https://scholar.gist.ac.kr/handle/local/7911 2025-12-08T07:47:01Z 2025-12-08T07:47:01Z MA, VAN LINH https://scholar.gist.ac.kr/handle/local/19889 2025-06-30T12:13:04Z 2024-12-31T15:00:00Z

Title: Visual Multi-Object Tracking with Re-Identification using Labeled Random Finite Sets Author(s): MA, VAN LINH Abstract: 본 논문은 베이즈 필터링(Bayesian filtering)을 통해 객체 출현 및 재출현 그리고 가림 현상과 같은 문제를 해결하는 향상된 온라인 비주얼 다중 객체 추적(multi-object tracking)을 기법을 소개한다. 먼저, 객체의 사라짐 및 재출현 문제 해결을 위해, 라벨링 된 확률 유한 집합(labeled random finite set) 필터링 알고리즘을 제안한다. 이는 객체 관리를위해객체들의특징을활용하고,객체검출(detection)에대해선형시간복잡도로 작동하는새로운모델을포함한다.또한,궤적의겹침과크기를고려한퍼지(fuzzy)검출 모델이 가림 현상(occlusion) 처리를 개선하며, 계산 시간을 최소화하기 위해 근사화된 필터도함께제시된다.다음장에서는제안하는다중객체추적알고리즘을 2D이미지에 서실제 3D좌표로확장한다.이접근법은단안카메라의 2D검출들을통합함으로써 3D 다중 객체 추적을 향상시키고, 카메라 재구성 시 검출기 재학습의 필요성을 제거한다. 이를통해트랙초기화및종료,재식별,가림현상처리를단일베이즈필터링재귀로통 합하고, 나아가 객체의 특징, 운동 데이터, 기하학적 가림 모델을 활용하여 제안 기법의 효율성을 높인다. 마지막으로, 2D 바운딩 박스 검출만을 사용하여 실시간 다중 카메라 다중 객체 추적(multi-camera multi-object tracking)을 위한 빠른 온라인 알고리즘이 고 안된다. 이 방법은 다중 센서(multi-sensor) generalized labeled multi-Bernoulli 필터를 – iii – 낮은복잡도와정확도손실없이구현가능하게하고,동적카메라구성에대한강건함을 보여준다.|This dissertation introduces advanced methods for online visual multi-object track- ing (MOT), addressing challenges like object appearance-reappearance and occlusion, using Bayesian filtering techniques. We first introduce an algorithm that leverages la- beled random finite set (LRFS) filtering to tackle disappearance and reappearance issues, incorporating a novel model that utilizes object features to manage reappearing objects with linear complexity relative to the number of detections. A fuzzy detection model is also introduced to enhance occlusion handling by considering track overlaps and sizes. To reduce computational time, we propose an approximation of this filter. In the next chapter, we extend our proposed multi-object tracking algorithm from 2D images to 3D real-world coordinates. More specifically, this approach enhances 3D multi-object tracking by integrating 2D detections from monocular cameras, elim- inating the need for detector retraining upon camera reconfiguration. This solution combines track initiation/termination, re-identification, and occlusion handling into a single Bayes filtering recursion, with improved efficiency through feature and kinematic incorporation, and a geometric occlusion model. In the last chapter, a rapid online al- gorithm is presented for real-time multi-camera multi-object tracking, using only 2D bounding box detections. This method simplifies the Multi-Sensor Generalized La- beled Multi-Bernoulli (MS-GLMB) filter to achieve a low-complexity implementation, demonstrating faster performance without accuracy loss, and robustness to dynamic camera configurations.

2024-12-31T15:00:00Z Hyun Myung Kim https://scholar.gist.ac.kr/handle/local/19883 2025-06-30T12:12:55Z 2022-12-31T15:00:00Z

