On-road visual vehicle tracking using Markov chain Monte Carlo with metropolis sampling

In this study, a method for vehicle tracking through video analysis based on Markov chain Monte Carlo (MCMC) particle filtering with metropolis sampling is proposed. The method handles multiple targets with low computational requirements and is, therefore, ideally suited for advanced-driver assistance systems that involve real-time operation. The method exploits the removed perspective domain given by inverse perspective mapping (IPM) to define a fast and efficient likelihood model. Additionally, the method encompasses an interaction model using Markov Random Fields (MRF) that allows treatment of dependencies between the motions of targets. The proposed method is tested in highway sequences and compared to state-of-the-art methods for vehicle tracking, i.e., independent target tracking with Kalman filtering (KF) and joint tracking with particle filtering. The results showed fewer tracking failures using the proposed method.

Development of a visual object tracking system using a compressive sensing framework for modeling the appearance

En el presente trabajo se aborda el problema del seguimiento de objetos, cuyo objetivo es encontrar la trayectoria de un objeto en una secuencia de video. Para ello, se ha desarrollado un método de seguimiento-por-detección que construye un modelo de apariencia en un dominio comprimido usando una nueva e innovadora técnica: “compressive sensing”. La única información necesaria es la situación del objeto a seguir en la primera imagen de la secuencia. El seguimiento de objetos es una aplicación típica del área de visión artificial con un desarrollo de bastantes años. Aun así, sigue siendo una tarea desafiante debido a varios factores: cambios de iluminación, oclusión parcial o total de los objetos y complejidad del fondo de la escena, los cuales deben ser considerados para conseguir un seguimiento robusto. Para lidiar lo más eficazmente posible con estos factores, hemos propuesto un algoritmo de tracking que entrena un clasificador Máquina Vector Soporte (“Support Vector Machine” o SVM en sus siglas en inglés) en modo online para separar los objetos del fondo de la escena. Con este fin, hemos generado nuestro modelo de apariencia por medio de un descriptor de características muy robusto que describe los objetos y el fondo devolviendo un vector de dimensiones muy altas. Por ello, se ha implementado seguidamente un paso para reducir la dimensionalidad de dichos vectores y así poder entrenar nuestro clasificador en un dominio mucho menor, al que denominamos domino comprimido. La reducción de la dimensionalidad de los vectores de características se basa en la teoría de “compressive sensing”, que dice que una señal con poca dispersión (pocos componentes distintos de cero) puede estar bien representada, e incluso puede ser reconstruida, a partir de un conjunto muy pequeño de muestras. La teoría de “compressive sensing” se ha aplicado satisfactoriamente en este trabajo y diferentes técnicas de medida y reconstrucción han sido probadas para evaluar nuestros vectores reducidos, de tal forma que se ha verificado que son capaces de preservar la información de los vectores originales. También incluimos una actualización del modelo de apariencia del objeto a seguir, mediante el reentrenamiento de nuestro clasificador en cada cuadro de la secuencia con muestras positivas y negativas, las cuales han sido obtenidas a partir de la posición predicha por el algoritmo de seguimiento en cada instante temporal. El algoritmo propuesto ha sido evaluado en distintas secuencias y comparado con otros algoritmos del estado del arte de seguimiento, para así demostrar el éxito de nuestro método.

Online Learning-based Robust Visual Tracking for Autonomous Landing of Unmanned Aerial Vehicles

Autonomous landing is a challenging and important technology for both military and civilian applications of Unmanned Aerial Vehicles (UAVs). In this paper, we present a novel online adaptive visual tracking algorithm for UAVs to land on an arbitrary field (that can be used as the helipad) autonomously at real-time frame rates of more than twenty frames per second. The integration of low-dimensional subspace representation method, online incremental learning approach and hierarchical tracking strategy allows the autolanding task to overcome the problems generated by the challenging situations such as significant appearance change, variant surrounding illumination, partial helipad occlusion, rapid pose variation, onboard mechanical vibration (no video stabilization), low computational capacity and delayed information communication between UAV and Ground Control Station (GCS). The tracking performance of this presented algorithm is evaluated with aerial images from real autolanding flights using manually- labelled ground truth database. The evaluation results show that this new algorithm is highly robust to track the helipad and accurate enough for closing the vision-based control loop.

