LEE- Bilgi ve Haberleşme Mühendisliği-Doktora
Bu koleksiyon için kalıcı URI
Gözat
Son Başvurular
1 - 5 / 7
-
ÖgePhysical layer design of energy efficient and secure FSO links for next generation communication systems(Graduate School, 2025-05-23)The ever-increasing demand for extremely high data rate networks has prompted the research community to explore new technologies capable of meeting this need. Free-space optical (FSO) communication systems are a promising technology for next-generation wireless networks due to their attractive features, such as ultra-high data rates, license-free spectrum, ease of deployment, and low cost. However, despite these appealing characteristics, the performance of FSO communication systems is hindered by several detrimental factors. Weather conditions, such as fog and rain, can significantly degrade FSO system performance. Atmospheric turbulence fading, caused by random variations in the atmosphere, also attenuates the power of the transmitted optical signal. Pointing errors, which arise from violations of the line-of-sight requirement between transmitting and receiving apertures, further degrade FSO performance. Another limiting factor is channel estimation errors, which stem from imperfect knowledge of the channel state information at the receiver. This thesis presents three distinct studies concerning the physical layer of FSO communication systems. In the first study, we investigate the performance of FSO communication systems over an imprecise Málaga turbulence-induced fading channel, which is a generalized form of both the Gamma-Gamma and K turbulence channels. The effect of pointing errors is also considered. In our analysis, we first derive the probability density function (PDF) and cumulative distribution function (CDF) of the channel's fading coefficient in the presence of channel estimation errors. We then obtain exact closed-form expressions for the average bit error rate (BER), outage probability (OP), and ergodic channel capacity to quantify the performance of the system under consideration. Additionally, to provide further insight into system performance in the high signal-to-noise ratio (SNR) regime, we derive asymptotic expressions for the BER and OP. Furthermore, the analytical results are successfully validated through Monte Carlo simulations. The results demonstrate the detrimental effects of the channel estimation errors on the performance of FSO communication systems. For instance, an estimation error of 5% results in approximately an 8 dB SNR loss at a BER of 5 × 10^−2. In the second study, spatial diversity techniques are proposed for FSO communication systems to combat the deteriorating effects, such as atmospheric turbulence and pointing errors. The performance of FSO communication systems with the Alamouti encoding scheme over the Málaga turbulence channel is investigated. We first derive the PDF of the end-to-end channel gain under atmospheric turbulence and pointing error conditions. Then, by capitalizing on this PDF, closed-form expressions for the average BER and OP of the proposed system are obtained. Additionally, to provide more insight, the asymptotic expressions for the average BER and OP are also derived. In the analysis, intensity modulation/direct detection and heterodyne detection techniques are considered, allowing the obtained results to cover both cases. Furthermore, the analytical results are successfully validated through Monte Carlo simulations. Our results highlight the performance gains that can be achieved when the Alamouti encoding scheme is employed in FSO communication systems. In the final study, we present a dual-hop decode-and-forward relaying-based FSO communication system. We consider utilizing simultaneous lightwave information and power transfer (SLIPT) with a time-splitting technique at the relay, where the direct current component of the received optical signal is harvested as transmit power for the relay. It is assumed that the FSO links experience Málaga turbulence channels with pointing errors. In order to evaluate the performance of the proposed communication system, closed-form expressions for OP, ergodic capacity, average BER, and throughput are derived. Additionally, to analyze the physical layer security of the proposed system, closed-form expressions for secrecy outage probability and strictly positive secrecy capacity are obtained. Finally, the accuracy of the derived analytical expressions is validated through Monte Carlo simulations. Results show that our proposed system model outperforms its non-SLIPT counterpart.
