporal continuity and visual cues to focus on a relevant subset of each frame. We determine some key positions in each frame, called attention samples, and object detection is performed only at scales with these positions as centers. These key positions are statistical samples from a density function that is estimated based on various visual cues, past experience and temporal continuity. This density estimation is modeled as a Bayesian Filtering problem and is carried out using Sequential Monte Carlo methods (also known as Particle Filtering), where a density is represented by a weighted sample set. The experiential sampling framework is inspired by Neisser’s perceptual cycle [2] and Itti-Koch’s static visual attention model[3]. In this work, we first use Basic Experiential Sampling as presented in[1]for object detection in video and show its limitations. To overcome these limitations, we extend the framework to effectively combine top-down and bottom-up visual attention phenomena. We use learning based detector’s response, which is a top-down cue, along with visual cues to improve attention estimate. To effectively handle multiple objects, we maintain a minimum number of attention samples per object. We propose to use motion as an alert cue to reduce the delay in detecting new objects entering the field of view. We use an inhibition map to avoid revisiting already attended regions. Finally, we improve detection accuracy by using a particle filter based detection scheme [4], also known as Track Before Detect (TBD). In this scheme, we compute likelihood of presence of the object based on current and past frame data. This likelihood is shown to be approximately equal to the product of average sample weights over past frames. Our framework results in a significant reduction in overall computation required by the object detector, with an improvement in accuracy while retaining its robustness. This enables the use of learning based image object detectors in real time video applications which otherwise are computationally expensive. We demonstrate the usefulness of this framework for frontal face detection in video. We use Viola-Jones’ frontal face detector[5] and color and motion visual cues. We show results for various cases such as sequences with single object, multiple objects, distracting background, moving camera, changing illumination, objects entering/exiting the frame, crossing objects, objects with pose variation and sequences with scene change. The main contributions of the thesis are i) We give an experiential sampling formulation for object detection in video. Many concepts like attention point and attention density which are vague in[1] are precisely defined. ii) We combine detector’s response along with visual cues to estimate attention. This is inspired by a combination of top-down and bottom-up attention maps in visual attention models. To the best of our knowledge, this is used for the first time for object detection in video. iii) In case of multiple objects, we highlight the problem with sample based density representation and solve by maintaining a minimum number of attention samples per object. iv) For objects first detected by the learning based detector, we propose to use a TBD scheme for their subsequent detections along with the learning based detector. This improves accuracy compared to using the learning based detector alone. This thesis is organized as follows . Chapter 1: In this chapter we present a brief survey of related work and define our problem. . Chapter 2: We present an overview of biological models that have motivated our work. . Chapter 3: We give the experiential sampling formulation as in previous work [1], show results and discuss its limitations. . Chapter 4: In this chapter, which is on Enhanced Experiential Sampling, we suggest enhancements to overcome limitations of basic experiential sampling. We propose track-before-detect scheme to improve detection accuracy. . Chapter 5: We conclude the thesis and give possible directions for future work in this area. . Appendix A: A description of video database used in this thesis. . Appendix B: A list of commonly used abbreviations and notations.