A Ground-Truth Video Dataset for the Development and Evaluation of Vision-based Sense-and-Avoid systems

The importance of vision-based systems for Sense-and-Avoid is increasing nowadays as remotely piloted and autonomous UAVs become part of the non-segregated airspace. The development and evaluation of these systems demand flight scenario images which are expensive and risky to obtain. Currently Augmented Reality techniques allow the compositing of real flight scenario images with 3D aircraft models to produce useful realistic images for system development and benchmarking purposes at a much lower cost and risk. With the techniques presented in this paper, 3D aircraft models are positioned firstly in a simulated 3D scene with controlled illumination and rendering parameters. Realistic simulated images are then obtained using an image processing algorithm which fuses the images obtained from the 3D scene with images from real UAV flights taking into account on board camera vibrations. Since the intruder and camera poses are user-defined, ground truth data is available. These ground truth annotations allow to develop and quantitatively evaluate aircraft detection and tracking algorithms. This paper presents the software developed to create a public dataset of 24 videos together with their annotations and some tracking application results.

Online Learning-based Robust Visual Tracking for Autonomous Landing of Unmanned Aerial Vehicles

Autonomous landing is a challenging and important technology for both military and civilian applications of Unmanned Aerial Vehicles (UAVs). In this paper, we present a novel online adaptive visual tracking algorithm for UAVs to land on an arbitrary field (that can be used as the helipad) autonomously at real-time frame rates of more than twenty frames per second. The integration of low-dimensional subspace representation method, online incremental learning approach and hierarchical tracking strategy allows the autolanding task to overcome the problems generated by the challenging situations such as significant appearance change, variant surrounding illumination, partial helipad occlusion, rapid pose variation, onboard mechanical vibration (no video stabilization), low computational capacity and delayed information communication between UAV and Ground Control Station (GCS). The tracking performance of this presented algorithm is evaluated with aerial images from real autolanding flights using manually- labelled ground truth database. The evaluation results show that this new algorithm is highly robust to track the helipad and accurate enough for closing the vision-based control loop.

Particle tracking in a salinity gradient: A method for measuring sinking rate of individual phytoplankton in the laboratory

This paper presents a new method to measure the sinking rates of individual phytoplankton “particles” (cells, chains, colonies, and aggregates) in the laboratory. Conventional particle tracking and high resolution video imaging were used to measure particle sinking rates and particle size. The stabilizing force of a very mild linear salinity gradient (1 ppt over 15 cm) prevented the formation of convection currents in the laboratory settling chamber. Whereas bulk settling methods such as SETCOL provide a single value of sinking rate for a population, this method allows the measurement of sinking rate and particle size for a large number of individual particles or phytoplankton within a population. The method has applications where sinking rates vary within a population, or where sinking rate-size relationships are important. Preliminary data from experiments with both laboratory and field samples of marine phytoplankton are presented here to illustrate the use of the technique, its applications, and limitations. Whereas this paper deals only with sinking phytoplankton, the method is equally valid for positively buoyant species, as well as nonbiological particles.

Video normals from colored lights

We present an algorithm and the associated single-view capture methodology to acquire the detailed 3D shape, bends, and wrinkles of deforming surfaces. Moving 3D data has been difficult to obtain by methods that rely on known surface features, structured light, or silhouettes. Multispectral photometric stereo is an attractive alternative because it can recover a dense normal field from an untextured surface. We show how to capture such data, which in turn allows us to demonstrate the strengths and limitations of our simple frame-to-frame registration over time. Experiments were performed on monocular video sequences of untextured cloth and faces with and without white makeup. Subjects were filmed under spatially separated red, green, and blue lights. Our first finding is that the color photometric stereo setup is able to produce smoothly varying per-frame reconstructions with high detail. Second, when these 3D reconstructions are augmented with 2D tracking results, one can register both the surfaces and relax the homogenous-color restriction of the single-hue subject. Quantitative and qualitative experiments explore both the practicality and limitations of this simple multispectral capture system.

