2024-2025

Advisor: Asst. Prof. Egemen Bilgin

Team: AYÇA TOGAR, GÖKÇENAZ GÜLER, PELİNSU AYDIN

Keywords: microwave imaging, food inspection, vivaldi antenna, inverse scattering, RTM, MUSIC

Report

Advisor: Asst. Prof. Egemen Bilgin

Team: ELİF KOÇAK, BESTE YAĞMUR AVANOĞLU

Keywords: Ground Penetrating Radar, Horn antenna, signal processing, excitation signal, IFFT, Hermittian process, ground removal techniques

Report

Advisor: Asst. Prof. Yusuf Aydın

Team: YİĞİT ŞAN, UTKU KEREM TAŞDELEN

Keywords: Fencing, Wireless, Scoreboard, Epee, ESP-NOW

Report

Advisor: Asst. Prof. Yusuf Aydın

Team: Onur Kurt, Ömer Mert Yıldız

Keywords: IoT, smart home systems, water leakage detection, energy monitoring, Arduino Cloud, ESP32, ESP8266, cloud-based monitoring

Report

Abstract: This project presents the design and implementation of an IoT-based smart monitoring system that addresses two major concerns in residential environments: early-stage water leakage detection and real-time energy parameter monitoring. The system architecture employs a distributed sensor network comprising ESP8266 modules for local water detection, an ESP32 board for real-time energy measurement, and an Arduino Uno R4 WiFi module serving as the centralized cloud-connected controller. Sensors including non-contact water level modules, ACS712 current sensors, and ZMPT101B voltage sensors were integrated to collect and transmit data. All communication between nodes and the central hub was established wirelessly using Wi-Fi protocols and synchronized with the Arduino IoT Cloud to facilitate remote access and control. The user interface was built using Arduino Cloud dashboards, allowing end users to monitor current sensor readings, receive alerts, and operate actuators via mobile or desktop platforms. The system demonstrated reliable performance under varied test conditions, validating its ability to respond to abnormal water or power events in real time. The overall outcome emphasizes the practicality of scalable, low-cost, and open-source IoT frameworks for increasing household safety, energy transparency, and remote accessibility in smart home applications.

Advisor: Asst. Prof. Yusuf Aydın

Team: Ali Emir Özen, Mahmut Küçükakyüz  

Keywords: IQuadruped robot, Automation, Computer vision, Courier, Robot arm.  

Report

Abstract: Although most of the documents exchanged between departments in institutions are transmitted and recorded digitally, there are many documents requiring original signature. Delivering these documents from one department to another is a task in itself. Additionally, such document delivery is often carried out by the secretary or the person who prepared the document. However, during this delivery process, the primary tasks of those delivering the documents may be interrupted, and their focus may be disrupted. In some companies, document tracking personnel are responsible for recording all incoming documents from companies and clients into a ledger or digital system and directing them to the relevant unit within the company. These individuals must have excellent time management skills. Documents need to be delivered to the appropriate unit, archived accurately when sent for storage, and tracked when necessary. Focus is the key element to ensure these tasks are performed flawlessly. However, as these individuals often need to leave their work areas due to the nature of their duties, they may experience physical and mental fatigue. At times, scenarios may arise where the person delivering the document has to wait because the intended recipient is not present or has to return later to deliver the document once the recipient becomes available. Such scenarios create distractions and disrupt focus. These delivery issues are not exclusive to companies; they also occur in sectors like factories and hospitals. In hospital environments, for instance, the transportation of materials such as documents, samples, and blood specimens are often carried out by doctors, nurses, or secretaries. This reduces the efficient utilization of healthcare professionals in their areas of expertise and leads to unnecessary fatigue.

Our project aims to develop an autonomous transportation solution to increase efficiency and improve workflows in institutional transportation within factories, schools, and hospitals. To achieve this goal, our project involves designing and developing an autonomous robot capable of securely and quickly transporting various lightweight items, such as documents, cargo, and equipment, from one point to another. By automating document delivery, which has been previously described as reducing efficiency and focus, our project seeks to enhance employee performance and provide a solution to this issue.

Advisor: Asst. Prof. Tuba Ayhan

Team: Muhammed Emin Bayraktar, Abdullah Erarslan, Tayyip Uzun, Buğra Akif Özesen

Keywords: Weight measurement, data retention, data transfer, automatic tare, tea picking machine

Report

Abstract: According to data published by the Ministry of Agriculture and Forestry in 2023, 1.27 million tons of fresh tea were harvested annually on 792,000 decares of land in the provinces of Rize, Trabzon, Artvin, and Giresun in the Eastern Black Sea Region in 2021 [1]. Over the years, to facilitate tea harvesting, various tea plucking machines powered by different fuels, sizes, and weights have been developed, replacing manually operated tea shears. Conversations with individuals involved in tea harvesting revealed that some workers in this region are paid based on the weight of the tea they collect. Employers mentioned that when workers paid by the weight of the tea work simultaneously in the same garden, they must exert additional effort and time to prevent the harvested tea from being mixed with that of other workers or the tea garden owner, which sometimes leads to disputes.

