해마신경망과 SURF를 이용한 실시간 얼굴인식

오인권,김현진,남궁재찬 한국정보기술학회 2012 한국정보기술학회논문지 Vol. No.

In this paper, A real-time face learning/recognition system will be proposed. Face learning and recognition of the system shall apply to the HNMA(Hippocampal Neuron Modeling Algorithm) using SURF(Speed Up Robust Features)’ features that can be extracted quickly local features from images. First, the face recognition database to be used for recognition process is learned in the hippocampus, and a SURF’ features set distributed in each image is stored separately, recognition is in progress through a matching process to the stored features set of SURF that are extracted from images for camera. Therefore, The system are shown to contain robust recognition performance against various environmental changes through SURF’ features. In the proposed method was the recognition results of 94.8% and colledted from the Internet, video recognition performance was 88.3%. 본 논문에서는 영상에서 지역적 특성을 빠르게 추출할 수 있는 SURF(Speed Up Robust Features)의 특징을 해마신경망 모델링 알고리즘(HNMA:Hippocampal Neuron Modeling Algorithm)에 적용하여 실시간으로 얼굴을 학습하고 인식하는 시스템을 제안한다. 먼저 인식에 사용될 얼굴 이미지 데이터베이스를 해마신경망에 학습시켜, 각 얼굴 영상에 분포하는 SURF 특징 집합을 구분하여 저장하고, 카메라를 통해서 입력되는 영상에서 얼굴을 추출하여 저장된 SURF 특징 집합과 정합과정을 통해, 인식하는 과정을 거친다. 지역적 특성을 갖는 SURF 특징 집합의 특성을 통해서 다양한 환경적 변화에서도 강건한 인식성능을 보이는 것으로 확인하였다. 본 연구가 제안하는 방법에서 제한된 환경에서의 인식성능은 94.8%의 인식결과를 보였고, 인터넷에서 수집한 동영상에서의 인식 성능은 88.3%의 인식결과를 보였다.

MPEG-4 기반의 영상전화기 구현을 위한 실시간 변환영역(객체) 추출에 관한 알고리즘

오인권,손영우,남궁 재찬,Oh, In-Gwon,Shon, Young-Woo,Namgung, Jae-Chan 한국통신학회 2004 韓國通信學會論文誌 Vol.29 No.1C

논문에서는 MPEG-4(Moving Picture Expert Group-4) 객체기반 부호화를 위하여 영상에서 실 시간적으로 변화영역(객체)을 추출하는 알고리즘에 대하여 제안한다. 기존의 객체 분리방법 들은 Off-Line 방법으로 객체를분리하므로 실시간 처리를 필요로 하는 영상전화나 영상회의 시스템에서는 사용할 수 없었다. 그리고 또 MPEG-4표준의 버전1에서 권장하는 객체분할 방식인 공간적인 분할(Spatial Segmentation)방법과 시간적인 분할(Temporal Segmentation)방법은 픽셀단위로 연산을 하므로 연산의 복잡도가 높아서 실시간 영상전송에 어렵다. 그러나 이 논문에서 제안하는 알고리즘은 연산단위를 픽셀단위로 연산하는 것이 아니라 매크로블록 단위로 연산이 이루어지므로 실시간 전송을 가능케 한다. 그러나MPEG-4권고 안에서 제시한 알고리즘처럼 이 번에 제안한 알고리즘도 한 영상에서 여러 개의 객체를 추출하는 것이 이루어지지 않았다. 그리고 전체 시스템 구성을 보면 크게 부호기와 복호기로 나누어지고 부호기에 본 논문에서 제안한 실시간 객체추출 알고리즘이 전처리 단으로 삽입되어 구현되었다. This paper proposes a algorithm to extract the variety region (object) from video for the real-time encoding of MPEG-4 based. The previous object segmentation methods cannot used the videophone or videoconference required by real-time processing. It is difficult to transfer a video to real-time because it increased complexity for the operation of each pixel on the spatial segmentation and temporal segmentation method proposed by MPEG-4 Working Group. But algorithm proposed for this thesis not operates a pixel unit but operates a macro block unit. Thus this enables real-time transfer. But this algorithm cannot extract several object for a image using proposed algorithm as previous algorithm. On system constructed by encoder and decoder. A proposed algorithm inserted for encoder as pre-process.

해마신경망과 SURF를 이용한 실시간 얼굴인식

오인권(In-Gwon Oh),김현진(Hyun-Jin Kim),남궁재찬(Jae-Chan Namgung) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.1

In this paper, A real-time face learning/recognition system will be proposed. Face learning and recognition of the system shall apply to the HNMA(Hippocampal Neuron Modeling Algorithm) using SURF(Speed Up Robust Features)’ features that can be extracted quickly local features from images. First, the face recognition database to be used for recognition process is learned in the hippocampus, and a SURF’ features set distributed in each image is stored separately, recognition is in progress through a matching process to the stored features set of SURF that are extracted from images for camera. Therefore, The system are shown to contain robust recognition performance against various environmental changes through SURF’ features. In the proposed method was the recognition results of 94.8% and colledted from the Internet, video recognition performance was 88.3%.

