V rámci této práce byl navržen a implementován systém pro analýzu dopravy z videa. Tento system umožňuje detekovat, sledovat a klasifikovat automobily. Systém je schopný detekovat pruhy z pohybu projíždějících automobilů a také je možné určit, zdali daný automobil jede v protisměru. Rychlost projíždějících automobilů je také měřena. Pro funkčnost systému není vyžadován žadný manuální vstup nebo kalibrace kamery, jelikož kamera je plně automacky zkalibrována pomocí úběžníků. Navržený systém pracuje s velkou přesností detekce, sledování a klasifikace automobilů a také rychlost automobilů je měřena s~malou chybou. Systém je schopný pracovat v reálném čase a je aktuálně využíván pro nepřetržité online sledování dopravy. Největším přínosem této práce je plně automatické měření rychlostí projíždějích vozidel.

In this modern era, land transports are increasing dramatically. Moreover, self-driven car or the Advanced Driving Assistance System (ADAS) is now the public demand. For these types of cars, road conditions detection is mandatory. On the other hand, compared to the number of vehicles, to increase the number of roads is not possible. Software is the only alternative solution. Road Conditions Detection system will help to solve the issues. For solving this problem, Image processing, and machine learning have been applied to develop a project namely, Detection of Road Conditions Using Image Processing and Machine Learning Techniques for Situation Awareness. Many issues could be considered for road conditions but the main focus will be on the detection of potholes, Maintenance sings and lane. Image processing and machine learning have been combined for our system for detecting in real-time. Machine learning has been applied to maintains signs detection. Image processing has been applied for detecting lanes and potholes. The detection system will provide a lane mark with colored lines, the pothole will be a marker with a red rectangular box and for a road Maintenance sign, the system will also provide information of aintenance sign as maintenance sing is detected. By observing all these scenarios, the driver will realize the road condition. On the other hand situation awareness is the ability to perceive information from it’s surrounding, takes decisions based on perceived information and it makes decision based on prediction.

Detection of multiple-lane markings (lane-line) on road surfaces is an essential aspect of autonomous vehicles. Although several approaches have been proposed to detect lanes, detecting multiple lane-lines consistently, particularly across a stream of frames and under varying lighting conditions is still a challenging problem. Since the road's markings are designed to be smooth and parallel, lane-line sampled features tend to be spatially and temporally correlated inside and between frames. In this thesis, we develop novel methods to model these spatial and temporal dependencies in the form of the target tracking problem. In fact, instead of resorting to the conventional method of processing each frame to detect lanes only in the space domain, we treat the overall problem as a Multiple Extended Target Tracking (METT) problem. In the first step, we modelled lane-lines as multiple "independent" extended targets and developed a spline mathematical model for the shape of the targets. We showed that expanding the estimations across the time domain could improve the result of estimation. We identify a set of control points for each spline, which will track over time. To overcome the clutter problem, we developed an integrated probabilistic data association fi lter (IPDAF) as our basis, and formulated a METT algorithm to track multiple splines corresponding to each lane-line.In the second part of our work, we investigated the coupling between multiple extended targets. We considered the non-parametric case and modeled target dependency using the Multi-Output Gaussian Process. We showed that considering dependency between extended targets could improve shape estimation results. We exploit the dependency between extended targets by proposing a novel recursive approach called the Multi-Output Spatio-Temporal Gaussian Process Kalman Filter (MO-STGP-KF). We used MO-STGP-KF to estimate and track multiple dependent lane markings that are possibly degraded or obscured by traffic. Our method tested for tracking multiple lane-lines but can be employed to track multiple dependent rigid-shape targets by using the measurement model in the radial space In the third section, we developed a Spatio-Temporal Joint Probabilistic Data Association Filter (ST-JPDAF). In multiple extended target tracking problems with clutter, sometimes extended targets share measurements: for example, in lane-line detection, when two-lane markings pass or merge together. In single-point target tracking, this problem can be solved using the famous Joint Probabilistic Data Association (JPDA) filter. In the single-point case, even when measurements are dependent, we can stack them in the coupled form of JPDA. In this last chapter, we expanded JPDA for tracking multiple dependent extended targets using an approach called ST-JPDAF. We managed dependency of measurements in space (inside a frame) and time (between frames) using different kernel functions, which can be learned using the trained data. This extension can be used to track the shape and dynamic of dependent extended targets within clutter when targets share measurements. The performance of the proposed methods in all three chapters are quanti ed on real data scenarios and their results are compared against well-known model-based, semi-supervised, and fully-supervised methods. The proposed methods offer very promising results. / Thesis / Doctor of Philosophy (PhD)