Title: Vari-focal Light Field Camera for extended depth of field Author(s): Hyun Myung Kim Abstract: In recent years, as interest and demand for the metaverse and self-driving cars such as AR/VR have grown rapidly, great efforts have been made around the world to practicalize 3D depth sensing cameras. The light field camera, one of the three-dimensional depth-sensing cameras, is a camera that can detect depth with only one image sensor and is the most popular technology in practical aspects such as volume, cost, and battery consumption. Light field cameras can extract depth information from images such as those measured by multiple cameras, as microlens arrays act as key optical components for conventional cameras. In the existing light field camera, research on increasing the measurement range, angle of view, and depth resolution in the depth extraction area in the image processing area has been actively conducted. However, research in the hardware area of light field cameras was insufficient. Therefore, in this paper, we propose a study on the hardware implementation and characteristics of light field cameras. First, we propose an efficient processing method for microlens arrays, which are key components of light field cameras. Second, we devised a method of applying a variable-focus lens to widen the measurement distance of the light field camera. Third, we present a method for optimizing the light alignment of the microlens array to broaden the angle of view of the light field camera. Finally, we present a miniaturization method to apply light field cameras to mobile applications. Successful experimental demonstrations represent significant advances in the field of light field camera technology.

2022-12-31T15:00:00Z Kyungho Won https://scholar.gist.ac.kr/handle/local/19850 2025-06-30T12:12:08Z 2021-12-31T15:00:00Z

Title: Towards End-to-End Brain-Computer Interface for the Public: Improving Convenience and Reducing Performance Variation Author(s): Kyungho Won Abstract: 뇌-컴퓨터 인터페이스는 몸 움직임 대신 사람의 뇌파로 외부 환경과 소통할 수 있게하는 인터페이스를 지칭한다. 뇌-컴퓨터 인터페이스는 몸 움직임이 불편한 사람에게 뇌파로 컴퓨터를 조작할 수 있게 함으로써 의사소통, 엔터테이먼트, 그리고 재활에도 활발하게 사용되어 왔다. 몸이 불편한 사용자 층 뿐만 아니라 뇌-컴퓨터 인터페이스는 뇌파 기반 게임에도 활용되며 일반 사용자들의 관심을 받아왔으나, 대중화되기 위한 장애물이 존재한다. 뇌-컴퓨터 인터페이스를 이용하기 위해 시행되는 교정 단계에서 유발되는 불편함과 사람간의 특성의 큰 차이점에서 유발되는 뇌-컴퓨터 인터페이스 성능의 변동성은 뇌-컴퓨터 인터페이스의 대중화를 방해하는 가장 큰 문제점들이며, 뇌-컴퓨터 인터페이스 연구는 오랫동안 한계점을 극복하기 위한 방법론들을 제안해왔다. 본 연구에서는 해당 문제점들을 해결하기 위해 뇌-컴퓨터 인터페이스 별 한계점 극복을 위한 연구들을 제안했다. 첫 번째로, 뇌-컴퓨터 인터페이스의 메커니즘을 이해하고 성능을 예측하기 위해 P300 기반 스펠러의 바이오마커를 탐색하는 연구를 통해 P300 스펠러의 주요 뇌파 특성을 제시했다. 두 번째로, 뇌-컴퓨터 인터페이스를 교정 단계 없이 즉시 이용하기 위한 제로트레이닝 뇌-컴퓨터 인터페이스의 성능 향상을 위해 전이 학습 시 유의미한 학습 데이터를 사용자 수준에서 선정하는 방법을 제안 해 성능 향상을 달성했다. 마지막으로 외부 자극을 이용해 뇌파를 변화시켜 뇌-컴퓨터 인터페이스 성능을 향상시키는 실험을 설계하고 진동자극 및 경두개 직류 전기 자극을 이용해 뇌-컴퓨터 인터페이스 저성능자의 성능을 향상시키고 자극의 종류별 효과를 비교하는 연구를 수행했다. 본 연구에서 공통적으로 관찰한 결과는 뇌-컴퓨터의 저성능자와 고성능자의 뇌파 특성이 구별될 정도로 다르다는 점과 저성능자와 고성능자를 함께 이용하는 것보다 분리해 이용했을 때 뇌-컴퓨터 인터페이스의 성능을 향상시킬 수 있다는 점 이었다. 또한, 저성능자와 고성능자는 뇌-컴퓨터 인터페이스를 하는 동안의 뇌파 패턴 뿐만 아니라, 외부 자극에 대해서도 다른 효과를 보이는 것으로 나타나 궁극적으로 뇌-컴퓨터 인터페이스를 개선하기 위해서는 저성능자와 고성능자를 분리시켜 각각에 대한 심도깊은 연구가 필요한 것으로 짐작된다.|Brain-computer interface (BCI) refers to an interface that enables people to communicate with the external environment through the brain activity instead of the body movements. BCI has facilitated people who need them by providing a tool to control computers or machines using brain activity for communication, entertainment, and rehabilitation. In addition to the patients with movement disorders, BCI has received the attention from the public as it was applied to BCI based game, but the challenging issues still remain. The most challenging issues blocking the popularization of the BCI are discomfort from the calibration session before using the BCI and the performance variation caused by the substantial inter-subject variability, and the BCI researchers have proposed the solutions to address these issues. In this study, we have proposed the research to overcome the challenging issues for a specific BCI system. First, we quantified the key EEG characteristics of the P300 Speller through a study exploring the biomarkers of the P300-based speller to understand the mechanism of the brain-computer interface and predict its performance. Second, we proposed a method to select meaningful subjects as learning data during transfer learning for zero-training motor imagery based BCI that users can use immediately without the calibration phase and achieved the improved BCI performance. Finally, we designed the experiment for enhancing motor imagery based BCI performance of low BCI performers through external stimulation, including vibrotactile stimulation and transcranial direct current stimulation (tDCS), and compared the stimulation effects on the brain activity. In this study, we commonly observed that the neurophysiological characteristics of inefficient BCI users and efficient BCI users were different enough to be classified, and subgroup analysis could enhance the BCI performance by dividing the subjects depending on the BCI performance and analyzing separately rather than applying the same method to all subjects. Moreover, we observed that inefficient BCI users and efficient BCI users showed different effects for the external stimulation as well as the brain patterns during the BCI task, indicating that in-depth investigation should be done for inefficient BCI users and efficient BCI users separately to improve the BCI.