-
ÖgeDirect pore-based identification for fingerprint matching process(Graduate School, 2023-09-15)Fingerprint, is considered one of the most crucial scientific tools in solving criminal cases. This biometric feature is composed of unique and distinctive patterns found on the fingertips of each individual. With advancing technology and progress in forensic sciences, fingerprint analysis plays a vital role in forensic investigations and the analysis of evidence at crime scenes. The fingerprint patterns of each individual start to develop in early stagesof life and never change thereafter. This fact makes fingerprints an exceptional means of identification. In criminal cases, fingerprint analysis is used to decipher traces, evidence, and clues at crime scenes. These analyses not only provide insights into how a crime was committed but also assist in identifying the culprits or individuals involved. Computer-based fingerprint identification systems yield faster and more accurate results compared to traditional methods, making fingerprint comparisons in large databases easier. These systems establish connections between fingerprints found at crime scenes and potential suspects' fingerprints, ensuring accurate results in forensic investigations. Furthermore, fingerprint analysis is not limited to criminal identification. It also proves highly useful in situations like identifying missing persons or victims and locating individuals lost in disaster areas. Therefore, fingerprint analysis has become an indispensable tool for justice systems, security agencies, and related institutions. Fingerprints are complex and unique characteristics that can uniquely identify individuals. This identification process typically consists of three levels, involving the features of ridge patterns, minutiae points, and pores. In the first level, the focus is on the ridge patterns present on the fingertips of each individual. Ridge patterns are raised areas on the fingertip surface and possess distinct shapes and flow directions for each person. These flow directions and shapes play a significant role in distinguishing one person's fingerprint from another's. In the second level, minutiae points are examined. Minutiae points are locations where ridge patterns intersect or bifurcate. These points can take various forms such as ridge endings, bifurcations, or ridge dots. Minutiae points are also utilized to determine the unique features of a fingerprint. The distinctiveness between two fingerprints often arises from differences in the types and positions of these points. The third level involves considering pores. Pores are tiny holes on the fingertip surface and have unique distributions in each fingerprint. The location and number of these holes are also used for fingerprint identification. These three levels constitute the fundamental components of fingerprint identification. This detailed examination and analysis demonstrate the high uniqueness of fingerprints and their effectiveness as a reliable biometric method for identity verification. However, there are some limitations and challenges faced by fingerprint identification technology. One primary issue is the potential inability of imaging systems to capture fingerprints with sufficient clarity and detail. Such a process can complicate data collection for imaging systems and consequently hinder providing a complete and accurate representation of fingerprints. Secondly, the insufficient level of porosity details for the third-level identification limits the widespread use of this form of identification. These factors highlight the challenges faced by fingerprint recognition technology. It's essential to enhance the sensitivity of imaging systems to obtain clearer and more detailed images even under varying conditions. Additionally, developing algorithms and data analysis methods to capture and process pore details more reliably is crucial. These solution-oriented approaches can unleash the potential of third-level identification, improving the reliability and widespread use of fingerprint recognition technology. Preserving pore-based fingerprint records in the unsolved cases database without subjecting them to any objective evaluation criteria underscores the importance of potential research in fingerprint analysis and identification. In this context, the presence of missing information in the said database emphasizes the inevitability of in-depth research in this field. Fingerprints captured in image records based on pores alone contain critical data reflecting individuals' unique biometric features. However, storing this data without subjecting it to any assessment process is of great significance for developing new methods and optimizing current technologies in the field of fingerprint analysis and identification. This way, the existence of incomplete data in the database highlights the necessity for comprehensive research in this area. Such gaps pave the way for the emergence of novel approaches and techniques, thereby enhancing the accuracy, precision, and reliability of fingerprint analysis. In this context, systematic examination and analysis of image records based on pores in the database are crucial for the development and reinforcement of fingerprint-based identification systems. The scientific