|
Face Detection
Štrba, Miroslav ; Juránek, Roman (referee) ; Hradiš, Michal (advisor)
This bachelor thesis contains overview of actual face detection methods using classifier. It also contains description of creating system for face detection. There are described different methods for classifier training in first part. There is analysis, which preceded creation of system focused on black-and-white picture, in second part. Implemented system is using WaldBoost algorithm and Haar features. There is option to use particle filter in video.
|
| |
|
Parametrization of Image Point Neighborhood
Zamazal, Zdeněk ; Bařina, David (referee) ; Zemčík, Pavel (advisor)
This master thesis is focused on parametrization of image point neighborhood. Some methods for point localization and point descriptors are described and summarized. Gabor filter is described in detail. The practical part of thesis is chiefly concerned with particle filter tracking system. The weight of each particle is determined by the Gabor filter.
|
|
Personal Navigation Based on Wireless Networks and Inertial Sensors
Kaňa, Zdeněk ; Raida, Zbyněk (referee) ; Soták,, Miloš (referee) ; Bradáč, Zdeněk (advisor)
Tato práce se zaměřuje na vývoj navigačního algoritmu pro systémy vhodné k lokalizaci osob v budovách a městských prostorech. Vzhledem k požadovaným nízkým nákladům na výsledný navigační systém byla uvažována integrace levných inerciálních senzorů a určování vzdálenosti na základě měření v bezdrátových sítích. Dále bylo předpokládáno, že bezdrátová síť bude určena k jiným účelům (např: měření a regulace), než lokalizace, proto bylo použito měření síly bezdrátového signálu. Kvůli snížení značné nepřesnosti této metody, byla navrhnuta technika mapování ztrát v bezdrátovém kanálu. Nejprve jsou shrnuty různé modely senzorů a prostředí a ty nejvhodnější jsou poté vybrány. Jejich efektivní a nové využití v navigační úloze a vhodná fůze všech dostupných informací jsou hlavní cíle této práce.
|
| |
| |
|
Application of Sequential Monte Carlo Estimation for Early Phase of Radiation Accident
Šmídl, Václav ; Hofman, Radek
The early phase of radiation accident is characterized by minimum number of measured data and high uncertainty in both atmospheric conditions and radiation situation. Our goal is to provide an accurate method of radiation situation assessment that is capable to respect the uncertainty and provide informative predictions of its evolution for the involved decision makers. We propose a state space model based on atmospheric dispersion model, numerical weather model with local corrections and random walk on the model corrections and release evolution. This model is highly nonlinear and is estimated using sequential Monte Carlo. Since the model is significantly more complex that previously considered models and its estimation with naive proposal densities become too computationally demanding. We propose to construct a proposal density using problem specific simplification followed by application of the Laplace approximation. Properties of the resulting estimation procedure are illustrated on a twin experiment.
|
|
Asimilace časoprostorového rozložení radionuklidů v časné fázi radiační nehody
Hofman, Radek ; Šmídl, Václav
Exploitation of the data assimilation methodology in the early phase of radiation accident is studied. When radioactive pollutants are released into the atmosphere, a radioactive plume is passing over the terrain. The released radioactive material causes pathway-specific irradiation which has detrimental effects on population health. In order to ensure efficiency of introduced countermeasures, it is necessary to predict spatial and temporal distribution of the aerial pollution and material already deposited on the ground. The predictions are made by the means of a numerical dispersion model with many inputs. Output of such a model is a prediction of radiation situation given in terms of radiological quantities. Exact values of the inputs are uncertain due to the stochastic nature of the dispersion, lack of accurate information, etc. Their subjective choice can introduce significant errors into the predictions and thus decrease the positive impact of the countermeasures.
|
| |
|
Odhad neznámé variance nelineárního stavového modelu pomocí marginalizovaného particle filteru
Šmídl, Václav
The problem of estimation of unknown covariance matrix of non-linear state-space model is studied. The proposed methodology is based on combination of Extended Kalman Filter with particle filter. It is shown that the approach is promising for limited number of unknown parameters. More demanding problems with completely unknown covariance structures can not be reliably estimated since the observed data do not carry enough information.
|