2021-12-31T15:00:00Z Suhyeong Cha https://scholar.gist.ac.kr/handle/local/19816 2025-06-30T12:11:21Z 2022-12-31T15:00:00Z

Title: Theoretical study of gallium-based wide bandgap channel devices using a deterministic multi-subband Boltzmann transport equation solver Author(s): Suhyeong Cha Abstract: Power semiconductors are widely used in automobiles, home appliances, and industrial systems, making a great contribution to the development of human civilization, and have now become part of our lives. Many studies are currently being conducted based on wide bandgap materials that are attracting increasing attention as alternatives to silicon-based power devices. Among them, GaN and β-Ga2O3 are key materials for next-generation power semiconductors with excellent physical properties. Simulations on high-frequency and high-power devices using these wide bandgap materials have been mainly performed through momentum-based transport models such as drift-diffusion (DD) and hydrodynamic models. However, with continuous development, miniaturized power devices now require more precise simulation methods. In this dissertation, a deterministic multi-subband Boltzmann transport equation is implemented to simulate gallium-based wide bandgap devices. The traditional Gummel-type iteration loop is used to obtain a self-consistent solution of the Poisson equation (PE), Schrödinger equation (SE), and Boltzmann transport equation (BTE). GaN-based high electron mobility transistors (HEMTs) and β-Ga2O3-based modulation-doped field-effect transistors (MODFETs) are simulated and the two-dimensional PE is applied to the entire device to obtain electrostatic potential. The one-dimensional SE is used for each position in the transport direction to account for the confinement effect of electrons by the vertical field of the gate contact in a thin channel. The transport properties in a non-equilibrium state are calculated through the one-dimensional BTE. The multidimensional distribution function of the BTE was expanded by the Fourier harmonics, which allows efficient computation by replacing the angular dependence of the two-dimensional momentum space with the harmonic coefficient. In addition, H-transformation, which gives numerical stability, is applied. This sequence of processes is described in chapter 2. In chapter 3, simulations are performed on a GaN-based HEMT with high electron density and high electron mobility based on the polarization effect by the heterostructure. Through comparison with actual benchmarked devices, it was confirmed that the latest miniaturized wide bandgap-based devices can be well simulated using the MS-BTE solver. In chapter 4, optimizations for β-Ga2O3-based MODFET devices are performed. A more realistic structure is considered by implementing a coupled DD and MS-BTE model. It was confirmed that the coupled model could provide more accurate simulation results in optimization studies compared with the calibrated DD model. Continuous research on high-power and high-frequency devices is aimed at miniaturization to improve performance and efficiency in wide bandgap-based devices, as was done in silicon. The deterministic MS-BTE solver of this dissertation can be an alternative to complement the limitations of momentum-based electron transport models commonly used for power semiconductor simulations.

2022-12-31T15:00:00Z