• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제12권3호
• /
• pp.250-255
• /
• 2012

Face recognition has wide applications in security and surveillance systems as well as in robot vision and machine interfaces. Conventional challenges in face recognition include pose, illumination, and expression, and face recognition at a distance involves additional challenges because long-distance images are often degraded due to poor focusing and motion blurring. This study investigates the effectiveness of applying photon-counting linear discriminant analysis (Pc-LDA) to face recognition in harsh environments. A related technique, Fisher linear discriminant analysis, has been found to be optimal, but it often suffers from the singularity problem because the number of available training images is generally much smaller than the number of pixels. Pc-LDA, on the other hand, realizes the Fisher criterion in high-dimensional space without any dimensionality reduction. Therefore, it provides more invariant solutions to image recognition under distortion and degradation. Two decision rules are employed: one is based on Euclidean distance; the other, on normalized correlation. In the experiments, the asymptotic equivalence of the photon-counting method to the Fisher method is verified with simulated data. Degraded facial images are employed to demonstrate the robustness of the photon-counting classifier in harsh environments. Four types of blurring point spread functions are applied to the test images in order to simulate long-distance acquisition. The results are compared with those of conventional Eigen face and Fisher face methods. The results indicate that Pc-LDA is better than conventional facial recognition techniques.

• 전자공학회논문지
• /
• 제52권10호
• /
• pp.129-136
• /
• 2015

본 논문에서는 조명 변이에 강인하게 동작 할 수 있는 하이브리드 얼굴 인식 방법을 제안한다. 이를 위해, 서로 다른 특성을 가진 조명 불변 특징 추출 방법으로부터 판별력 있는 특징들을 추출한다. 개별 방법들의 장점들을 효과적으로 활용하기 위해, 판별 거리 척도를 이용하여 각 특징들의 분별력을 측정하여 분별력이 높은 특징들로만 복합 특징을 구성하여 얼굴 인식에 사용한다. Multi-PIE, Yale B, AR, yale database들에 대한 실험 결과, 제안한 방법은 모든 database에 대해 개별 조명 불변 특징 방법들보다 우수한 인식 성능을 보여 주었다.

• 한국지능시스템학회논문지
• /
• 제14권6호
• /
• pp.783-788
• /
• 2004

영상에서 얼굴이 있는 위치를 찾거나 얼굴을 검출하기 위한 많은 방법들이 연구되고 있다. 영상에서 얼굴 검출은 얼굴의 크기, 얼굴이 있는 위치, 그리고 다양한 포즈, 조명 상태 등의 변화에 따라 달라진다 따라서 얼굴 검출과 인식에 있어서의 어려운 점은 얼굴의 크기와 위치, 거리, 조명, 포즈 때문에 나타나는 것이다. 본 논문에서는 다양한 얼굴 크기와 얼굴이 있는 위치 등에 강인한 얼굴 검출을 위해 피셔의 선형 판별 함수를 이용하는 방법을 제안한다. 선형 판별식을 이용하여 효과적으로 얼굴을 검출하기 위해서는 학습 방법 및 학습에 사용되는 데이터들의 구성이 중요하다. 그 이유는, 얼굴 검출을 위해 사용되는 학습 데이터들은 조명과 포즈에 영향을 받기 때문에 얼굴의 특징들을 반영하는 학습 데이터들의 구성이 중요하다. 따라서 본 논문에서는 복잡한 배경과 다양한 크기의 얼굴을 검출하기 위한 계층적인 방법을 제시하며, 효과적인 피셔 판별 분석을 위하여 얼굴과 비얼굴 학습 데이터의 효율적인 분류 방법을 제안한다.

