• 전자공학회논문지B
• /
• 제31B권5호
• /
• pp.141-148
• /
• 1994

Both the neural network classifier utilizing multi-layer perceptron and the linear tree classifier composed of hierarchically structured linear discriminating functions can form arbitrarily complex decision boundaries in the feature space and have very similar decision making processes. In this paper, a new method for automatically choosing the number of neurons in the hidden layers and for initalzing the connection weights between the layres and its supporting theory are presented by mapping the sequential structure of the linear tree classifier to the parallel structure of the neural networks having one or two hidden layers. Experimental results on the real data obtained from the military ship images show that this method is effective, and that three exists no siginificant difference in the classification acuracy of both classifiers.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권9호
• /
• pp.3243-3257
• /
• 2021

Deep neural networks provide excellent performance in pattern recognition, audio classification, and image recognition. It is important that they accurately recognize input data, particularly when they are used in autonomous vehicles or for medical services. In this study, we propose a data correction method for increasing the accuracy of an unknown classifier by modifying the input data without changing the classifier. This method modifies the input data slightly so that the unknown classifier will correctly recognize the input data. It is an ensemble method that has the characteristic of transferability to an unknown classifier by generating corrected data that are correctly recognized by several classifiers that are known in advance. We tested our method using MNIST and CIFAR-10 as experimental data. The experimental results exhibit that the accuracy of the unknown classifier is a 100% correct recognition rate owing to the data correction generated by the proposed method, which minimizes data distortion to maintain the data's recognizability by humans.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권7호
• /
• pp.1858-1872
• /
• 2023

With the advance of radar technologies, the availability of synthetic aperture radar (SAR) images increases. To improve application of SAR images, a management system for SAR images is proposed in this paper. The system provides trainable land cover classification module and display of SAR images on the map. Users of the system can create their own classifier with their data, and obtain the classified results of newly captured SAR images by applying the classifier to the images. The classifier is based on convolutional neural network structure. Since there are differences among SAR images depending on capturing method and devices, a fixed classifier cannot cover all types of SAR land cover classification problems. Thus, it is adopted to create each user's classifier. In our experiments, it is shown that the module works well with two different SAR datasets. With this system, SAR data and land cover classification results are managed and easily displayed.

• 한국정보과학회논문지:컴퓨팅의 실제 및 레터
• /
• 제16권10호
• /
• pp.955-961
• /
• 2010

본 논문에서는 HOG 특정벡터와 영상분할을 이용한 부스팅 분류기반의 자동차영역 검출 알고리즘의 연구에 대해서 기술한다. 입력된 영상으로부터 차량을 검출하기위해 먼저 분할 후 합병(split-merge) 방법을 적용하여 영상을 분할한다. 그리고 가장 큰 두 영역을 검색 영역에서 제외하여 처리 속도를 향상 시킨다. 각 영역에 대해 HOG(histogram of oriented gradient) 특정을 추출한다. 분류기는 두 개의 모집단을 분류하는데 많이 사용되고 있는 AdaBoost 방법을 사용한다. 제안방법의 성능 평가를 위해 537개의 영상을 사용하여 분류기를 학습하였으며, 또한 학습에 사용하지 않은 비학습영상 500개를 사용하여 인식률을 구하였다. 실험결과 비학습영상에 대해 98.34%의 인식률을 얻었다. 결론적으로 제안된 방법이 지능형 자동차 제어 시스템에서 차량의 위치를 찾는 방법으로 활용될 수 있다.

• 전기학회논문지
• /
• 제64권5호
• /
• pp.755-765
• /
• 2015

This paper is concerned with Interval Type-2 Radial Basis Function Neural Network classifier realized with the aid of Census Transform(CT) and (2D)2LDA methods. CT is considered to improve performance of face recognition in a variety of illumination variations. (2D)2LDA is applied to transform high dimensional image into low-dimensional image which is used as input data to the proposed pattern classifier. Receptive fields in hidden layer are formed as interval type-2 membership function. We use the coefficients of linear polynomial function as the connection weights of the proposed networks, and the coefficients and their ensuing spreads are learned through Conjugate Gradient Method(CGM). Moreover, the parameters such as fuzzification coefficient and the number of input variables are optimized by Artificial Bee Colony(ABC). In order to evaluate the performance of the proposed classifier, Yale B dataset which consists of images obtained under diverse state of illumination environment is applied. We show that the results of the proposed model have much more superb performance and robust characteristic than those reported in the previous studies.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 2005년도 ICCAS
• /
• pp.2373-2378
• /
• 2005

An emotion detection algorithm using frontal facial image is presented in this paper. The algorithm is composed of three main stages: image processing stage and facial feature extraction stage, and emotion detection stage. In image processing stage, the face region and facial component is extracted by using fuzzy color filter, virtual face model, and histogram analysis method. The features for emotion detection are extracted from facial component in facial feature extraction stage. In emotion detection stage, the fuzzy classifier is adopted to recognize emotion from extracted features. It is shown by experiment results that the proposed algorithm can detect emotion well.

