<p>This master's thesis develops an algorithm for tracking of cars robust enough to handle turning cars. It is implemented in the image processing environment Image Processing Application Programming Interface (IPAPI) for use with the WITAS project. </p><p>Firstly, algorithms, comparable with one currently used in the WITAS-project, are studied. The focus is on how rotation, that originates from the turning of the cars, affects tracking performance. The algorithms studied all perform an exhaustive search over a region, close to the last known position of the object being tracked, to find a match. After this, an iterative algorithm, based on the idea that a car can only rotate, translate and change scale, is introduced. The algorithm estimates the parameters describing this rotation, translation, and change of scale, iteratively. The iterative process needs a initial parameter estimate that is accurate enough for the algorithm to converge. The developed algorithm is based on an earlier publication on the subject, however the mathematical description, and deduction, of it is taken one step further than in this publication. </p><p>The iterative algorithm used performs well under the assumption that the data used fulfills some basic criteria. These demands comprises: placement of camera, template size as well as how the parameters may vary between two observations. The iterative algorithm is also potentially faster than exhaustive search methods, because few iterations are needed when the parameters change slowly. Better initial parameters should improve stability and speed of convergation. Other suggestions that could give better performance is discussed, e.g., methods to better extract the target from the surroundings.</p>

This master's thesis develops an algorithm for tracking of cars robust enough to handle turning cars. It is implemented in the image processing environment Image Processing Application Programming Interface (IPAPI) for use with the WITAS project. Firstly, algorithms, comparable with one currently used in the WITAS-project, are studied. The focus is on how rotation, that originates from the turning of the cars, affects tracking performance. The algorithms studied all perform an exhaustive search over a region, close to the last known position of the object being tracked, to find a match. After this, an iterative algorithm, based on the idea that a car can only rotate, translate and change scale, is introduced. The algorithm estimates the parameters describing this rotation, translation, and change of scale, iteratively. The iterative process needs a initial parameter estimate that is accurate enough for the algorithm to converge. The developed algorithm is based on an earlier publication on the subject, however the mathematical description, and deduction, of it is taken one step further than in this publication. The iterative algorithm used performs well under the assumption that the data used fulfills some basic criteria. These demands comprises: placement of camera, template size as well as how the parameters may vary between two observations. The iterative algorithm is also potentially faster than exhaustive search methods, because few iterations are needed when the parameters change slowly. Better initial parameters should improve stability and speed of convergation. Other suggestions that could give better performance is discussed, e.g., methods to better extract the target from the surroundings.

This thesis describes design and implementation of method, which determines beginning of heart beats in echocardiographic record. Design of this method is built around pyramidal Lucas-Kanade algorithm and fast Fourier transform. This method is implemented in C++ language with OpenCV and FFTW libraries. Analysis of the implementation has shown, that this method is sensitive to anomalies in echocardiographic record. This method is developed as a part of the project for an analysis of echocardiographic records for st. Anna hospital at Brno.

The reliance on GPS for Unmanned Aerial Vehicle (UAV) localization limits the areas of application to places with a stable GPS signal. The emergence of deep learning in computer vision has made deep learning methods for visual UAV navigation a promising candidate for autonomous GPS denied localization. These method locate using images taken by a mounted camera on the UAV. Most works in the field evaluate localization ability in urban environments dense with artificial structures. This thesis analyses the localization ability of one such method over agricultural fields and forests in comparison to urban areas to investigate whether such systems rely on artificial structure or if they can function in a general environment. The localization technique is based on the deep feature Lucas-Kanade algorithm and uses convolutional neural network extracted feature representations of images taken by the UAV and satellite images to place the UAV within the satellite image for a position estimate. A network interpretation method is also applied to the problem to investigate whether it can help explain what causes the potential differences in localization accuracy between the areas. The investigation finds that the localization method is applicable in both forests and agricultural fields and pinpoints other factors than the prevalence of artificial structure that are more important for accurate localization. Further, a potential improvement to the algorithm is proposed that is shown to notably improve localization accuracy in certain conditions. It is based on obtaining a second position estimate by reversing the optimization direction and choosing the better of the two based on a loss function. / Obemannade luftburna fordon (UAV) är generellt beroende av GPS för autonom lokalisering vilket begränsar deras användningsområden till platser med en stabil GPS signal. Framväxten av djupinlärning inom datorseende har gjort djupinlärningsbaserade metoder för visuell UAV navigation en lovande kandidat för UAV lokalisering oberoende av GPS. De flesta vetenskapliga artiklar inom området utvärderar lokaliseringsförmågan i urbana miljöer som är fyllda med artificiella strukturer såsom hus och vägar. I den här uppsatsen analyseras lokaliseringsförmågan av en sådan metod över jordbruksområden och skog i förhållande till urbana miljöer för att undersöka om sådana system är beroende av artificiell struktur för att lokalisera korrekt. Lokaliseringsmetoden är baserad på Lukas-Kanade-algoritmen på djupa repesentationer. Konvolverande neurala nätverk tränas för att extrahera representationer av UAV- och satellitbilder som är mer passande för att bestämma förhållandet mellan kamerapositionerna med Lukas-Kanade algoritmen. En nätverkstolkningsmetod appliceras även på problemet för att undersöka huruvida det kan användas för att förklara eventuella skillnader i lokaliseringsförmåga mellan områdena. Undersökningen finner att lokaliseringsmetoden fungerar väl i jordbruksområden och skog och fastställer andra faktorer som är viktigare för välfungerande lokalisering än förekomsten av artificiella strukturer. Ytterligare föreslås en potentiell förbättring till algoritmen som visas kunna förbättra lokaliseringsnoggrannheten markant i vissa förhållanden. Förbättringen är baserad på att utvinna en andra positionsuppskattning genom att omvända optimeringsriktningen och välja den bättre av de två baserat på en förlustfunktion.