스틱 텐서 보팅 정규화를 이용한 얼굴인식률 개선

오인권(In-gwon Oh),손영우(Young-woo Shon),이한우(Han-woo Lee) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.11

Face recognition is significantly affected by environmental factors such as lighting, distance, pose, accessories, and noise. Among them, the distance of a camera and a subject of recognition can be seen as more fundamental problem than other environmental factors. Improving the recognition rate according to the distance mainly rely on image interpolation. In this paper, tensor voting normalization method is proposed to minimize the lost information from the enlargement of the face image. Tensor voting is a way to compensate the damaged areas through voting among stick tensors on the distribution of the image. This is the best effective way to minimize the damaged areas occurs in the process of enlarging image. By the experiments, we are able to check that tensor voting interpolation is the effective method to compensate a small-size image compared to the existed edge adaptive interpolation. It causes to improve the face recognition rate.

SIFT 특징벡터와 해마신경망을 이용한 얼굴 표정인식

오인권(In-Gwon Oh),남궁재찬(Jae-Chan Namgung) 한국정보기술학회 2012 한국정보기술학회논문지 Vol.10 No.2

In this paper, the characteristics of SIFT(Scale Invariant Feature Transform) algorithm applied to the hippocampus neural modeling algorithm(HNMA) to recognize facial expressions of the system is proposed. The system roughly consists of three parts; the first is a part to extract the features from a facial image (extraction), the second is a part to learn the features (learning) and the last is a part to recognize a facial image using knowledges based on the learned features (recognition). Features extraction part is extracted from the eyes and mouth area to obtain SIFT features, it is configured to be compared with the trained images. In learning part, Features of images input by hippocampus structure is classified according to the favorable characteristics as long-term memory and short-term memory. In this study, 93.5% of the proposed method, face recognition results showed the amount of training data, the more recognition you receive for that improvement can be seen.

이동통신네트워크에서 핸드오프 알고리즘을 이용한 적응형 동적 재라우팅 기법에 관한 연구

김일태,오인권,이진우 건국대학교 1997 학술논문집 : 건국대 대학원 Vol.45 No.1

In consequence of cellular network architecture evolution, a cell size is getting small such as micro/pico cell in new generation communication(FPLMTS). Due to the frequent handoffs in the pico cell, re-connection setup delay and the number of pre-established connections for the seamless handoff might be increased on ATM-based transport network. To reduce handoff processing delay and the number of pre-established connections, this paper proposes a new re-routing scheme. In the scheme, in order to obtain paths for handoff prior to handoff, the new cell architecture and the scheme of searching paths is also suggested. We conclude by comparing the proposed re-routing scheme to the existing schemes in the literature.

음폐수처리와 바이오에너지 생산을 위한 미생물연료전지

이기환,오인권,최정동 대한환경공학회 2017 대한환경공학회 학술발표논문집 Vol.2017 No.11

ICA-SIFT를 이용한 물체인식

김현오(Hyun-O Kim),오인권(In-Gwon Oh),남궁재찬(Jae-Chan Namkung) 한국정보기술학회 2010 한국정보기술학회논문지 Vol.8 No.4

A study on the object recognition is a main interest in the field of the image processing and various applied researches. The one method of the object recognitions is feature based object recognition. SIFT (Scale Invariant Feature Transform) algorithm which is robust in illumination and has the quality which is not relationship to size and rotation of image was proposed. However, there is a disadvantage that feature matching performance is gone down when camera movies or view angle of object changes. This paper presents ICA-SIFT(Independent Component Analysis-Scale Invariant Feature Transform) algorithm to improve recognition performance. it can be obtained more distinctive feature matching performance through this method and improved in accuracy as ave. 7.9% than SIFT. also, it can be possible that object recognition is more correct using reliable matching data through eliminating mismatching data.

PCA&LDA-SIFT 알고리즘을 이용한 얼굴인식 향상에 관한 연구

이세진(Se-Jin Lee),오인권(In-Gwon Oh),남궁재찬(Jae-Chan Namgung) 한국정보기술학회 2008 Proceedings of KIIT Conference Vol.2008 No.-

PCA&LDA-SIFT 알고리즘을 이용한 얼굴인식 성능의 향상

이세진(Se-Jin Lee),오인권(In-Gwon Oh),남궁재찬(Jae-Chan Namkung) 한국정보기술학회 2010 한국정보기술학회논문지 Vol.8 No.6

Face recognition is actively being studied in image processing, pattern recognition, computer vision and neural network. It much developed through various studies in the past. Yet it is possible only in restricted environment because of changes in illumination or changes in image including facial size according to distance. This paper used among methods of recognizing face SIFT(Scale Invariant Feature Transform) algorithm which is characterized by being unyielding against illumination and invariant against the size and rotation of an image. However, the algorithm with the disadvantage of deteriorating keypoint matching performance becomes a factor of lowering recognition performance. To improve it, we proposed an algorithm of grafting a PCA&LDA fusion model into SIFT algorithm descriptor, and compared and analyzed it in performance with the existing recognition algorithms such as PCA(Principal Component Analysis), LDA(Linear Discirminant Analysis) and SIFT.