• 한국정보과학회논문지:소프트웨어및응용
• /
• 제33권5호
• /
• pp.481-488
• /
• 2006

부공간 투영기술(subspace projection)을 이용한 얼굴인식기술의 성능은 이들 기저영상들(basis images)의 특징과 밀접한 관련이 있다. 특히 표정변화와 같은 국부적 왜곡이나 오클루전이 있는 경우의 인식성능은 기저영상들의 특징에 의해 영향을 받게 된다. 부공간 투영기반의 얼굴인식 방법이 오클루전이나 표정변화와 같은 국부적인 왜곡발생에 강인하려면 부분국부적 표현(part-based local representation)의 기저벡터를 갖는 것이 중요하다. 본 연구에서는 국부적 왜곡과 오클루전에 강인한 효과적인 부분국부적 표현방법을 제안한다. 제안한 방법을 LS-ICA(locally salient ICA) 방법이라고 명명하였다. LS-ICA방법은 ICA 구조I의 기저영상을 구하는 과정에서 공간적인 국부성(locality)의 제약조건을 부과함으로써 부분국부적 기저영상(part-based local basis images)을 얻는 방법이다. 결과적으로 공간적으로 현저한 특징만을 포함하는 기저영상을 사용하게 되며, 이는 "Recognition by Parts"의 방법론과 유사하다. LS-ICA방법과 LNMF(Localized Non-negative Matrix Factorization)와 LFA(Local Feature Analysis)와 같은 기존의 부분 표현방법(part-based representation)들에 대해 다양한 얼굴영상 데이타베이스를 사용하여 실험한 결과, LS-ICA방법이 기존의 방법에 비하여 높은 인식성능을 보였으며, 특히 오클루전이나 국부적인 변형이 포함된 얼굴영상에서 뛰어난 인식성능을 보였다.

• 한국정보보호학회:학술대회논문집
• /
• 한국정보보호학회 2006년도 하계학술대회
• /
• pp.251-257
• /
• 2006

The performance of face recognition methods using subspace projection is directly related to the characteristics of their basis images, especially in the cases of local distortion or partial occlusion. In order for a subspace projection method to be robust to local distortion and partial occlusion, the basis images generated by the method should exhibit a part-based local representation. We propose an effective part-based local representation method named locally salient ICA (LS-ICA) method for face recognition that is robust to local distortion and partial occlusion. The LS-ICA method only employs locally salient information from important facial parts in order to maximize the benefit of applying the idea of 'recognition by parts.' It creates part-based local basis images by imposing additional localization constraint in the process of computing ICA architecture I basis images. We have contrasted the LS-ICA method with other part-based representations such as LNMF (Localized Non-negative Matrix Factorization)and LFA (Local Feature Analysis). Experimental results show that the LS-ICA method performs better than PCA, ICA architecture I, ICA architecture II, LFA, and LNMF methods, especially in the cases of partial occlusions and local distortion

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권7호
• /
• pp.3172-3193
• /
• 2018

Recently, several studies have shown that linear representation based approaches are very effective and efficient for image classification. One of these linear-representation-based approaches is the Collaborative representation (CR) method. The existing algorithms based on CR have two major problems that degrade their classification performance. First problem arises due to the limited number of available training samples. The large variations, caused by illumintion and expression changes, among query and training samples leads to poor classification performance. Second problem occurs when an image is partially noised (contiguous occlusion), as some part of the given image become corrupt the classification performance also degrades. We aim to extend the collaborative representation framework under limited training samples face recognition problem. Our proposed solution will generate virtual samples and intra-class variations from training data to model the variations effectively between query and training samples. For robust classification, the image patches have been utilized to compute representation to address partial occlusion as it leads to more accurate classification results. The proposed method computes representation based on local regions in the images as opposed to CR, which computes representation based on global solution involving entire images. Furthermore, the proposed solution also integrates the locality structure into CR, using Euclidian distance between the query and training samples. Intuitively, if the query sample can be represented by selecting its nearest neighbours, lie on a same linear subspace then the resulting representation will be more discriminate and accurately classify the query sample. Hence our proposed framework model the limited sample face recognition problem into sufficient training samples problem using virtual samples and intra-class variations, generated from training samples that will result in improved classification accuracy as evident from experimental results. Moreover, it compute representation based on local image patches for robust classification and is expected to greatly increase the classification performance for face recognition task.

