http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Development of a DAQ system for a plasma display panel-based X-ray detector (PXD)
Lee, Hakjae,Jung, Young-Jun,Eom, Sangheum,Kang, Jungwon,Lee, Kisung Elsevier 2015 Nuclear Instruments & Methods in Physics Research. Vol.784 No.-
<P><B>Abstract</B></P> <P>Recently, a novel plasma display panel (PDP)-based X-ray detector (PXD) was developed. The goal of this study is to develop a data acquisition system for use with the PXD as an imaging detector. Since the prototype detector does not have any barrier ribs or a switching device in a detector pixel, a novel pixelation scheme—the line-scan method—is developed for this new detector. To implement line scanning, a multichannel high-voltage switching circuit and a multichannel charge-acquisition circuit are developed. These two circuits are controlled by an FPGA-based digital signal processing board, from which the information about the charge and position of each pixel can be sent to a PC. FPGA-based baseline compensation and switching noise rejection algorithms are used to improve the signal-to-noise ratio (SNR). The characteristic curve of the entire PXD system is acquired, and the correlation coefficients between the X-ray dose, and the signal intensity and the SNR were determined to be approximately 0.99 and 52.9, respectively.</P> <P><B>Highlights</B></P> <P> <UL> <LI> We developed a data acquisition circuit for a novel X-ray imaging detector. </LI> <LI> Line scan, noise rejection, and data transmission methods have been implemented by the FPGA. </LI> <LI> The linearity and SNR of the proposed detector system have been measured quantitatively. </LI> </UL> </P>
Simulation Study of Plasma Display Panel-Based Flat Panel X-Ray Detector
Hakjae Lee,Kisung Lee,Sangheum Eom,Hanho Park,Jungwon Kang IEEE 2013 IEEE transactions on nuclear science Vol.60 No.2
<P>Screen-film-based radiography is being rapidly replaced by digital radiography (DR). Thin-film-transistors (TFT) with amorphous silicon (a-Si) or amorphous selenium (a-Se) are usually used in DR X-ray imaging systems. Another flat panel display, plasma display panel (PDP), has a structure that is similar to that of the conventional gas type radiation detectors, and can be manufactured with lower costs than the TFT-based detector panels. The motivation of this study was to develop a cost-effective DR detector using the PDP. In order to apply the PDP technologies in gaseous detectors for X-ray imaging, we modified the pixel's structure and optimized the materials inside the PDP panel. To maximize the signal's intensity, we re-designed the panel's structure based on the gas microstrip detector (GMD), and estimated the performance of the proposed detector using the Monte Carlo simulation method. Signal intensity of gaseous detector is determined by the amount of ionization as well as by the avalanche effect. The ionization and avalanche processes were simulated using the Geant4 and Garfield, respectively. Four types of gas mixtures and various values of electric fields have been explored. The results show that a higher proportion of Xe helps to generate more ionization electrons. The results also suggest that the electric field, which is applied between anode and cathode strips, is a dominant factor for the avalanche effect to occur. In this study, the GMD structure was adopted for the plasma-display-panel-based X-ray detector. A quantitative verification of the effectiveness of the proposed structure was performed as well.</P>
인공신경망을 이용한 모바일 매니퓰레이터 통합제어 알고리즘에 관한 연구
이상흠(Sangheum Lee),박현준(Hyeonjun Park),김동한(Donghan Kim) 대한전자공학회 2020 대한전자공학회 학술대회 Vol.2020 No.8
Redundancy makes it difficult to control mobile manipulator. When it performs a simple mission, there is no control problem. However, if it faces a crowded environment, it might have control problems. In this study, integrated control algorithms were studied as a method of artificial neural network(ANN) training based on mobile manipulator kinematics data. As the goal position of the end effector of the manipulator is determined, the mobile base is moved to the position where the kinematic solution exists. In that way, the robotic system can be controlled by a single control input.
의료영상 분할 모델의 도메인 일반화 성능 향상을 위한 자기 지도 학습의 활용
이예진(Yejin Lee),이상우(Sangwoo Lee),황상흠(Sangheum Hwang) 대한산업공학회 2021 대한산업공학회지 Vol.47 No.2
In recent years, deep learning technology has been widely used for medical image analysis. However, deep neural networks tend to produce lower generalization performance for data in novel domains, which is a frequent scenario in the field of medical imaging since the domain can be easily shifted by a patient’s physical characteristics and image acquisition equipment. Meanwhile, self-supervised learning is recently known not only to further enhance the performance of a model, but also to improve the robustness of it. Based on this finding, we empirically demonstrated that a model’s domain generalization performance can be improved by using self-supervised pre-training in this study. Moreover, we additionally found that data augmentation applied to the pretext task can significantly impact on domain generalization performance of a model.
대조적 손실 함수를 활용한 영역 분할 모델의 도메인 강건성 개선
이상우(Sangwoo Lee),이예진(Yejin Lee),황상흠(Sangheum Hwang) 대한전기학회 2021 대한전기학회 학술대회 논문집 Vol.2021 No.10
본 논문에서는 도메인 변화에 대한 영역 분할 모델의 강건성을 개선하기 위해 대조적 손실 함수를 영역 분할 문제에 적용하는 새로운 알고리즘을 제안한다. 인코더로부터 추출된 특징 벡터를 클래스에 따라 서로 대조하는 손실 함수의 최소화 과정을 통해 임베딩 공간에서 특징 벡터들이 의미적으로 구분되는 임베딩 성질을 학습한다. 제안한 방법의 효과성을 검증하기 위해 도메인 일반화 연구에서 표준적으로 사용되는 평가 프로토콜 하에 실험을 진행하였으며, 폐 영역 분할 데이터 셋과 U-Net, U-Net++ 모델 구조에서 제안한 방법이 도메인 강건성을 개선하는 데에 효과가 있음을 정량적, 정성적으로 입증하였다.
Similarity based Deep Neural Networks
Seungyeon Lee,Eunji Jo,Sangheum Hwang,Gyeong Bok Jung,Dohyun Kim 한국지능시스템학회 2021 INTERNATIONAL JOURNAL of FUZZY LOGIC and INTELLIGE Vol.21 No.3
Deep neural networks (DNNs) have recently attracted attention in various areas. Their hierarchical architecture is used to model complex nonlinear relationships in high-dimensional data. DNNs generally require large numbers of data to train millions of parameters. However, the training of a DNN with a small number of high-dimensional data can result in an overfitting. To alleviate this problem, we propose a similarity-based DNN that can effectively reduce the dimensionality of the data. The proposed method utilizes a kernel function to calculate pairwise similarities of observations as input, and the nonlinearity based on the similarities is then explored using a DNN. Experiment results show that the proposed method performs effectively regardless of the dataset used, implying that it can be applied as an alternative when learning a small number of high-dimensional data.