National Repository of Grey Literature 8 records found  Search took 0.00 seconds. 
Analysis of robot position errors
Němec, Jakub ; Krejsa, Jiří (referee) ; Šolc, František (advisor)
This bachelor’s thesis deals with the comparison of several robot position error computation methods. The first part describes individual computation methods. The next parts are dedicated to the individual described methods and factors that influence the formation and size of position errors.
Image Descriptors and their Usage for Object Detection
Bature, Jonathan ; Druckmüllerová, Hana (referee) ; Procházková, Jana (advisor)
Detekce obličeje v obrazech je široce prozkoumaným tématem v počítačovém vidění. Algoritmus která umožnila rozpoznání obličeje a stanovila nové standardy v této oblasti byla Viola-Jones algoritmus. Tato práce popisuje a vysvětluje vlastní implementaci obličeje detektor založený na algoritmu Viola Jones pomocí Matlab CascadeObjectDetector. Tato diplomová práce přispívá ke zkoumání technik, jako je hlavní součást analýza (PCA) pro rozměrovou redukci deskriptorů pro stanovení detekce objektu systém, který dosahuje nejlepšího kompromisu mezi výkonem a rychlostí. V našem přístupu, detekce obličeje se provádí analýzou hlavních součástí (PCA). Obraz obličeje je promítnutý do prostoru obličeje, který kóduje nejlepší variace známých obrazů obličeje. The prostor obličeje je definován vlastní tváří. Vlastní tvář je sada obličejových vlastních vektorů které nemusí odpovídat běžným rysům obličeje, jako jsou oči, nos a rty. Systém funguje tak, že promítá předem extrahované obličejové obrazy do řady obličejových prostorů které představují velké odchylky mezi známými obrazy obličeje. Obličeje jsou klasifikovány jako známé nebo neznámé tváře po porovnání s existujícím obrázkem tváře v databázi.
Odometer module for mobile robot
Davídek, Daniel ; Burian, František (referee) ; Jílek, Tomáš (advisor)
The main objective of this work was the development and realization of odometric module for mobile robot envMap. The module processes signals from the incremental encoder and then calculates the position and direction of the robot together with the probability of its occurrence. The calculated data with a time stamp obtained from the GPS, are then sent to the above-system via the serial interface. For the realization the development board STM32F407G was chosen. f this board can be configured to function as an encoder input. This minimizes the possibility of encoder data loss when external interrupts are handled. The STM32F4 will not be any pull-back in the later development as there are plenty of peripherals on STM32F4. They were designed with two printed circuit boards: the reduction for the GPS module and shield-board for STM32F4. The equations for calculating odometric information and the probability of occurrence are included in this work. To calculate the movement of the robot, the equation considering only the movement on the plane surface is chosen. Later in this thesis there is also description of the source code for the STM32 and its functionality. In the results there is a complete connecting diagram along with the description of the utility program ODOMAT and communication messages. Finally, measurements correcting systematic error in parameters of the robot are described and there is the final odometry measurements compared with GPS localization as ground truth.
Image Descriptors and their Usage for Object Detection
Bature, Jonathan ; Druckmüllerová, Hana (referee) ; Procházková, Jana (advisor)
Detekce obličeje v obrazech je široce prozkoumaným tématem v počítačovém vidění. Algoritmus která umožnila rozpoznání obličeje a stanovila nové standardy v této oblasti byla Viola-Jones algoritmus. Tato práce popisuje a vysvětluje vlastní implementaci obličeje detektor založený na algoritmu Viola Jones pomocí Matlab CascadeObjectDetector. Tato diplomová práce přispívá ke zkoumání technik, jako je hlavní součást analýza (PCA) pro rozměrovou redukci deskriptorů pro stanovení detekce objektu systém, který dosahuje nejlepšího kompromisu mezi výkonem a rychlostí. V našem přístupu, detekce obličeje se provádí analýzou hlavních součástí (PCA). Obraz obličeje je promítnutý do prostoru obličeje, který kóduje nejlepší variace známých obrazů obličeje. The prostor obličeje je definován vlastní tváří. Vlastní tvář je sada obličejových vlastních vektorů které nemusí odpovídat běžným rysům obličeje, jako jsou oči, nos a rty. Systém funguje tak, že promítá předem extrahované obličejové obrazy do řady obličejových prostorů které představují velké odchylky mezi známými obrazy obličeje. Obličeje jsou klasifikovány jako známé nebo neznámé tváře po porovnání s existujícím obrázkem tváře v databázi.
Portfolio optimization for an P2P investor on Zonky
Jonáš, Filip ; Polák, Petr (advisor) ; Máková, Barbora (referee)
This thesis analyzes the Czech peer-to-peer lending platform Zonky. The goal was to find the optimal portfolio for a risk-averse investor investing in Zonky loans. For this purpose, the Modern portfolio theory from Markowitz was used. Based on the provided loan book containing information about loans which Zonky has provided since its foundation we examined the statistical properties of the individual risk categories represented by the interest rate charged. The optimization was done using the Excel Solver tool assuming that the loan categories are uncorrelated as well as considering the correlation we found using the variance- covariance matrix. For both cases, the portfolio minimizing the standard deviation as well as the portfolio which maximizes the Sharpe ratio was found. Generally, both types of portfolios were comprised mainly of loans with lower interest rate. According to our results, it seems that such loans offer better relationship between risk and return compared to categories which are riskier. Also, we showed that the platform's recovery rate has a significant impact on the performance of the loan categories especially of those which are among the riskiest. Furthermore, we demonstrated that the correlation between individual risk categories should not be ignored when a portfolio...
Odometer module for mobile robot
Davídek, Daniel ; Burian, František (referee) ; Jílek, Tomáš (advisor)
The main objective of this work was the development and realization of odometric module for mobile robot envMap. The module processes signals from the incremental encoder and then calculates the position and direction of the robot together with the probability of its occurrence. The calculated data with a time stamp obtained from the GPS, are then sent to the above-system via the serial interface. For the realization the development board STM32F407G was chosen. f this board can be configured to function as an encoder input. This minimizes the possibility of encoder data loss when external interrupts are handled. The STM32F4 will not be any pull-back in the later development as there are plenty of peripherals on STM32F4. They were designed with two printed circuit boards: the reduction for the GPS module and shield-board for STM32F4. The equations for calculating odometric information and the probability of occurrence are included in this work. To calculate the movement of the robot, the equation considering only the movement on the plane surface is chosen. Later in this thesis there is also description of the source code for the STM32 and its functionality. In the results there is a complete connecting diagram along with the description of the utility program ODOMAT and communication messages. Finally, measurements correcting systematic error in parameters of the robot are described and there is the final odometry measurements compared with GPS localization as ground truth.
Analysis of robot position errors
Němec, Jakub ; Krejsa, Jiří (referee) ; Šolc, František (advisor)
This bachelor’s thesis deals with the comparison of several robot position error computation methods. The first part describes individual computation methods. The next parts are dedicated to the individual described methods and factors that influence the formation and size of position errors.

Interested in being notified about new results for this query?
Subscribe to the RSS feed.