Ing. Lukáš Brchl

Ing. Oldřich Malec

Department of Software Engineering

This bachelor thesis focuses on design and implementation of a fleet management module to existing application Dronetag using modern frontend web technologies - mainly TypeScript and React. The module allows users to form organizations, manage the organization, and plan missions. The thesis is divided into smaller specific goals. First, we conduct research in which existing solutions on the market are analyzed and their advantages and disadvantages are listed. The following chapter Module Analysis puts focus on software requirements and use cases and defines what the module should do. Module Design chapter places a lot of emphasis on graphic and software design, describes the architecture, and provides a high-level overview of the designed solution. The implementation part of the thesis explains specific implementation decisions, introduces the libraries used in the project, and elaborates on more advanced components. The last chapter describes the testing process and provides a report on actual testing with real pilots and users. The last section of the chapter suggests future improvements to the module.

Thesis on DSpace

Ing. Lukáš Brchl

prof. Dr. Ing. Petr Kroha, CSc.

Department of Software Engineering

This bachelor's thesis investigates the usage of object detection algorithms on images captured by an infrared camera placed on a drone. The solution will help to automate the analysis of captured data, targeting to increase the effectiveness of rescue operations. During the completion of the task, I developed a Python desktop application, that realizes chosen detection methods. The methods selection was based on an analysis of current approaches and take advantage of the existing detection systems. The application was used to measure the accuracy and performance of these approaches on the dataset created as a part of the thesis. In the end, the conclusion evaluates the possibility to use image detection on a thermogram, in a real-world application. The single-stage Region Proposal Convolutional Network showed the best result and was chosen for future development.

Thesis on DSpace

Ing. Jan Matoušek

Department of Software Engineering

This bachelor's thesis focuses on analyzing, designing, implementing and testing drone fleet management solution. Fleet management solution should allow grouping drones and pilots into organizations and coordination of flights with multiple pilots involved. Existing fleet management solutions are selected and their analysis is then performed, assessing their strengths and weaknesses. Current Dronetag platform inner workings are analyzed for proper understanding of system constraints put on the developed solution. Strengths and weaknesses of the analyzed platforms and knowledge of the Dronetag platform is then applied for fleet management solution analysis. Needed extensions to the Dronetag platform are then designed and implemented. Backend extension is implemented in Python using the Django framework, mobile application extension then uses the Flutter framework. Finally, the working prototype is tested by real pilots using defined test scenarios. Output from execution of these scenarios is then used to propose further user experience improvements and behavior changes.

Thesis on DSpace

Ing. Martin Daňhel, Ph.D.

Department of Digital Design

This thesis deals with implementation of embedded application for reception of identification and localization messages from drones. The thesis summarizes European standard for direct remote identification and descripes implementation of application for reception of these messages. The application is implemented on the ESP32 platform, specifically on the ESP32-C3 SoC. Solution is written in C and is written as a module for real-time operating system Zephyr. The module is able to function independently within larger projects. Messages are received using Bluetooth 4, Bluetooth 5 Long Range, Wi-Fi Beacon a Wi-Fi Aware technologies. Receiving of both Wi-Fi and Bluetooth was done both on single ESP32-C3 SoC and using separate Bluetooth Stack architecture. One ESP32-C3 was running application and reception of Wi-Fi messages and other was used for reception of Bluetooth Message. Both devices are communicating between themselves using HCI UART transport protocol. During the testing, it was found that when receiving Wi-Fi and Bluetooth at the same time on single SoC, there were lost messages. In the case of a separate Bluetooth Stack architecture, this problem has been solved.

Thesis on DSpace

prof. RNDr. Pavel Surynek, Ph.D.

Department of Applied Mathematics

Quality control is an integral part of SMT electronics manufacturing. Automated inspection systems are available but costly. In this thesis, we propose a hardware and software solution that is less expensive and easier to implement. Our image capturing system is composed of 3D-printed parts and additional electronic parts for illumination. The proposed defect detection method consists of a simple thresholding-based localization method and a random forest classifier, which uses features extracted with a pre-trained VGG-16 neural network. Our model reaches an 80~\% accuracy on our dataset, showing the success of our solution but also exposing shortcomings of our dataset and method. The thesis concludes with an analysis of results and suggestions for future improvements.

Thesis on DSpace

doc. Ing. Pavel Hrabák, Ph.D.