community can focus on creating new algorithms and methods using this missing data, taking steps toward improving the forensic investigation process and finding more effective solutions in the field of security. Preserving image records based on pores in the unsolved cases database without subjecting them to evaluation criteria is the cornerstone of progress in fingerprint analysis and identification. Careful examination of these records can contribute to the development of more reliable and effective security applications. The research presented offers an innovative approach that goes beyond traditional fingerprint identification methods. In this study, an original dataset was created using a hyperspectral imaging system called "DocuCenter NIRVIS" and the "Projectina Image Acquisition-7000" software, where pores are more thoroughly examined compared to classical fingerprint identification methods. This dataset served as a foundation for a direct pore-based identification system developed for fingerprint matching. While traditional fingerprint identification methods generally rely on general skin surface characteristics, this research emphasizes the focus on pores using hyperspectral imaging. Hyperspectral imaging provides high-resolution images at different wavelengths, allowing for more detailed pore features to be captured. This enabled a more precise analysis of the unique pore patterns of each individual's fingerprint, leading to more reliable identification results. The DocuCenter NIRVIS device and the Projectina Image Acquisition-7000 software created a unique dataset using this hyperspectral imaging approach. This dataset includes hyperspectral images of fingerprints from different individuals, capturing data that defines the unique features of each pore. This dataset forms the basis for the development and testing of a direct pore-based identification system for fingerprint matching. The research presented offers a novel approach aiming to overcome the limitations of traditional fingerprint identification methods and provide a more accurate, precise, and reliable identification system. In this study, pores within the created dataset of 1050 fingerprint images were manually marked using the "Computer Vision Annotation Tool." This stage involves identifying and labeling each pore on each fingerprint. Manual marking was chosen to ensure the accurate and precise identification of pores. Following pore marking, an iterative nearest neighbor algorithm-based scoring system was applied. This algorithm identifies similar pore patterns by comparing different fingerprints within the dataset. This enables the matching of the pores contained in one fingerprint with similar pore patterns in other fingerprints. This step allows for the determination of the unique features of each pore and the comparison of various fingerprints. The scoring system evaluates the similar pore matches identified by the iterative nearest neighbor algorithm. A score is calculated for each match, indicating the degree of similarity between the pores. High-scoring matches are used to identify common pore patterns between different fingerprints. This stage is critical for making more precise matches between different fingerprints and identifying crime scene fingerprints. The research ensures that manually marked pores are successfully analyzed using the iterative nearest neighbor algorithm-based scoring system. This approach not only results in more accurate fingerprint matching but also represents a crucial step in identifying crime scene fingerprints. According to the findings obtained, there is a direct correlation between the number of analyzed pores and the accuracy of marking. These results clearly demonstrate that more detailed and accurate marking of pores directly impacts the reliability and success of matches. The results indicate that increasing the number of analyzed pores and enhancing the accuracy of their marking led to a significant increase in match scores. In other words, more detailed and accurate identification of each pore on each fingerprint has improved the accuracy and reliability of fingerprint matching. These findings demonstrate that precise marking of pores improves the detection of similar patterns and enhances the accuracy of the matching algorithm, leading to more reliable results.Additionally, the query results indicate that the scores for subsequent ranked fingerprint images in the database are notably lower after matching fingerprints. This emphasizes the effectiveness and superiority of the proposed pore-based identification approach compared to other methods. The results show that high scores are achieved for matching fingerprints, while the scores for subsequent fingerprint images in the database are notably lower. This observation highlights the effectiveness of the proposed pore-based approach in identifying and matching unique pore patterns more effectively than other methods. As the study highlights, the detailed analysis of pores and the accurate identification of similar pore patterns contribute to enhancing the accuracy and reliability of fingerprint identification. Therefore, the developed pore-based approach has the potential to yield more accurate, precise, and reliable results in fingerprint identification.