Over the years, in order to facilitate the tea picking work, various electric or petrol tea picking machines with different fuels, sizes and weights have been produced instead of fully manpowered tea shears. Workers now collect tea with these electric tea picking machines instead of tea shears. However, the workers cannot measure and record the weight of the tea collected while picking the tea. The planned electric tea picking machines with a measurement and record keeping unit are promising to prevent this waste of time and controversy. The targeted measuring and record keeping unit will automatically tare after each unloading, this automatic taring process avoids the problem that the tea remaining in the machine at each unloading process is recorded as extra in the next unloading process. During the daily operation, data can be kept in the machine memory together with the date and time information, and this way of keeping data prevents the discussions that may arise between the employer and the employee and provides concrete data information. This data stored in the memory will be transferred wirelessly to the employer's device when the employer wants to retrieve the data. This data transfer will allow employers to monitor performance objectively and provide a more organised working environment by preventing the mixing of the tea collected by the workers. 

A microcontroller based system will be developed to achieve these planned targets. This system consists of IMU, load cell, wireless communication unit and LCD display. During the project, each of these parts will be connected to the microcontroller and configured to fulfil its task. The method to be followed in the project is to make each unit (weighing, record keeping, tare, wireless transmission) work separately, integrate and test as a whole. In project management, the team works together and meets with academic and industrial advisors at regular intervals, and at the end of each work package, reports and discusses the work package outputs with the advisors. 

There is no similar example in the literature/market that can measure the weight of tea collected by integrating into a tea collector. The unique aspects of the project are that the target device can be integrated into existing tea collectors, update the tare as the hopper is emptied, record the weighing records and transfer them wirelessly. The measurement and record keeping unit will make the use of existing tea picking machines more attractive and reduce disputes during tea picking.

Considering the researches and objectives, it is foreseen that our project will turn into a marketable product. If a marketable and patentable product emerges, an application will be made for a patent.

Advisor: Asst. Prof. Yusuf Aydin

Team: Eren Aktaş, Kerem Emre, Ece Kaya

Keywords: Ornithopter, biomimetic UAV, flapping wing, surveillance systems, prototyping, unmanned aerial vehicle

Report

Abstract: This project aims to design and prototype a biomimetic unmanned aerial vehicle (ornithopter) that mimics the flapping wing motion of birds for use in surveillance and reconnaissance missions. Traditional unmanned aerial vehicles (UAVs), such as rotor-based or fixed-wing drones, often face limitations due to high noise levels, ease of detection, and restricted maneuverability. In contrast, ornithopters offer several advantages, including quieter operation, visual similarity to birds, and improved stealth, making them highly suitable for applications such as border security, intelligence operations, and wildlife monitoring.

Within the scope of the project, both mechanical and electronic subsystems of the ornithopter were designed and integrated. Key components include the flapping wing mechanism, a tail control system operated by servo motors, manual control via an RC transmitter, and basic sensor integration to gather motion and positioning data. A functional prototype was developed that is capable of generating lift and directional control.

Testing revealed that the flapping system could produce a degree of lift, although performance was limited due to mechanical friction, insufficient motor power, and structural constraints. On the other hand, the tail control mechanism successfully performed as intended, enabling directional adjustments with precision. Further development, including the addition of autonomous control algorithms and more advanced sensors, is expected to enhance flight stability and overall performance.

This project contributes to the advancement of biomimetic UAV design in Turkey and offers a foundation for future research in the field. The developed prototype holds potential for use in defense, environmental monitoring, and post-disaster search-and-rescue operations, aligning with the growing need for low-detectability aerial surveillance systems.

Advisor: Asst. Prof. Yusuf Aydin

Team: Yağmur Akbaş Furkan Arslan, Aslıhan Emine Mete

Keywords: PV Panel, Robot, Cleaning, Maintenance, Image Processing

Report

Abstract: This paper presents the development of a robot system designed to clean solar panels and ensure their effective maintenance. The system addresses general environmental factors - such as dust and dirt - which reduces the efficiency of the panel. Major components of the project include thermal image processing, implementation of a centralized remote control interface, and a cleaning module design adapted to solar panel surfaces, especially for solar panel surfaces to detect discrepancy. The proposed solution reduces manual labor and promotes sustainable operation of solar power systems.

