• 대한의용생체공학회:의공학회지
• /
• 제22권4호
• /
• pp.331-338
• /
• 2001

In this paper, we have proposed a new computer retina model reflecting the mechanism of transient amacrine cell on the basis of a conventional computer retina model to understand mechanism of visual information processing. The conventional computer retina model contained most of mechanism for other retina models and it was verified with the physiological data. However, we found that a conventional computer retina model doesn't have the mechanism of amacrine cell that was likely to respond to moving stimulus. In proposed model, therefore, a conventional computer model that considered from photoreceptors to bipolar cells and a new computer model that considered for transient amacrine cell and ganglion cell was combined. As we compared the physiological data with the results of computer simulation of transient amacrine cell about fixed stimulus and moving stimulus, we confirmed that the proposed new computer retina model was normally operated.

• International Journal of Internet, Broadcasting and Communication
• /
• 제14권2호
• /
• pp.136-146
• /
• 2022

Some surface defects make the weak quality of steel materials. To limit these defects, we advocate a one-stage detector model RetinaNet among diverse detection algorithms in deep learning. There are several backbones in the RetinaNet model. We acknowledged two backbones, which are ResNet50 and VGG19. To validate our model, we compared and analyzed several traditional models, one-stage models like YOLO and SSD models and two-stage models like Faster-RCNN, EDDN, and Xception models, with simulations based on steel individual classes. We also performed the correlation of the time factor between one-stage and two-stage models. Comparative analysis shows that the proposed model achieves excellent results on the dataset of the Northeastern University surface defect detection dataset. We would like to work on different backbones to check the efficiency of the model for real world, increasing the datasets through augmentation and focus on improving our limitation.

• 대한전자공학회:학술대회논문집
• /
• 대한전자공학회 2000년도 ITC-CSCC -2
• /
• pp.637-640
• /
• 2000

ERG signal represents the responses of the each layer of retina for the visual stimulus and accumulated responses according to the signal processing occurring in the retina. By investigating the reaction types of each wave of the ERG, various kinds of information for the diagnosis and the signal processing mechanisms in the retina can be obtained. In this paper, the ERG signal is generated by simulating of the volume conductor field of response of each retina layer and summing of them algebraically. The retina model used for simulation is Shah’s Computer Retina model which is one of the most reliable models recently developed. The generated ERG is compared with the typical ERG and shows a very close similarity. By changing the parameters of the retina model, the diagnostic investigation is performed with the variation of the ERG waveform.

• 전자공학회논문지SC
• /
• 제39권6호
• /
• pp.49-58
• /
• 2002

본 논문에서는 망막의 전기생리학적 기전을 바탕으로 실제 망막과 유사한 기능 및 응답 특성을 갖는 망막 모델을 제안하였다. 시세포에서 양극세포까지는 기존에 연구된 여러 망막 모델들을 종합하여 모델링하였고, 3 NDP 기전을 이용하여 움직임 정보를 검출한다고 알려져 있는 아마크린세포와 양극세포 터미널에 관한 새로운 모델을 제안하였다. 이 모델의 평가를 위하여, 공간상의 동적 자극과 정지 자극에 대한 응답 특성을 비교 분석을 하였을 뿐만 아니라 자극의 움직임 속도에 따른 특성에 대한 분석도 수행하여 제안한 망막 모델을 검증하였다. 본 연구결과는 움직임 정보를 검출하기 위한 비전 시스템에 대한 인간 시각 시스템의 적용 및 생체에 이식할 수 있는 인공 망막의 개발을 위한 기초 연구에 이용될 수 있을 것이다.

• 대한의용생체공학회:의공학회지
• /
• 제9권1호
• /
• pp.17-24
• /
• 1988

This paper describes a neural network modelling of a vertebrate retina using a discrete-time and discrete-space approach based on neuro-anatomical data, and the computer simulations of the model which approximate the frog/amphibian negro-physiological data. It then compares them and describes how such a model can be beneficially used for confirming the hypothesis of a given neural system and further predict yet unknown experimental data.

• Journal of information and communication convergence engineering
• /
• 제2권2호
• /
• pp.132-137
• /
• 2004

Most of the current computer vision theories are based on hypotheses that are difficult to apply to the real world, and they simply imitate a coarse form of the human visual system. As a result, they have not been showing satisfying results. In the human visual system, there is a mechanism that processes information due to memory degradation with time and limited storage space. Starting from research on the human visual system, this study analyzes a mechanism that processes input information when information is transferred from the retina to ganglion cells. In this study, a model for the characteristics of ganglion cells in the retina is proposed after considering the structure of the retina and the efficiency of storage space. The MNIST database of handwritten letters is used as data for this research, and ART2 and SOM as recognizers. The results of this study show that the proposed recognition model is not much different from the general recognition model in terms of recognition rate, but the efficiency of storage space can be improved by constructing a mechanism that processes input information.