-
ÖgeCharacterization of different shape objects using EM pulse for several different scenarios(Graduate School, 2024-10-07)This thesis presents a thorough investigation into the interaction of electromagnetic pulses (EMP) with various shaped objects, with a particular focus on computational method for characterization. The comprehensive dissertation is anchored in the principles of electromagnetics, with Maxwell's equations serving as the foundation for formulating the theoretical framework essential for analyzing radiation and diffraction phenomena. Initially, the study delves into Maxwell's equations within the context of a uniform dielectric medium characterized by permittivity and permeability. These equations are crucial for describing the behavior of electric and magnetic fields in various media and are presented in their frequency domain forms. The main framework of the thesis revolves around the formulation of problems related to EM radiation and diffraction. By applying the curl operator to Maxwell's equations, the study derives expressions that elucidate how electromagnetic fields propagate and interact with different objects. The thesis meticulously examines the boundary conditions at the surfaces of different objects, essential for deriving the integral equations governing radiation and scattering phenomena. A significant portion of the thesis is devoted to the derivation of functions, which are indispensable for solving the integral equations related to EM problems. These functions assist in characterizing the fields generated by sources in different configurations. The thesis also details the computation of electric and magnetic vector potentials, which are pivotal in understanding how EM fields can be represented and manipulated mathematically. The integral equations, which form the basis for many of the computational methods used in the study, are carefully derived and explained. These foundational steps are critical as they set the stage for the complex simulations and analyses that follow, ensuring that the theoretical underpinnings are robust and reliable. In examining the scattering of EM pulses, the thesis employs various boundary conditions to ensure accurate simulation of the scenarios. These conditions include the continuity of tangential components of electric and magnetic fields across interfaces, which is critical for solving the resulting linear algebraic equations. The inversion of these equations provides the amplitudes of the unknown fields, facilitating the calculation of fields both inside and outside the objects. The handling of boundary conditions is crucial for ensuring the physical accuracy of the simulations, and the thesis provides a detailed account of how these conditions are implemented. This meticulous approach ensures that the simulations reflect realistic physical behaviors, which is paramount for the reliability of the results. The thesis progresses to apply these theoretical constructs to practical problems involving different shapes of objects. The Method of Auxiliary Sources (MAS) is extensively utilized for this purpose. MAS is an efficient computational technique that approximates the scattered fields by placing hypothetical sources around the object. This method's efficacy is particularly highlighted in its application to scenarios where objects need to be rendered invisible over a broad frequency range—a novel extension of MAS within this thesis. Numerical research explores the potential of time domain analysis in addition to frequency domain methods. Moreover, the research extends to shape reconstruction of objects using EM pulses. By analyzing the frequency response of the dielectric objects to EM pulses, the thesis provides a comprehensive characterization of their scattering properties. This dual approach allows for a more detailed understanding of the interactions between EM pulses and objects, offering insights that are not readily apparent from time-domain analysis alone. The frequency-domain analysis also helps identify resonant frequencies at which the scattering characteristics are particularly pronounced, valuable for applications in sensing and detection technologies. In practical applications, the thesis demonstrates the feasibility of using the MAS method for real-time object characterization. The research includes developing algorithm that can quickly process Gaussian EMP signal to reconstruct object shapes and determine material properties. This algorithm is tested using results generated from numerical simulations. The results show that the MAS-based algorithms can accurately and efficiently characterize objects in real-time, making them suitable for deployment in field applications where rapid assessment is required. The thesis delves into shape reconstruction using time-domain analysis, which involves measuring the time delay between the first and second echoes of an incident EMP to determine the object's dimensions. This method proves particularly accurate for objects with low ellipticity. When the object's permittivity is known, both the shape and dimensions can be accurately reconstructed; otherwise, only the shape can be inferred. This technique is vital for non-invasive applications such as medical imaging, where precision and safety are crucial. The research demonstrates how time-domain analysis can significantly enhance shape reconstruction accuracy, offering a promising avenue for further research and practical applications. The thesis also explores the practical implications of its findings, particularly in the field of radar and stealth technology. By optimizing the parameters of objects such as elliptical cylinders and dielectric ellipsoids with high ellipticity, the MAS method effectively minimizes the scattering echoes, making the object less detectable by radar. This has significant implications for military and defense applications, where reducing the radar cross-section of objects is of paramount importance. The thesis provides detailed case studies of how these principles can be applied