Advisor: Asst. Prof. Yusuf Aydin

Team: Aleks Demir,Ata Berk Özçelik,Gizem Güre

Keywords: Autonomous robot, Indoor delivery, ROS 2, A* algorithm, PD controller, RFID authentication, Obstacle avoidance, LiDAR, SLAM, Path planning, Secure access, Mobile robotics, Robotics, Wheeled robot

Report

Abstract: This project illustrates the design and construction of a campus autonomous delivery robot with the aim of reducing manual labor in procedural delivery work while exercising safe and contactless transfer of items. The robot incorporates LiDAR perception and operates on a SLAM algorithm to map and track its surroundings precisely. In order to generate paths, A* was used instead of Dijkstra based on comparative testing, on the grounds that it is faster under dynamic conditions. A PD controller specially programmed for the robot enables it to traverse the planned path smoothly with sensitive motion. In order to address challenges such as navigation in narrow corridors and crowded environments, dynamic obstacle avoidance is achieved through costmap-based inflation layers, which were tuned experimentally. Delivery safety is ensured by an RFID-based access system in relation to each destination, and a servo-lock ensures that only approved users are able to collect goods from the robot. The system was tested on an experimental indoor setup for various navigation scenarios. The final deployment exhibited regular path planning, safe maneuvering, and successful deliveries. The modules enable scalability in many operational environments, such as inter-department supply and equipment transportation. The project further reduces labor costs and enhances delivery efficiency indoors in relation to sustainability.

Advisor: Asst. Prof. Yusuf Aydin

Team: Batuhan Erden, Abdullah Taha Aygün

Keywords: Robotic Hand, Haptic Feedback, Wireless Communication

Report

Abstract: The limitations in the ability of current robotic systems to perform complex tasks cause people to continue working in dangerous environments, even in tasks that sometimes require simple object handling. This situation can potentially cause serious injury to people working in these hazardous environments or potentially result in serious damage to the environment. To overcome this problem, robots that can imitate human movement in various ways are being developed. In this senior design project, a similar robotic system was developed in the form of a hand. As a solution to the problem mentioned above, this paper presents the design, implementation, and performance of a robotic hand system that can move its fingers wirelessly with high accuracy controlled by a glove worn by the user, while at the same time receiving a form of force feedback. This allows for users to control the robot that handles objects with high precision in environments where it is dangerous for humans to work or in jobs that require high accuracy without a human presence is required.

The system consists of two main modules that communicate wirelessly over the Bluetooth Low Energy (BLE) protocol; a control glove worn by the operator, and a robotic hand. The Arduino Nano ESP32-based control glove measures the bending angle of each finger thanks to five flex sensors placed in special pockets sewn onto the glove. At the same time, based on the data received from the robotic hand, five servos placed on the user's arm provides real-time force feedback to the user by pulling strings connected to the fingertips. As for the robotic hand, it was aimed to successfully mimic the movements of the fingers of the user with low latency and high accuracy, and to provide a force feedback for the user with the FSR sensors placed to the fingertips of the robotic hand to control to pressure robotic hand applies to the objects which it grips and handles.

Advisor: Asst. Prof. Serap Kırbız

Team: Alp Eren Kartal, Atamert Kaya, Enes Kadri Aksaçlı, Hasan Cenk Yiğit

Keywords: Security System, Image Processing, Body (Object) Detection, Face Detection, Biometric Verification, Anti Spoofing.

Report

Abstract: Advanced Room Security System is a multi-layered security system developed for places requiring special security such as bank rooms and server rooms. The system consists of the integration of multi-layered image processing-based security systems. The project, which consists of three stages: body detection, face detection and face recognition, aims to find solutions to the weaknesses of traditional security systems and to bring a new perspective to security systems. In addition, it is aimed to increase the system security level with anti-spoofing work. This report includes the theoretical background, methodology, design, application, testing and results obtained in the completion of the project.

Advisor: Asst. Prof. Egemen Bilgin

Team: Ahmet Resul Metin , İrem Aydın

Keywords: Radar Cross Section (RCS) , Frequency Selective Surface (FSS), Patch Antenna , Radome

Report

Abstract: This project investigates the integration of Frequency Selective Surfaces (FSS) in order to reduce the Radar Cross Section (RCS) of antenna systems. In the first stage, only the antenna structure was considered and RCS reduction was achieved with an FSS design operating in a single frequency band. In the second stage, the radome structure, which protects the antenna from the external environment, was included in the study, aiming to reduce the radar visibility of the structure consisting of the antenna and radome together. The ring-shaped FSS elements integrated on the radome were designed to be reflective at the 10 GHz frequency and permeable at the 5 GHz frequency where the antenna operates. In this way, effective RCS reduction was achieved at a certain radar frequency without deteriorating the antenna performance. Simulation results show that the single-band FSS integrated radome design provides an RCS reduction of approximately −16 dB. In addition, a dual-band FSS structure that can provide RCS reduction in two different frequency bands was also tested with the analyses performed in the CST Studio environment. However, the main focus of the project was the comprehensive examination of the single-band design. This work provides significant contributions to the fields of defense, aviation and communication by providing a low-cost and passive stealth solution.