Automatic and adaptable registration of live RGBD video streams sharing partially overlapping views

In this thesis, we introduce DeReEs-4v, an algorithm for unsupervised and automatic registration of two video frames captured depth-sensing cameras. DeReEs-4V receives two RGBD video streams from two depth-sensing cameras arbitrary located in an indoor space that share a minimum amount of 25% overlap between their captured scenes. The motivation of this research is to employ multiple depth-sensing cameras to enlarge the field of view and acquire a more complete and accurate 3D information of the environment. A typical way to combine multiple views from different cameras is through manual calibration. However, this process is time-consuming and may require some technical knowledge. Moreover, calibration has to be repeated when the location or position of the cameras change. In this research, we demonstrate how DeReEs-4V registration can be used to find the transformation of the view of one camera with respect to the other at interactive rates. Our algorithm automatically finds the 3D transformation to match the views from two cameras, requires no human interference, and is robust to camera movements while capturing. To validate this approach, a thorough examination of the system performance under different scenarios is presented. The system presented here supports any application that might benefit from the wider field-of-view provided by the combined scene from both cameras, including applications in 3D telepresence, gaming, people tracking, videoconferencing and computer vision.

Robust indoor speaker recognition in a network of audio and video sensors

Situational awareness is achieved naturally by the human senses of sight and hearing in combination. Automatic scene understanding aims at replicating this human ability using microphones and cameras in cooperation. In this paper, audio and video signals are fused and integrated at different levels of semantic abstractions. We detect and track a speaker who is relatively unconstrained, i.e., free to move indoors within an area larger than the comparable reported work, which is usually limited to round table meetings. The system is relatively simple: consisting of just 4 microphone pairs and a single camera. Results show that the overall multimodal tracker is more reliable than single modality systems, tolerating large occlusions and cross-talk. System evaluation is performed on both single and multi-modality tracking. The performance improvement given by the audio–video integration and fusion is quantified in terms of tracking precision and accuracy as well as speaker diarisation error rate and precision–recall (recognition). Improvements vs. the closest works are evaluated: 56% sound source localisation computational cost over an audio only system, 8% speaker diarisation error rate over an audio only speaker recognition unit and 36% on the precision–recall metric over an audio–video dominant speaker recognition method.

Detection of frontal faces in video streams

[EN]This paper describes an approach for detection of frontal faces in real time (20-35Hz) for further processing. This approach makes use of a combination of previous detection tracking and color for selecting interest areas. On those areas, later facial features such as eyes, nose and mouth are searched based on geometric tests, appearance veri cation, temporal and spatial coherence. The system makes use of very simple techniques applied in a cascade approach, combined and coordinated with temporal information for improving performance. This module is a component of a complete system designed for detection, tracking and identi cation of individuals [1].