• 한국정보통신학회논문지
• /
• 제23권6호
• /
• pp.682-689
• /
• 2019

부분 기반 영상 표현(part-based image representation)에서는 영상의 부분적인 모습을 기저 벡터로 표현하고 기저 벡터의 선형 조합으로 영상을 분해하며, 이 때 기저 벡터의 계수가 곧 물체의 부분적인 특징을 의미하게 된다. 본 논문에는 부분 기반 영상 표현 기법인 비음수 행렬 분해(non-negative matrix factorization, NMF)를 이용하여 얼굴 영상을 표현하고 신경망 기법을 적용하여 가려진 얼굴을 인식하는 얼굴 인식을 제안한다. 표준 비음수 행렬 분해, 투영 경사 비음수 행렬 분해, 직교 비음수 행렬 분해를 이용하여 얼굴 영상을 표현하였고, 각 기법의 성능을 비교하였다. 인식기로는 학습벡터양자화 신경망을 사용하였으며, 인식기에서의 거리 척도로는 유클리디언 거리를 사용하였다. 실험 결과, 전통적인 얼굴 인식 방법에 비하여 제안한 기법이 가려진 얼굴 인식에 보다 강인함을 보인다.

• 한국CDE학회논문집
• /
• 제9권1호
• /
• pp.51-61
• /
• 2004

Mechanical parts are often grouped into part families based on the similarity of their shapes, to support efficient manufacturing process planning and design modification. The 2-part sequence papers present similarity assessment techniques to support part family classification for machined parts. These exploit the multiple feature decompositions obtained by the feature recognition method using convex decomposition. Convex decomposition provides a hierarchical volumetric representation of a part, organized in an outside-in hierarchy. It provides local accessibility directions, which supports abstract and qualitative similarity assessment. It is converted to a Form Feature Decomposition (FFD), which represents a part using form features intrinsic to the shape of the part. This supports abstract and qualitative similarity assessment using positive feature volumes.. FFD is converted to Negative Feature Decomposition (NFD), which represents a part as a base component and negative machining features. This supports a detailed, quantitative similarity assessment technique that measures the similarity between machined parts and associated machining processes implied by two parts' NFDs. Features of the NFD are organized into branch groups to capture the NFD hierarchy and feature interrelations. Branch groups of two parts' NFDs are matched to obtain pairs, and then features within each pair of branch groups are compared, exploiting feature type, size, machining direction, and other information relevant to machining processes. This paper, the second one of the two companion papers, describes the similarity assessment method using NFD.

• 한국CDE학회논문집
• /
• 제9권1호
• /
• pp.44-50
• /
• 2004

Mechanical parts are often grouped into part families based on the similarity of their shapes, to support efficient manufacturing process planning and design modification. The 2-part sequence papers present similarity assessment techniques to support part family classification for machined parts. These exploit the multiple feature decompositions obtained by the feature recognition method using convex decomposition. Convex decomposition provides a hierarchical volumetric representation of a part, organized in an outside-in hierarchy. It provides local accessibility directions, which supports abstract and qualitative similarity assessment. It is converted to a Form Feature Decomposition (FFD), which represents a part using form features intrinsic to the shape of the part. This supports abstract and qualitative similarity assessment using positive feature volumes. FFD is converted to Negative Feature Decomposition (NFD), which represents a part as a base component and negative machining features. This supports a detailed, quantitative similarity assessment technique that measures the similarity between machined parts and associated machining processes implied by two parts' NFDs. Features of the NFD are organized into branch groups to capture the NFD hierarchy and feature interrelations. Branch groups of two parts' NFDs are matched to obtain pairs, and then features within each pair of branch groups are compared, exploiting feature type, size, machining direction, and other information relevant to machining processes. This paper, the first one of the two companion papers, describes the similarity assessment methods using convex decomposition and FFD.