Department of Applied Mathematics

This thesis aims to design a framework for tracking people based on a stream from a single stationary camera, with the secondary goal of extracting age and gender information for tracked people. The focus of this work is on the retail shop environment. The main algorithm follows the tracking by detection approach. The matching of detections to tracks is done based on spatial and visual information from convolutional neural networks. Kalman filter is used for robust state representation and updates. We evaluate the algorithm with multiple detector models on a dataset collected from the target environment. We also evaluate the performance improvements from using the TensorRT optimization framework. The resulting application achieves 0.91 MOTA on the testing dataset, with frame rate of 13 FPS on the Jetson NX platform.

Thesis on DSpace

doc. Ing. Kamil Dedecius, Ph.D.

Department of Applied Mathematics

This thesis aims to design an algorithm for the detection of faults on UAVs using vibration analysis. Defects in the propulsion system of an UAV, such as damaged propellers or motors, can cause loss of thrust, higher power consumption, and in the worst case, lead to a crash of the aircraft. Which may cause severe damage to the aircraft itself, and pose a safety hazard to nearby people. Vibrations are measured using the 3 axis MEMS capacitive onboard accelerom- eter of a Blip development board. A custom Zephyr RTOS based application is designed and implemented for the data collection purposes. Three different datasets are collected using mu- tiple different devices, as vibration sources. Two different machine learning based models are tested and evaluated on the collected datasets. Both models reach above 95% test accuracy when trained using samples of all fault classes.

Thesis on DSpace

doc. Ing. Jakub Kraus, Ph.D.

Department of Software Engineering

Following the new pan-European legislation, manned aircraft will be required to broadcast their position in U-space. For which, however, many aircraft still need to be equipped with the necessary technology. In this thesis the chassis of a new battery-powered tool is designed. It will serve as a portable identification device with the extra ability to record flight and alert on collision traffic. In the study, existing products are researched, and interviews with pilots are carried out to define the required parameters. The outcome is the first 3D-printed prototype of the device enclosure, which will serve testing purposes and further development into a commercial product. Furthemore, the user interface of a companion mobile application is designed. The concept of the interface includes product configuration, a surrounding air traffic map display and a digital pilot logbook. This thesis is part of a project led by Dronetag, which is developing the device.

Thesis on DSpace

Ing. Jakub Novák

Department of Applied Mathematics

Computer vision makes an important part in Surface Mount Technology processes including Pick and Place (P&P) operations. For P&P operation it is crucial to introduce strong computer vision solution to detect and align components to correct position before they are placed on the board. In the field of open-source software for DYI P&P machines the most popular tool is OpenPNP. Its computer vision takes form of a pipeline of OpenCV operations applied on the captured image of the component. This solution is not reliable and behind the latest trends of object detection. This work aims to improve the part detection process in OpenPNP tool by implementing better pipelines management system and introducing an alternative computer vision method that would be more robust to outer conditions. The prior research focused on two main aspects; user experience with OpenPNP object offset detection, and current trends in object detection that could be used for P&P detection purposes. Based on the OpenPNP community experience and discussions I designed and implemented a new solution that will allow users to manage OpenCV pipelines by making them reusable. In other part of the work, I trained four detectors for rotated objects and measure their performance upon custom dataset captured by P&P machine. The best detection results were achieved using R3Det detector. I compared the detector results with current OpenCV solution and proved that solution using machine learning can be more robust alternative able to cope with real-time scenarios without the need of regular user input.

Thesis on DSpace

Department of Software Engineering

The thesis aims to design and implement a mobile application that will de- tect wireless broadcasts from unmanned aerial vehicles. According to new regulations, pilots need to register, and their drones are required to transmit the position and identification in a way that data can be acquired by mobile phones. The solution will gather data from the surroundings, giving users complete information about aerial traffic in their vicinity. The application will help to make drone operations safer because the general public may use it to identify operators and make them accountable for their actions in case the security or privacy is threatened.

Thesis on DSpace

Ing. Magda Friedjungová, Ph.D.

Department of Applied Mathematics

The goal of this thesis is the creation of a system, which is able to detect and track persons using information from a stationary camera. This system is also able to extract biometric information of age and gender from the detections. This can be useful for example in a commercial setting, where a retail store can use this information to predict customer behavior and/or plan marketing strategies.

Thesis on DSpace

Ing. Lukáš Brchl

doc. RNDr. Pavel Surynek, Ph.D.

Department of Applied Mathematics

In recent years, the applications of vehicle detection and tracking have been increasingly used for monitoring traffic situations in populated areas and mo- torways, planning traffic on the roads and for various civil uses. Combining recordings from surveillance cameras and modern algorithms for detection and tracking objects, we can aggregate statistics on the number of vehicles in each area, average speed, type, colour, brand, trajectories and more. This work deals with the issue of detection and monitoring of vehicles in normal operation from the video recordings. The first part focuses mainly on the analysis of algorithms, structures, models and existing solutions to the given topic; describes different types of neural networks used in advanced models. Based on the analysis, in the second part of the work it implements a solution for vehicle detection and tracking using the DETR model, which is an absolute novelty among models for object detection and classification, which uses the Transformer deep learning model. In the implementation for training and evaluation, the VisDrone dataset was used.