Performance analysis in soccer: applications of player tracking technology

The aim of this thesis was threefold, firstly, to compare current player tracking technology in a single game of soccer. Secondly, to investigate the running requirements of elite women’s soccer, in particular the use and application of athlete tracking devices. Finally, how can game style be quantified and defined. Study One compared four different match analysis systems commonly used in both research and applied settings: video-based time-motion analysis, a semi-automated multiple camera based system, and two commercially available Global Positioning System (GPS) based player tracking systems at 1 Hertz (Hz) and 5 Hz respectively. A comparison was made between each of the systems when recording the same game. Total distance covered during the match for the four systems ranged from 10 830 ± 770 m (semi-automated multiple camera based system) to 9 510 ± 740m (video-based time-motion analysis). At running speeds categorised as high-intensity running (>15 km⋅h-1), the semi-automated multiple camera based system reported the highest distance of 2 650 ± 530 m with video-based time-motion analysis reporting the least amount of distance covered with 1 610 ± 370 m. At speeds considered to be sprinting (>20 km⋅h-1), the video-based time-motion analysis reported the highest value (420 ± 170 m) and 1 Hz GPS units the lowest value (230 ± 160 m). These results demonstrate there are differences in the determination of the absolute distances, and that comparison of results between match analysis systems should be made with caution. Currently, there is no criterion measure for these match analysis methods and as such it was not possible to determine if one system was more accurate than another. Study Two provided an opportunity to apply player-tracking technology (GPS) to measure activity profiles and determine the physical demands of Australian international level women soccer players. In four international women’s soccer games, data was collected on a total of 15 Australian women soccer players using a 5 Hz GPS based athlete tracking device. Results indicated that Australian women soccer players covered 9 140 ± 1 030 m during 90 min of play. The total distance covered by Australian women was less than the 10 300 m reportedly covered by female soccer players in the Danish First Division. However, there was no apparent difference in the estimated "#$%&', as measured by multi-stage shuttle tests, between these studies. This study suggests that contextual information, including the “game style” of both the team and opposition may influence physical performance in games. Study Three examined the effect the level of the opposition had on the physical output of Australian women soccer players. In total, 58 game files from 5 Hz athlete-tracking devices from 13 international matches were collected. These files were analysed to examine relationships between physical demands, represented by total distance covered, high intensity running (HIR) and distances covered sprinting, and the level of the opposition, as represented by the Fédération Internationale de Football Association (FIFA) ranking at the time of the match. Higher-ranking opponents elicited less high-speed running and greater low-speed activity compared to playing teams of similar or lower ranking. The results are important to coaches and practitioners in the preparation of players for international competition, and showed that the differing physical demands required were dependent on the level of the opponents. The results also highlighted the need for continued research in the area of integrating contextual information in team sports and demonstrated that soccer can be described as having dynamic and interactive systems. The influence of playing strategy, tactics and subsequently the overall game style was highlighted as playing a significant part in the physical demands of the players. Study Four explored the concept of game style in field sports such as soccer. The aim of this study was to provide an applied framework with suggested metrics for use by coaches, media, practitioners and sports scientists. Based on the findings of Studies 1- 3 and a systematic review of the relevant literature, a theoretical framework was developed to better understand how a team’s game style could be quantified. Soccer games can be broken into key moments of play, and for each of these moments we categorised metrics that provide insight to success or otherwise, to help quantify and measure different methods of playing styles. This study highlights that to date, there had been no clear definition of game style in team sports and as such a novel definition of game style is proposed that can be used by coaches, sport scientists, performance analysts, media and general public. Studies 1-3 outline four common methods of measuring the physical demands in soccer: video based time motion analysis, GPS at 1 Hz and at 5 Hz and semiautomated multiple camera based systems. As there are no semi-automated multiple camera based systems available in Australia, primarily due to cost and logistical reasons, GPS is widely accepted for use in team sports in tracking player movements in training and competition environments. This research identified that, although there are some limitations, GPS player-tracking technology may be a valuable tool in assessing running demands in soccer players and subsequently contribute to our understanding of game style. The results of the research undertaken also reinforce the differences between methods used to analyse player movement patterns in field sports such as soccer and demonstrate that the results from different systems such as GPS based athlete tracking devices and semi-automated multiple camera based systems cannot be used interchangeably. Indeed, the magnitude of measurement differences between methods suggests that significant measurement error is evident. This was apparent even when the same technologies are used which measure at different sampling rates, such as GPS systems using either 1 Hz or 5 Hz frequencies of measurement. It was also recognised that other factors influence how team sport athletes behave within an interactive system. These factors included the strength of the opposition and their style of play. In turn, these can impact the physical demands of players that change from game to game, and even within games depending on these contextual features. Finally, the concept of what is game style and how it might be measured was examined. Game style was defined as "the characteristic playing pattern demonstrated by a team during games. It will be regularly repeated in specific situational contexts such that measurement of variables reflecting game style will be relatively stable. Variables of importance are player and ball movements, interaction of players, and will generally involve elements of speed, time and space (location)".

Feature-based tracking of urethral motion in low-resolution trans-perineal ultrasound

This paper describes a novel algorithm for tracking the motion of the urethra from trans-perineal ultrasound. Our work is based on the structure-from-motion paradigm and therefore handles well structures with ill-defined and partially missing boundaries. The proposed approach is particularly well-suited for video sequences of low resolution and variable levels of blurriness introduced by anatomical motion of variable speed. Our tracking method identifies feature points on a frame by frame basis using the SURF detector/descriptor. Inter-frame correspondence is achieved using nearest-neighbor matching in the feature space. The motion is estimated using a non-linear bi-quadratic model, which adequately describes the deformable motion of the urethra. Experimental results are promising and show that our algorithm performs well when compared to manual tracking.

Feature-based tracking of urethral motion in low-resolution trans-perineal ultrasound

This paper describes a novel algorithm for tracking the motion of the urethra from trans-perineal ultrasound. Our work is based on the structure-from-motion paradigm and therefore handles well structures with ill-defined and partially missing boundaries. The proposed approach is particularly well-suited for video sequences of low resolution and variable levels of blurriness introduced by anatomical motion of variable speed. Our tracking method identifies feature points on a frame by frame basis using the SURF detector/descriptor. Inter-frame correspondence is achieved using nearest-neighbor matching in the feature space. The motion is estimated using a non-linear bi-quadratic model, which adequately describes the deformable motion of the urethra. Experimental results are promising and show that our algorithm performs well when compared to manual tracking.