to real-world scenarios, enhancing the understanding of EMP interactions with various materials and shapes. The case studies are comprehensive, covering different dielectric properties and configurations, and provide valuable insights into the practical applications of the MAS method. One of the other significant applications of the MAS method in this research is the reduction of front echo in Gaussian EMP scattering from 3D dielectric ellipsoids with high ellipticity. By optimizing the parameters of the ellipsoids, such as size and dielectric permittivity, the MAS method effectively minimizes the scattering echoes, making the object less detectable by radar. This has important implications for military and defense applications, particularly in stealth technology and missile design. The thesis presents detailed numerical results to demonstrate the efficacy of the MAS method. Simulations were performed using a specially designed software suite, and the results were visualized to show the electric field amplitudes for dielectric ellipsoids. These results highlight significant reductions in front echo, validating the proposed method's practical utility. Numerical results form a substantial part of the thesis, showcasing the practical applications of the derived formulations. These results not only validate the theoretical models but also demonstrate the capability of the computational method to handle complex scattering problems. The numerical results are presented in a series of detailed graphs and tables, illustrating the effectiveness of the MAS method in reducing computational complexity while maintaining high accuracy. The results highlight the precision with which the MAS method can model the scattering behavior, making it a powerful tool for various applications. Also, the advantages of MAS are demonstrated through a variety of numerical experiments and comparisons with other methods, such as the Method of Moments (MoM) and the Finite-Difference Time- Domain (FDTD) method. These comparisons are essential as they validate the MAS method against established techniques, showcasing its advantages in terms of computational efficiency and accuracy. In summary, this thesis offers a comprehensive analysis of EMP interactions with different shaped objects. It combines rigorous theoretical formulations with advanced computational method, providing significant contributions to the field of electromagnetic research. The methods developed and the numerical results obtained have potential applications in areas such as radar detection, stealth technology, telecommunications, medical imaging, geoscience and material science. The study not only advances technological capabilities but also enriches scientific understanding of electromagnetic interactions with complex objects. The detailed explanations, extensive numerical simulations, and practical applications presented in the thesis make it a valuable resource for researchers and practitioners in the field of electromagnetics. The findings and methodologies presented in this thesis have the potential to influence future research and development in the field, offering new insights and tools for tackling complex electromagnetic problems.
-
ÖgeKesirli kalkülüs ve derin değerlendirme yaklaşımı ile havacılık verilerinin modellenmesi, etki faktörlerinin analizi ve öngörü çalışması(Lisansüstü Eğitim Enstitüsü, 2024-05-21)Bu tez çalışmasında; taşınan hava yolcuları sayısı, hava kargo miktarı, uçuş sayısı, varış noktası sayısı, uluslararası yolcu sayısı, uluslararası varış noktası sayısı, uluslararası uçuş sayısı, arz edilen koltuk kilometre (AKK) bilgisi, AKK uluslararası bilgisi 2011-2019 yılları arasında en yoğun havalimanlarına sahip sekiz ülke olan Almanya, Fransa, İngiltere, Amerika, Hindistan, Singapur, Çin ve Türkiye için seçilmiştir. Seçilen havacılık verileri dikkate alınarak her bir faktörün hesaplanmasında kesirli hesaplama ve derin değerlendirme yöntemlerinden faydalanılacaktır. Çalışmada bildiğimiz kadarıyla türevleri de içeren çok fonksiyonlu olarak henüz ele alınmamış kesirli türevden faydalanılarak; Türevli Çoklu Derin Değerlendirme Metodolojisi (MDAM wd) ile geliştirilen modelleme, öngörü ve etki analizi yapmamıza imkan verecektir. Bu tez çalışmasında yukarıda belirlediğimiz veri seti ile geliştirdiğimiz matematiksel yaklaşımların uygulaması yapılacaktır. İlk olarak modellemek istediğimiz faktörü, etkilendiğini değerlendirdiğimiz kendi ve diğer faktörlerin geçmiş değerleriyle ilişkilendirip böylece her bir faktörün diğer faktörlerin geçmiş değerleriyle etkileşimini ve aynı zamanda geçmiş verilerindeki değişimlerinin ağırlıklı etkisini bulacağız. Sonrasında ise geleceğe ilişkin tahminlerde bulunulacaktır. Bu çalışmada havacılık faktörleri çok az bir sapma ile modellenerek doğru sonuçlar elde edilmiştir, öyle ki modellenen sekiz ülke içinde maksimum hata % 0,015927302 olarak Almanya'nın Hava taşımacılığı, kargo faktörü için bulunmuştur. Faktörlerin öngörülmesi ile ilgili olarak, toplam 72 öngörü için hataların %90.278'i %10'dan küçük olarak bulunmuştur. Önerilen yöntemin sonucu oldukça tatmin edicidir ve kıyaslama yaptığımız önceki türev olmayan yönteme (MDAM) göre daha iyi sonuçlar vermektedir. Ayrıca, 2019 yılı için elde edilen öngörü sonuçları ışığında en öngörülebilir ülkenin İngiltere olduğu tespit edildi; ikinci sırada Almanya, üçüncü sırada ABD, dördüncü sırada Fransa, beşinci sırada Çin, altıncı sırada Singapur, yedinci sırada Türkiye ve sonuncu sırada ise Hindistan oldu. Örneklem için seçilen sekiz ülke içinden en az öngörülebilir faktörler ise Türkiye'nin uluslararası yolcuları ve Singapur'un uçuş sayısı ve uluslararası uçuş sayısı olmuştur. Etki analizi sonuçları verilmiş ve ayrıca ülkeler arasında ortak veya değişken olduğu tespit edilen bazı faktör etkileri çalışmada tartışılmıştır. Faktörlerin diğer faktörler üzerindeki ağırlıklı ortalama etkisi açısından ve geriye dönük olarak 2019 yılı ve 𝑙=2 yıl için önerdiğimiz varsayım çerçevesinde, güçlü pozitif (olumlu) ve güçlü negatif (olumsuz) etkiler değerlendirilmiştir.