Tracking gaze of an individual provides important information in understanding the behavior of that person. Gaze tracking has been widely used in a variety of applications from tracking consumers gaze fixation on advertisements, controlling human-computer devices, to understanding behaviors of patients with various types of visual and/or neurological disorders such as autism. Gaze pattern can be identified using different methods but most of them require the use of specialized equipments which can be prohibitively expensive for some applications. In this dissertation, we investigate the possibility of using sclera and iris regions captured in a webcam sequence to estimate gaze pattern. The sclera and iris regions in the video frame are first extracted by using an adaptive thresholding technique. The gaze pattern is then determined based on areas of different sclera and iris regions and distances between tracked points along the irises. The technique is novel as sclera regions are often ignored in eye tracking literature while we have demonstrated that they can be easily extracted from images captured by low-cost camera and are useful in determining the gaze pattern. The accuracy and computational efficiency of the proposed technique is demonstrated by experiments with human subjects.

Video stabilization is an important aspect of video processing, especially for handheld devices where unintended camera movement can significantly degrade the resulting recording. This paper investigates four image based methods for video stabilization. The study explores the Lukas-Kanade, Inverse Compositional Lukas-Kanade, Farnebäck Optical Flow, and GMFlow methods, evaluating their sub-pixel accuracy, real-time performance, and robustness to in-frame motion such as a person walking in front of the camera. The results indicate that while all methods achieve sub-pixel precision, real-time execution on a mobile phone is not feasible with the current implementations. Furthermore, the methods exhibit varying levels of difficulty in handling in-frame motion, with RANSAC-based approaches partially compensating for non-camera-induced movement. The paper also discusses the potential of machine learning techniques, represented by GMFlow, in enhancing stabilization quality at the cost of computational complexity. The findings offer valuable insights for the development of more efficient and robust video stabilization solutions.

In the semestral thesis is given basic methods of non-contact measurement heart rate. There is explained Eulerian video magnification method deals with the visualization of the pulse in the videosequence of face. The semestral thesis describes algorithm Viola-Jones face detection in images and algorithm Kanade-Lucas-Tomasi for tracking faces in the videosequence. Part of the work includes design and realization of measurement. There is explained realization of the program and documented execution results, which are discussed. From the results it is designed to guide for optimal recording.

This thesis is devoted to learn detectors by object tracking in video sequence. In this thesis, we discuss methods for object tracking, object detection and online learning and possibilities of their using in sophisticated techniques, which combine object tracking and online learning detectors.

This thesis deals with an automatic counting of cylinders placed on the back of a truck using images taken by a camera mounted above the car. To achieve this goal, an SVM classifier based on HOG image descriptors has been trained to detect the cylinders. Further, a tracking method based on optical flow estimation has been designed to track the cylinders through image sequences. The result of the thesis is an application that counts bottles with precision 93,08 % placed on the truck and visualizes results of the detection.

The topic of this thesis is the recognition and detection of moving object and persons in video sequence and in the static image. Designed application uses the combination of background model for movement detection, histograms of oriented gradients method for person recognition and Lucas-Kanade method for object tracking.