• 한국정보통신학회논문지
• /
• 제26권4호
• /
• pp.519-525
• /
• 2022

영상 내 사람의 얼굴을 검출하는 얼굴 검출 작업은 다양한 영상 처리 어플리케이션 내 전처리 또는 핵심 과정으로 사용되고 있다. 최근 딥러닝 기술의 발달로 높은 성능을 내고 있는 신경망 모델은 2차원 영상에 의존적이며, 카메라 품질이 떨어지거나, 얼굴의 초점을 제대로 잡지 못하는 등의 영상 내 노이즈가 발생할 경우, 제대로 얼굴을 검출하지 못할 수 있다. 본 논문에서는 2차원 영상의 의존성을 낮추기 위해 깊이 정보를 함께 사용하는 얼굴 검출 방법에 대해 제안한다. 제안하는 모델은 기존 공개된 얼굴 검출 데이터 셋을 이용하여 깊이 정보를 사전에 생성 및 전처리 과정을 거친 후 학습하였으며, 그 결과, 평균 정밀도 기준 FRN 모델은 89.16%로 87.95%의 성능을 보인 RetinaNet 모델보다 약 1.2% 정도의 성능이 향상되었음을 확인하였다.

• 전력전자학회논문지
• /
• 제8권1호
• /
• pp.17-23
• /
• 2003

컴퓨터 모니터를 통해 안구망막의 형태와 두께를 보기 위해서 레이저의 경로차를 이용하는 SLO 장비가 사용되고 있다. 이러한 방법으로 망막의 선명한 3차원 영상을 보기 위해서는 레이저 광경로 시스템의 정확한 동기제어가 필요하다. 이 영상을 얻기 위해서는 평면주사를 하는데 있어서 정밀동기제어가 매우 중요하다. 본 논문에서는 안구의 3차원 영상을 만들기 위해 갈바노미터의 동기제어를 구현한다. 또한 갈바노미터의 간략한 수학적 모델의 타당성을 보인다.

• Journal of Multimedia Information System
• /
• 제8권2호
• /
• pp.79-84
• /
• 2021

Detecting brain tumors of different sizes is a challenging task. This study aimed to identify brain tumors using detection algorithms. Most studies in this area use segmentation; however, we utilized detection owing to its advantages. Data were obtained from 64 patients and 11,200 MR images. The deep learning model used was RetinaNet, which is based on ResNet152. The model learned three different types of pre-processing images: normal, general histogram equalization, and contrast-limited adaptive histogram equalization (CLAHE). The three types of images were compared to determine the pre-processing technique that exhibits the best performance in the deep learning algorithms. During pre-processing, we converted the MR images from DICOM to JPG format. Additionally, we regulated the window level and width. The model compared the pre-processed images to determine which images showed adequate performance; CLAHE showed the best performance, with a sensitivity of 81.79%. The RetinaNet model for detecting brain tumors through deep learning algorithms demonstrated satisfactory performance in finding lesions. In future, we plan to develop a new model for improving the detection performance using well-processed data. This study lays the groundwork for future detection technologies that can help doctors find lesions more easily in clinical tasks.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2000년도 제15차 학술회의논문집
• /
• pp.496-496
• /
• 2000

We propose a neuron computer for tracking motion of particles in multi-dimensional space. The neuron computer is constructed of neural networks and their connections, which is a simplified model of the brain. The neuron computer is assemblage of neural networks, it includes a control unit, and the actions of the unit are represented by instructions. We designed a neuron computer to recognize and predict motion of particles. The recognition unit is constructed of neuron-array, encoder, and control part. The neuron-array is a model of the retina, and particles crease an image on the array, where the image is binary. The encoder picks one particle from the array, and translates the particle's location to Cartesian coordinates, which is scaled in [0, 1] intervals. Next, the encoder picks another particle, and does same process. The ordering and reduction of complex processes are executed by instructions. The instructions are held in the control part. The prediction unit is constructed of a multi-layer neural network and a feedback loop, where real time learning is executed. The particles' future locations are forecasted by coordinate values. The neuron computer can chase maximum 100 particles that take evasions.