2018-2019

Advisor: Prof. Dr. Engin Türe

Team: Kerem Özdemir, Etka Eren Altıntaş, Mustafa Melih Metin

Abstract: While drone technology is one of today's indispensable, and it is one of the building blocks of the future. Already from agriculture to safety, transportation to industry, they are in our service in different areas of life. However, the biggest problem of these devices is flight time. In this study, flight time of drones is aimed to be increased without changing basic characteristics such as battery, weight, size. The main reason for this aim is that the drone's functionality is not diminished. Such that; in other studies, solar panels were mounted on drones. However, this situation has led to an increase in the weight and size of drones and reduced functionality. In this study, solar energy was stored and drone was recharged using an auxiliary vehicle. The wireless charging module is used to charge the drone. In this way, the drone continued its normal operation and its structure was intact. Arduino software and module are used to help the drone and vehicle to find each other. The vehicle is designed to be autonomous so that the vehicle can carry the battery without any problem or crash. In addition, it has continued to work without any interference, even in unpredictable natural conditions. A drone in the Quark Micro brand and model has been charged in 15 minutes.

Keywords: Drone, Arduino, Solar Power, Wireless Charge, Autonomous Vehicle, Solar Panel

Report I, Report II

Advisors: Prof. Dr. Engin Türe, Dr. Tuba Ayhan

Team: Mustafa Can Gülbaş, Diren Erdem Aydın

Abstract: Nowadays, accessibility to water and power are crucial for humanity especially for people who earn their daily bread from depended industries to water and power, like farmers. Under current conditions, requirements of agriculture sector are not ensured in accordance with today’s technology. Since the global demand from agriculture is increasing with population growth, traditional irrigation techniques are not able to satisfy the global demand even if these old techniques satisfy the global demand, there are serious concerns about water and energy wastage. With the combination of automation technology, new generation irrigation techniques such as drip irrigation or sprinkler irrigation, and renewable energy sources, is the scope to reduce issues regarding water and power crisis. In this paper, a micro-controller based solar powered smart irrigation system and control features, is designed and prototype is implemented. Sensors enabled trough the controller, the prototype is fed by ordinary battery, system is connected to solar photovoltaic and able to communicate with several devices about system’s status.

Keywords: Smart Irrigation, Solar Power, Sensors, Water&Energy Saving, Microcontroller, Communication

Report I, Report II

Advisor: Dr. Tuba Ayhan

Team: Göktuğ Kayacan, Remzi Orak

Abstract: With the most recent improvements on the machine learning technologies more advanced algorithms such as neural networks came under a spotlight once again. Although most neural network-based machine learning methods are considered heavily resource dependent, it is possible to implement lighter but more restricted implementation of a neural network. Implementing such an algorithm on a platform which operates purely on hardware basis is a good way to demonstrate how powerful such an algorithm could be. Our goal is to implement an area and resource efficient neural network on a FPGA and use this neural network to apply pattern recognition based on the MNIST dataset. The final system is capable of recognizing given digits with high precision.

Keywords: FPGA, Neural Network, Machine Learning, Touchpad, MNIST, Pattern recognition, Verilog, Python, MATLAB, Optimization, Hardware, ASIC.

Report I, Report II

Advisors: Dr. Ebru Arısoy, Dr. Tuba Ayhan

Team: Mehmet Taylan Eğer, Sena Akbulut, Sena Koyuncu

Abstract: The proposed project offers a low-cost alternative to smart-meters focusing on giving real-time information about the consumption. The project was developed around the idea of raising awareness on over-consumption. With only a camera and a Raspberry Pi, an adapted system was created to traditional gas, electricity and water meters. So, it is not expected to change the system of the house. The consumption data is being extracted from the images using OCR algorithm after some image processing is being done such as filtering and editing. The data is being recorded in the file which is named corresponding date. This procedure is being executed on hourly basis. At the end of the day, the file is being imported to the user via email. The user is being expected to download the file to the path provided on email. The system offers an interface where the daily, monthly and annual graphs of usage can be seen. Using this interface, the consumption data in the txt files are being graphed. Providing the consumption data to the users is the key of the system. By the continuous use of the accessible consumption graphics, more accurate billings and more control over the consumption by users are being predicted.

Keywords: Raspberry Pi, OCR, GUI, Image Processing

Report I, Report II

Advisors: Dr. Ebru Arısoy, Dr. Tuba Ayhan

Team: Atakan Kara, Mustafa Talha Tunçak, Arda Kalo, Caner Coşkun

Abstract: This article includes Signal Detection and Localization using prototype which is car. Signal detection device was thought to find electronical devices that include wireless signal communication. Electronical device communication is provided by signals. Signals have transmitting and receiving power. According to Friis’s Equation, there is an exponential relation between distance and power. The relation is used to determine location information of receivers. Almost all electrical devices care telecommunication signal therefore this study focused on detection their received signal to have information about receiver’s location. Location detection is done by mapping.

Keywords: Signal detection, GSM (Global System for Mobile Communication), Radio frequency.

Report I, Report II

Advisors: Dr. Ebru Arısoy, Dr. Serap Kırbız

Team: Uygar Akkoç, Mustafa Utku Öztürk, Bengisu Özmelleş

Abstract: This report describes a detection and recognition system on human faces which processes rapidly in real time. Human face contains the specific features required for use in the recognition and detection systems. Goal of the face detection part is to determine the size and location of a face in the given video. For the implementation of that part, object detection method based on Haar Cascades proposed by Paul Viola and Michael Jones used. As a result of this process, the data is used for recognition algorithm’s input. A face recognition system based on the eigenfaces approach is proposed. This approach is done by comparing the characteristic features of human faces. Eigenface method is divided into two different sections: training and prediction. In the training, the set of images with human faces are scanned and identified. As a result of training, eigenfaces are created. In the prediction section, the similarity percentage is measured by comparing the dataset of human faces and the input image. If the system gives a percentage above a predetermined level for the face it sees, it indicates that it recognizes the person. If it does not recognize the face that it sees, system will offer option to add this contact to dataset with the GUI (Graphical User Interface) that we created. With GUI, user enters the mail address for to receive a notification about the guests.

Keywords: Face Detection, Face Recognition, Eigenface, Haar Cascades

Report I, Report II