• 한국물환경학회지
• /
• 제27권2호
• /
• pp.147-158
• /
• 2011

This study analyzed the linkage characteristics between road runoff and the nearest streams in mountain regions using a discriminant analysis. The road-to-stream linkage is an important characteristic to evaluate whether the contaminant on road surface is transported directly into the nearby channel system. This study evaluated a total of 51 drainage outlets of mountain roads near the Soyanggang Dam. The linkage between road and stream, slope and width of road, and other information necessary for the discriminant analysis have been collected by in situ investigation and by analyzing the Digital Elevation Model. Finally, as independent variables in the discriminant analysis, the contributing road representing the road characteristics (similar to the runoff from the road drainage outlet) and the distance and slope of the connecting channel between road and nearest stream were selected. Among these three, the distance was found to have the highest discriminant power, the contributing road the lowest. Using the discriminant function derived, 40 out of 51 cases (78.4%) were correctly discriminated and the remaining 11 cases (21.6%) were wrongly discriminated. Reasons of wrongly discriminated cases were mainly due to change in drainage outlet direction, excessive runoff, change in road-to-stream path, etc. This result also indicates that the road-to-stream linkage can be introduced or prohibited by exactly the same way.

• 한국도로학회논문집
• /
• 제18권3호
• /
• pp.127-134
• /
• 2016

PURPOSES : Low visibility caused by dark surroundings at nighttime affects the likelihood of accidents, and various efforts, such as installing road safety facilities, have been made to reduce accidents at night. Despite these efforts, the nighttime severity index (SI) in Korea was higher than the daytime SI during 2011-2014. This study determined the factors affecting daytime and nighttime accident severity through a discriminant analysis. METHODS : Discriminant analysis. RESULTS : First, drowsiness, lack of attention, and lighting facilities affected both daytime and nighttime accident severity. Accidents were found to be caused by a low ability to recognize the driving conditions and a low obstacle avoidance capability. Second, road conditions and speeding affected only the daytime accident severity. Third, failure to maintain a safe distance significantly affected daytime accident severity and nonsignificantly affected nighttime accident severity. The majority of such accidents were caused by rear-end collisions of vehicles driving in the same direction; given the low relative speed difference in such cases, the shock imparted by the accidents was minimal. CONCLUSIONS : Accidents caused by a failure to maintain a safe distance has lower severity than do accidents caused by other factors.

• 한국지능시스템학회논문지
• /
• 제24권6호
• /
• pp.622-627
• /
• 2014

본 논문은 웨이블릿 변환과 선형 판별 분석법 그리고 유전알고리즘을 이용하여 걸음걸이 인식률을 향상시키는 방법을 제안한다. 걸음걸이 에너지 영상에서 웨이블릿 변환으로 분해된 4개의 대역을 얻는다. 분해된 대역을 선형 판별 분석법으로 영상의 특징을 추출한다. 추출된 4개 대역의 특징들과 학습영상의 특징들 사이의 유클리디안 거리를 계산하고, 각 대역에서 계산된 거리 값에 유전알고리즘으로 최적화된 4개의 가중치를 부여한다. 4개 대역의 거리 값과 가중치와의 선형결합으로 계산된 새로운 거리 값을 바탕으로 최근접 이웃 분류 방법을 이용하여 인식 실험을 수행한다. 실험 결과에서 가중치 융합 전 인식률 보다 융합 후 인식률이 더 높은 것을 확인 할 수 있다.

• Korean Chemical Engineering Research
• /
• 제46권2호
• /
• pp.383-388
• /
• 2008

화학공정의 안전하고 효율적인 운전에 관심이 커지면서 공정이상의 원인을 조기에 진단하기 위한 다양한 이상진단방법이 연구되어 왔다. 최근에는 통계적 모델 등 정량적 데이터에 기반한 이상진단방법이 많이 연구되고 있으나, 특정 조업영역에서 얻어진 통계적 모델을 다른 조업영역에 적용하면 오진단이 많아지게 된다. 따라서 공정특성상 다양한 조업영역이 존재하는 화학공정에 데이터기반 방법론을 적용하기에는 어려움이 있어 화학공정의 조업영역 판별법이 요구되고 있다. 이 연구에서는 유클리드 거리(Euclidean distance), FDA(Fisher's discriminant analysis), PCA(principal component analysis)의 통계모델과 이 모델들에 공정변수의 동특성을 반영한 모델을 제안하였다. 6개의 조업모드를 가진 TE(tennessee eastman) 공정에 대한 사례연구를 통해 동특성을 반영한 PCA 모델의 성능이 가장 우수함을 확인하였다.