Advisor: Asst. Prof. Egemen Bilgin

Team: Ara Arda Köseoğlu,Kerem Erden,Seren Paşaoğlu,Eren Taşel

Keywords: Time Difference of Arrival (TDOA), Source Localization, Positioning, Acoustic Wave Propagation, Optimization Techniques, Levenberg-Marquardt Optimization, Particle Swarm Optimization.

Report

Abstract: The aim of this project is to approximate the location of an unknown sound source using the time difference of arrival (TDOA) method. Accurately locating sound sources plays an important role in many areas, from security systems to robotic navigation. The speed of sound propagation and different acoustic obstacles can make location estimation difficult. The TDOA method is a technique that stands out with its high accuracy in such systems. This method is based on the principle that sound from a single source reaches receivers at different times. This difference in arrival times is related to the different distances between the source and each receiver. In the case of using three or more microphones as receivers, the location of the source can be calculated with hyperbolic equations based on these arrival time differences. In the TDOA method, a hyperbola is created for each microphone pair, and the intersection point of these hyperbolic curves gives the location of the source. In the project, a 2-dimensional modeled microphone array will be used to perceive the sound produced by the source by at least 3 receivers and convert it into an electrical signal. Based on the phase differences between these signals, the time difference between them will be determined, and hyperbolic graphs showing the location will be created accordingly. The intersection points of these graphs will be analyzed to obtain a result about where the sound source is. The microphones will be placed in both rectangular and linear positions, and sound reception tests will be performed in two different directions. The noise and reflection effects that occur in the instantaneous sound reception of the system create a high margin of error in finding the source of the sound. This causes the sound source to be located in the wrong place. Optimization techniques found in the literature will be applied to the system to minimize these effects. Levenberg-Marquardt optimization and Partial Swarm Optimization techniques were used in the project. Using more than one optimization technique, which technique gives better results in which location will be compared.

Advisor: Asst. Prof. Serap Kırbız

Team: Gökdeniz Ersoy, Mehmet Alper Tatar, Eray Tonbul

Keywords: Fatigue detection, Sleep detection, Sleep, Image processing, EAR,  PERCLOS, MAR, Facial Landmark, CNN 

Report

Abstract: Within the scope of this project, it is aimed to prevent traffic accidents caused by these symptoms by detecting the signs of sleep and fatigue that individuals who actively drive vehicles may encounter while driving. Approximately 20% of traffic accidents worldwide occur due to sleep and fatigue.  The Driver Sleep Detection System developed in this direction is a system that monitors the driver's eyelid movements, head position and yawning behaviors in real time and warns the driver both in writing and audibly in case of possible fatigue. The system takes eyelid opening (EAR) and lip opening (MAR) measurements in order to determine personalized threshold values  before driving. Based on these threshold values, the system warns the driver if critical levels are exceeded in the analyzes made during driving. Eye closure status is analyzed with the EAR (Eye Aspect Ratio) and PERCLOS (Percent Eye Closure) methods, and both the open/closed status of the eyes and the duration of closure are calculated. The MAR (Mouth Aspect Ratio) method was used for yawn detection, and certain reference points on the face (Facial Landmarks) were used to evaluate the head position. In order to increase the accuracy of these analyses made with classical methods, deep learning-based Convolutional Neural Network (CNN) models were also integrated into the project. With two separate CNN models developed for eye and mouth conditions, the overall accuracy of the system reached 97.7%. This project provides an effective and applicable solution that contributes to traffic safety by detecting driver fatigue with high accuracy.

Advisor: Asst. Prof. Serap Kırbız

Team: Ece Çalışkan,   Mert Kaan Pullu 

Keywords: real-time detection, YOLOv5, helmet detection, PPE, computer vision, workplace safety, object detection, industrial security 

Report

Abstract: PerceptiGuard II is a real-time computer vision system developed to enhance safety and security in industrial environments. Building on the initial trespasser detection system, this upgraded version integrates personal protective equipment (PPE) detection—specifically helmets—using a YOLOv5-based model. The system using  a custom dataset configured for 4 helmet types and achieves consistent detection performance independent of  various lighting conditions. Has an avg. confidence of 80% or higher  ideal conditions and real-time process speed more than  15 frames per second, PerceptiGuard II demonstrates practical trustability for on-site deployment. The solution operates on cost-effective, easy to buy and set up hardware and open-source tools, making it both accessible and scalable. Potential future enhancements include detecting additional PPE types and improving low-light performance.