-
ÖgeDistilling knowledge of neural networks for image analysis, model compression, data protection and minimization(Graduate School, 2024-07-04)Knowledge distillation is an effective tool for model training, which refers to the process of knowledge transfer between models. In the context of knowledge distillation, the model to be trained with the injected knowledge is named student, where the teacher refers to the model whose knowledge is acquired. It can be exploited for various aims including improving model performance, accelerating the model, and reducing model parameters. Further, with the advent of diverse distillation schemes, it can be efficiently applied in various scenarios and problems. Thus, it has a wide range of application fields including computer vision and natural language processing. This thesis comprises the studies conducted on numerous problems of knowledge distillation, as well as the literature review. The first problem we focus on is hint position selection as an essential element in hint distillation, which is transferring features extracted in intermediate layers, namely hints. First, we demonstrate the importance of the determination of the hint positions. Then, we propose an efficient hint point selection methodology based on layer clustering. For this purpose, we exploit the k-means algorithm with specially designed metrics for layer comparison. We validate our approach by conducting comprehensive experiments utilizing various architectures for teacher-student pairs, hint types, and hint distillation methods, on two well-known image classification datasets. The results indicate that the proposed method achieves superior performance compared to the conventional approach. Another problem focused on in this thesis is model stealing, which refers to acquiring knowledge of a model that is desired to be protected due to the privacy concerns or commercial purposes. Since knowledge distillation can be exploited for model stealing, the concept of the undistillable teacher has been introduced recently, which aims to protect the model from stealing its knowledge via distillation. To contribute to this field, we propose an approach called averager student, whose goal is distilling the undistillable teacher, in this thesis. We evaluate the proposed approach for given teachers which are undistillable or normal. The results suggest that the proposed method outperforms the compared methods whose aim is the same as ours. The last problem we addressed is cross distillation, which means the distillation process between teacher and student models that operate on different modalities. In this work, we introduce a cross distillation scheme that transfers the compressed domain knowledge to the pixel domain. Further, we employ hint distillation which utilizes our previously proposed hint selection method. We evaluate our approach on two computer vision tasks, that are object detection and recognition. The results demonstrate that compressed domain knowledge can be efficiently exploited in a task in the pixel domain via the proposed approach. The proposed approaches in the context of the thesis, contribute to studies on image analysis, model compression, data protection, and minimization. First, our study on the selection of efficient hint positions aims to improve model compression performance, although the proposed approach can also be employed for other distillation schemes. The gains of our method in terms of model compression are presented as well as the performance results of the proposed algorithm. Then, our work on model stealing targets to contribute to the literature on model intellectual property (IP) protection and data protection, where we introduce an algorithm to distill a protected model's knowledge. Moreover, our study on cross distillation provides a contribution to data protection and minimization studies, where we propose a distillation methodology that utilizes compressed domain knowledge on pixel domain problems. Our approach demonstrates a technique that expands limited knowledge by employing different modality data instead of more samples. Since we utilize compressed domain images and eliminate the need for more samples to boost performance, we prevent the use of more data that may be personal or sensitive.