• 대한기계학회논문집
• /
• 제17권10호
• /
• pp.2446-2456
• /
• 1993

By virtue of progress of computer science, CAD/CAM technology has been developed greatly in each area. But the problems in the integration of CAD/CAM are not yet solved completely. The reason is that the exchange of data between CAD and CAM is difficult because the domains of design and manufacturing are different in nature. To solve this problem, a feature based modeller is developed in this study, which makes it possible to communicate between design and manufacturing through features. The modeller has feature, the concept of semi-bounded plane is introduced, and implemented as a B-rep sheet model using half-edge data structure. The features are then created on a part by local modification of the boundary on a part based on feature template information. This approach generalizes the modelling of features in a geometry model.

• 한국통신학회논문지
• /
• 제37권7C호
• /
• pp.554-564
• /
• 2012

차량인식은 차량 후보영역 검출단계와 검출된 후보 영역에서 특징을 기반으로 차량을 검증하는 차량 검증단계로 나누어진다. 선형 변환 방식의 특징은 차원 감소 효과와 통계적인 특징을 지니게 되어, 이동이나 회전에 강인한 특성을 갖는다. 선형 변환 방식 중 비음수행렬분해(Non-negative Matrix Factorization, NMF)는 부분 기반 표현 방식으로 차량의 국소적인 특징을 기저벡터로 사용하여 희소성을 갖는 특징을 추출할 수 있기 때문에 도심영상에서 발생하는 차폐 영역에 따른 인식률 저하를 방지할 수 있다. 본 논문에서는 차량 인식에 적합한 NMF 특징 추출 방법을 제안하고, 인식률을 검증하였다. 또한 희소성 제약 조건을 이용하여 기저 벡터에 희소성을 가지는 SNMF(Sparse NMF)와 LVQ2(Learning Vector Quantization) 신경 회로망을 결합하여 차폐 영역에 대한 차량 인식 효율을 기존의 NMF를 이용한 방법과 비교하였다. NMF를 이용하는 특징은 차량이 혼재되어 차폐 영역이 빈번히 발생하는 도심에서도 강건한 특징임을 보였다.

• 대한산업공학회지
• /
• 제18권2호
• /
• pp.11-22
• /
• 1992

Up to now, most researches on production automation have concentrated on local automation, e. g. CAD, CAM, robotics, etc. However, to achieve total automation it is required to link each local modules such as CAD, CAM into a unified and integrated system. One such missing link is between CAD and computer vision system. This thesis is an attempt to link the gap between CAD and computer vision system. In this paper, we propose algorithms that carry out edge detection, thinning and pruning from the image data of manufactured parts, which are obtained from video camera and then transmitted to computer. We also propose a feature extraction and surface determination algorithm which extract informations from the image data. The informations are compatible to IGES CAD data. In addition, we suggest a methodology to reduce search efforts for CAD data bases. The methodology is based on graph submatching algorithm in GEFG(Generalized Edge Face Graph) representation for each part.

• 정보처리학회논문지A
• /
• 제8A권1호
• /
• pp.48-55
• /
• 2001

Using BDD, we can represent Boolean functions uniquely and compactly, Hence, BDD have become widely used for CAD applications, such as logic synthesis, formal verification, and etc. The size of the BDD representation for a function is very sensitive to the choice of orderings on the input variables. Therefore, it is very important to find a good variable ordering which minimize the size of the BDD. Since finding an optimal ordering is NP-complete, several heuristic algorithms have been proposed to find good variable orderings. In this paper, we propose a variable ordering algorithm based on the $\mu$O(microcanonical optimization). $\mu$O consists of two distinct procedures that are alternately applied : Initialization and Sampling. The initialization phase is to executes a fast local search, the sampling phase leaves the local optimum obtained in the previous initialization while remaining close to that area of search space. The proposed algorithm has been experimented on well known benchmark circuits and shows superior performance compared to a algorithm based on simulated annealing.