• 전자공학회논문지CI
• /
• 제49권2호
• /
• pp.96-102
• /
• 2012

영상 분류를 위한 다단계 특성벡터 기반의 분류기 모델(Partitioned Feature-based Classification Model with Expertise Table: PFC-ET)의 성능을 더욱 향상시킨 진보된 형태의 분류기 통합모델 (Classifier Integration Model: CIM)이 본 논문에서 제안되었다. CIM은 PFC-ET과 같이 주어진 데이터에서 추출된 전체의 특징벡터를 연결하여 이용하지 않고, 같은 성질의 특징 벡터들끼리 모아서, 각각의 국지적 학습기를 통하여 분류에 이용한다. PFC-ET에서 분류판단 확률행렬에 의한 오류를 최소화하기위해 국지적 분류기로 사용되는 군집화 알고리즘의 멤버 비율을 사용하여 최종적인 분류의 정확도를 높이는 방안을 제안한다. 제안된 CIM의 성능을 검증하기 위하여, Caltech 데이터에 대한 일반적인 영상 분류와 6 클래스 위성 영상 분류 문제에 대한 실험을 진행하였다. 제안된 CIM은 기존의 PFC 와 PFC-ET 모델과 비교한 실험에서 분류 정확도와 후처리 문제의 복잡성 면에서 향상된 성능을 보여주었다.

• 전기학회논문지
• /
• 제63권6호
• /
• pp.797-803
• /
• 2014

In this paper, 3D face recognition system is designed by using polynomial based on RBFNNs. In case of 2D face recognition, the recognition performance reduced by the external environmental factors such as illumination and facial pose. In order to compensate for these shortcomings of 2D face recognition, 3D face recognition. In the preprocessing part, according to the change of each position angle the obtained 3D face image shapes are changed into front image shapes through pose compensation. the depth data of face image shape by using Multiple Point Signature is extracted. Overall face depth information is obtained by using two or more reference points. The direct use of the extracted data an high-dimensional data leads to the deterioration of learning speed as well as recognition performance. We exploit principle component analysis(PCA) algorithm to conduct the dimension reduction of high-dimensional data. Parameter optimization is carried out with the aid of PSO for effective training and recognition. The proposed pattern classifier is experimented with and evaluated by using dataset obtained in IC & CI Lab.

• 한국지능시스템학회논문지
• /
• 제9권5호
• /
• pp.472-479
• /
• 1999

본 논문에서는 뉴로-퍼지 알고리즘을 이용한 원격탐사 화상의 지표면 패턴분류기를 제안한다. 제안된 패턴 분류기는 일반적인 퍼지 인식기를 가지고 있는 3층 전방향 신경회로망 구조로 되어 있고 가중치들은 퍼지집합으로 구성된다. 이러한 퍼지-뉴로 패턴분류 시스템을 Visual C++ 환경을 구현한다. 성능평가를 위해 기존의 역전파 학습기능을 가진 신경회로망과 Maximum-likelihood 알고리즘을 이용해처리한 결과와비교분석한다. 대표적인 지표면 특징을 나타내는 8개의 클래스에 대해 훈련집합을 선정하고 각각의 분류 알고리즘에 같은 훈련집합을 사용하여 학습시킨 후 실험화상을 적용하여 지표면 특징을 8개의 클래스로 분류하였다. 실험결과 제안된 뉴로-퍼지 분류기는 여러개의 클래스로 혼합된 패턴에 대해서 기존의 분류기들에 비해 보다 더 좋은 성능을 보인다.

• ETRI Journal
• /
• 제21권4호
• /
• pp.41-51
• /
• 1999

This research is an attempt to obtain more accurate land cover information from LANDSAT images. Canonical correlation analysis, which has not been widely used in the image classification community, was applied to the classification of a LANDSAT images. It was found that it is easy to select training areas on the classification using canonical correlation analysis in comparison with the maximum likelihood classifier of $ERDAS^{(R)}$ software. In other words, the selected positions of training areas hardly affect the classification results using canonical correlation analysis. when the same training areas are used, the mapping accuracy of the canonical correlation classification results compared with the ground truth data is not lower than that of the maximum likelihood classifier. The kappa analysis for the canonical correlation classifier and the maximum likelihood classifier showed that the two methods are alike in classification accuracy. However, the canonical correlation classifier has better points than the maximum likelihood classifier in classification characteristics. Therefore, the classification using canonical correlation analysis applied in this research is effective for the extraction of land cover information from LANDSAT images and will be able to be put to practical use.