• 한국환경복원기술학회지
• /
• 제26권6호
• /
• pp.189-203
• /
• 2023

This study analyzed potential breeding sites for black-faced spoonbills on 70 non-breeding, uninhabited islands in Incheon, Korea, in order to suggest potential breeding sites for black-faced spoonbills, whose breeding population has recently been increasing. By comparing the environmental characteristics of breeding and non-breeding areas identified through a literature search, we developed a discriminant to identify potential breeding areas for black-faced spoonbills. Among a total of eight environmental variables(Island area, distance from land, distance to mudflat, distance to rice field, distance to sea route, depth of water, mudflat area, rice field area), the variables that influenced the selection of breeding sites for black-faced spoonbills were average water depth, tidal flat area, and paddy field area. As a result of discriminant analysis of breeding islands using these variables, the accuracy was found to be quite high at 80%. As a result of applying the developed discriminant to non-breeding islands located in the Incheon region, a total of 9 islands(Yongrando, Goseokdo, Beolyeom, Joreumseom, Goeriseom, Hambakdo, Moido, Bigajido, Ahyeom) were identified as potential breeding grounds for spoonbills. The research results can be used as basic data for future management of black-faced spoonbill breeding sites and selection of alternative habitats.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권2호
• /
• pp.742-762
• /
• 2015

In this study, we propose a distance metric learning approach called discriminant metric learning (DML) for face verification, which addresses a binary-class problem for classifying whether or not two input images are of the same subject. The critical issue for solving this problem is determining the method to be used for measuring the distance between two images. Among various methods, the large margin nearest neighbor (LMNN) method is a state-of-the-art algorithm. However, to compensate the LMNN's entangled data distribution due to high levels of appearance variations in unconstrained environments, DML's goal is to penalize violations of the negative pair distance relationship, i.e., the images with different labels, while being integrated with LMNN to model the distance relation between positive pairs, i.e., the images with the same label. The likelihoods of the input images, estimated using DML and LMNN metrics, are then weighted and combined for further analysis. Additionally, rather than using the k-nearest neighbor (k-NN) classification mechanism, we propose a verification mechanism that measures the correlation of the class label distribution of neighbors to reduce the false negative rate of positive pairs. From the experimental results, we see that DML can modify the relation of negative pairs in the original LMNN space and compensate for LMNN's performance on faces with large variances, such as pose and expression.

• Biomedical Engineering Letters
• /
• 제8권4호
• /
• pp.373-382
• /
• 2018

Since epileptic seizure is unpredictable and paroxysmal, an automatic system for seizure detecting could be of great significance and assistance to patients and medical staff. In this paper, a novel method is proposed for multichannel patient-specific seizure detection applying the earth mover's distance (EMD) in scalp EEG. Firstly, the wavelet decomposition is executed to the original EEGs with five scales, the scale 3, 4 and 5 are selected and transformed into histograms and afterwards the distances between histograms in pairs are computed applying the earth mover's distance as effective features. Then, the EMD features are sent to the classifier based on the Bayesian linear discriminant analysis (BLDA) for classification, and an efficient postprocessing procedure is applied to improve the detection system precision, finally. To evaluate the performance of the proposed method, the CHB-MIT scalp EEG database with 958 h EEG recordings from 23 epileptic patients is used and a relatively satisfactory detection rate is achieved with the average sensitivity of 95.65% and false detection rate of 0.68/h. The good performance of this algorithm indicates the potential application for seizure monitoring in clinical practice.