Lane detection is a crucial task in the field of autonomous driving and advanced driver assistance systems. In recent years, convolutional neural networks (CNNs) have been the primary approach for solving this problem. However, interesting findings from recent research works regarding the use of Transformer models and attention-based mechanisms have shown to be beneficial in the task of semantic segmentation of the road lane markings. In this work, we investigate the effectiveness of incorporating a Vision Transformer (ViT) to process feature maps extracted by a CNN network for lane detection. We compare the performance of a baseline CNN-based lane detection model with that of a hybrid CNN-ViT pipeline and test the model over a well known dataset. Furthermore, we explore the impact of incorporating temporal information from a road scene on a lane detection model’s predictive performance. We propose a post-processing technique that utilizes information from previous frames to improve the accuracy of the lane detection model. Our results show that incorporating temporal information noticeably improves the model’s performance, and manages to make effective corrections over the originally predicted lane masks. Our SegNet backbone, exploiting the proposed post-processing mechanism, reached an F1 scoreof 0.52 and Intersection-over-Union (IoU) of 0.36 over the TuSimple test set. However, the findings from the testing of our CNN-ViT pipeline and a relevant ablation study, do indicate that this hybrid approach might not be a good fit for lane detection. More specifically, the ViT module fails to exploit the feature sextracted by our CNN backbone and therefore, our hybrid pipeline results in less accurate lane marking spredictions.

This thesis describes the design and subsequent implementation of a unit for traffic monitoring using Raspberry PI. First section provides a quick overview of assistance systems, which use a road lane detection techniques. Next there is a description of two diferent methods for road lane detection. Follow the description of monitoring scene. Then the work describe the practical part including the design and realization of supporting electronics, selecting of each components, including the modifying of cameras, mechanical design and creating of unit. Another section is about selection and installation of appropriate software components necessary for running of the unit and the selection of development tools for creating user application. After description of graphical user interafce, there is a description of road lanes detection algorithm. At the end of the thesis is summarized a reliability of unit in real traffic situation. At the appendix there are technical drawings, describing the unit.

A vehicles ability to detect and estimate its surroundings is important for ensuring the safety of the vehicle and passengers regardless of the level of vehicle autonomy. With an improved road and lane estimation, advanced driver-assistance systems will be able to provide earlier and more accurate warnings and actions to prevent a possible accident. Current lane boundary estimations rely on camera and inertial sensor data to detect and estimate relevant lane boundaries in the vehicles surroundings. The current lane boundary estimation system struggles to provide correct estimations at distances exceeding 75 meters and has a performance which is affected by environmental effects. The methods in this thesis show how map data, together with sensor fusion with radar, camera, inertial measurement unit and global navigation satellite system data is able to provide an improvement to the lane boundary estimations. The map based estimation system is implemented and evaluated for high speed roads (highways and country roads) where lane boundary estimations for distances above 75 meters are needed. The results are conducted in a simulate environment and show how the map based system is able to correct unreliable sensor input to provide more precise boundary estimations. The map based system is also able to provide an up to 36% relative increase in correctly identified objects within ego vehicles lane between 12.5-150 meters in front of ego vehicle. The results indicate the ability to extend the horizon in which driver-assistance functions are able to operate, thus increasing the safety of future autonomous or semi-autonomous vehicles. Future work within the subject is needed to apply map based estimations on urban areas. The precision of such an system also relies on precise positional data. Incorporation of more precise global navigation data would be able to show an increased performance. / Ett fordons förmåga att upptäcka och uppskatta sin omgivning är viktig för att säkerställa fordonets och passagerarnas säkerhet oavsett fordonets autonominivå. Med en förbättrad väg- och körfältsuppskattning kommer avancerade förarassistanssystem att kunna ge tidigare och mer exakta varningar och åtgärder för att förhindra en eventuell olycka. Aktuella estimeringar av körfältsgränser är beroende av kamera och tröghetssensordata för att upptäcka och uppskatta relevanta körfältsgränser i fordonets omgivning. Det nuvarande estimerings-systemet upvisar inkorrekta uppskattningar på avstånd över 75 meter och har en prestanda som påverkas av den omgivande miljön. Metoderna i detta examensarbete visar hur kartdata, tillsammans med sensorfusion av radar, kamera, tröghetsmätenhet och globala satellitnavigeringsdata, kan ge en förbättrad estimering av körfältsgränser. Det kartbaserade systemet är implementerat och utvärderat för höghastighetsvägar (motorvägar och landsvägar) där estimeringar av körfältsgränser för avstånd över 75 meter behövs. Resultaten utförs i en simulerad miljö och visar hur det kartbaserade systemet kan korrigera opålitlig sensorinmatning för att ge mer exakta gränsuppskattningar. Systemet kan också ge en upp till 36% relativ ökning av korrekt identifierade objekt inom ego-fordonets körfält mellan 12.5-150 meter framför ego-fordonet. Resultaten indikerar förmågan att förlänga horisonten som förarassistansfunktioner kan fungera i, vilket ökar säkerheten för framtida autonoma eller halvautonoma fordon. Framtida arbeten inom ämnet behövs för att tillämpa kartbaserade uppskattningar på tätorter. Precisionen hos ett sådant system är också beroende av mer exakt positionsdata. Inkorporering av mer exakt global navigationsdata skulle i detta fall kunna visa en ökad sytemprestanda.