Thesis on DSpace

Ing. Michal Valenta, Ph.D.

Department of Software Engineering

This thesis describes analysis, design and implementation of application for Apple Watch smart watches. This application will help drone pilots with better orientation in airspace. The thesis also includes research of other similar applications for WatchOS operating system, requirements and use cases specification and design proposal. Design proposal was tested with several testers, who helped with final appearance of application. Realization part includes architectural design proposal for WatchOS, communication with associated mobile application, describtion of web service and implementation details. The whole solution is tested with several testers and also with people who work in flight industry.

Thesis on DSpace

Ing. Lukáš Brchl

Ing. Jakub Žitný

Department of Software Engineering

This master's thesis focuses on cross-platform mobile application development in the Flutter framework for safer drone operations. This thesis contains a full application analysis, design, and implementation. Its main functionality is the managing drones and identification devices attached to drones, showing restricted flight zones, and planning and watching drone flights. The analysis emphasizes the analysis of the existing web platform and all infrastructure, which provides data to the client and third-party appli- cations. The design contains a description of the application structure and uses architecture patterns of the Flutter framework. The implementation de- scribes details about realization and contains detailed launch instructions in the development, test, and production environments. In addition, it contains a description of the necessary configuration files. These files differ across the various environments and have a security disposition.

Thesis on DSpace

Ing. Lukáš Brchl

Ing. Pavel Hrabák, Ph.D.

Department of Applied Mathematics

This thesis deals with the implementation of an application for vehicle recognition from images captured with flying unmanned aerial vehicle (UAV). In the first section of the thesis is an overview of already existing solutions from this topic. This section is focused mainly on algorithms used in this thesis. The next section is an overview of already implemented solutions for automatic car detection on parking lots. Besides, the thesis proposes and implements an algorithm for the automatic detection of parked vehicles. The result of the thesis is four models and approaches implemented for automatic car detection from UAV on parking lots. For training was used CARPK dataset which contains more than 90 000 captured cars on more than 1500 photos. Furthermore, the work proposes an approach to detecting objects in a video, by creating an orthophoto map, which is subsequently provided detection. The whole thesis is implemented in the Python language.

Thesis on DSpace

doc. Ing. Štěpán Starosta, Ph.D.

Department of Applied Mathematics

The electric utility companies spend large amounts of money and effort every year to ensure the safe and uninterrupted operation of the electric power infrastructure. The most common source of outages is vegetation damaging power lines, for example, fallen trees. For this reason, companies perform regular inspections and maintenance of power line corridors, especially in forests and densely vegetated areas, creating a high demand for inexpensive and highly automated methods of power line corridor surveys. This work aims to create a robust algorithm for automatic detection of vegetation encroachment in the power line corridor using an Unmanned Aerial Vehicle (UAV), the techniques of photogrammetry, and computer vision. The study will cover the workflow for power line corridor inspection from comprehensive guidelines for data acquisition through power line 3D reconstruction to vegetation encroachment detection and visualization of the results.

Thesis on DSpace

Ing. Jakub Žitný

Department of Applied Mathematics

This diploma thesis focuses on facial landmark detection with the aim to create a system for automated speech therapy. Important part of this system is the ability to detect and track tongue movement. The Intel RealSense D415 depth camera is used for this purpose. The resulting application written in the Python language is able to evaluate several typical speech therapy exercises either from live camera footage or prerecorded videos in the .bag format. An experimental analysis of two selected exercises was performed using video data of several volunteers including both adults and children. On average the system was able to recognize 65% of the performed exercise repetitions without any of the exercises getting evaluated prematurely.

Thesis on DSpace

doc. Ing. Pavel Kordík, Ph.D.

Department of Applied Mathematics

The automotive industry, a vital and influential sector in the global economy, relies heavily on cutting-edge technological advancements and innovative methodologies to drive its growth and development. This study delves into the potential of advanced image processing techniques, encompassing illumination homogenization, image segmentation, and subsequent vehicle modeling using NeRF technology. The primary objective of this research is to develop a proof-of-concept solution that harnesses state-of-the-art technologies for vehicle processing and modeling, demonstrating their efficacy and superiority over traditional approaches. The developed solution underscores the benefits of utilizing novel technologies, showcasing their potential to transform the automotive industry by offering enhanced accuracy, efficiency, and overall performance compared to conventional methods.

Thesis on DSpace