• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.959-971
• /
• 2019

CNN (Convolutional Neural Network) is one representative deep learning algorithm, which can extract high-level spatial and spectral features, and has been applied for hyperspectral image classification. However, one significant drawback behind the application of CNNs in hyperspectral images is the high dimensionality of the data, which increases the training time and processing complexity. To address this problem, several CNN based hyperspectral image classification studies have exploited PCA (Principal Component Analysis) for dimensionality reduction. One limitation to this is that the spectral information of the original image can be lost through PCA. Although it is clear that the use of PCA affects the accuracy and the CNN training time, the impact of PCA for CNN based hyperspectral image classification has been understudied. The purpose of this study is to analyze the quantitative effect of PCA in CNN for hyperspectral image classification. The hyperspectral images were first transformed through PCA and applied into the CNN model by varying the size of the reduced dimensionality. In addition, 2D-CNN and 3D-CNN frameworks were applied to analyze the sensitivity of the PCA with respect to the convolution kernel in the model. Experimental results were evaluated based on classification accuracy, learning time, variance ratio, and training process. The size of the reduced dimensionality was the most efficient when the explained variance ratio recorded 99.7%~99.8%. Since the 3D kernel had higher classification accuracy in the original-CNN than the PCA-CNN in comparison to the 2D-CNN, the results revealed that the dimensionality reduction was relatively less effective in 3D kernel.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.9 no.3
• /
• pp.67-72
• /
• 2010

This paper presents biometrical driver verification system in car experiment through analysis of speech, and face information. We have used Mel-scale Frequency Cesptral Coefficients (MFCCs) for speaker verification using speech information. For face verification, face region is detected by AdaBoost algorithm and dimension-reduced feature vector is extracted by using principal component analysis only from face region. In this paper, we apply the extracted speech- and face feature vectors to an SVM kernel with Gaussian Mixture Models(GMM) supervector. The experimental results of the proposed approach show a clear improvement compared to a simple GMM or SVM approach.

• KOREAN JOURNAL OF CROP SCIENCE
• /
• v.59 no.3
• /
• pp.312-318
• /
• 2014

We evaluated the morphological characteristics of 156 maize inbred lines, which were developed to breeding normal maize variety at Maize Experiment Station, Gangwon Agricultural Research and Extension Services, by examining 11 quantitative and three qualitative characteristics. On the evaluation of three qualitative traits for 156 maize inbred lines, most inbred lines showed yellow (85 and 84 inbred lines) at tassel color (QL1) and silk color (QL2), and showed semi erect (105 inbred lines) at plant type (QL3). While, the evaluation of 11 quantitative traits for 156 maize inbred lines, they showed the morphological variation in days of tasseling (QN1, 56.5 to 76.0 days), days of silking (QN2, 59.0 to 85.5 days), stem thickness (QN3, 12.7 to 42.9 mm), plant height (QN4, 111.8 to 239.8 cm), ear height (QN5, 48.2 to 126.5 cm), 100 kernel weight (QN6, 14.9 to 36.4 g), ear length (QN7, 10.0 to 79.0 cm), setted kernel length (QN8, 8.0 to 70.5 cm), ear thickness (QN9, 4.0 to 22.0 cm), total kernel weight (QN10, 22.0 to 490.0 kg) and water content (QN11, 9.3 to 11.9%), respectively. As a result, 11 inbred lines (00hf3, 00hf19, 00hf30, 00hf36, 02S8069, 02S8072, 02S8090, 02S8099, 05S10011, 06S8085-6, 07S8011) in the 156 normal maize inbred lines have showed comparatively high values. While, the results of PCA (principal component analysis) indicated that the ear length (QN7), setted kernel length (QN8), ear thickness (QN9) and total kernel weight (QN10) greatly contributed in positive direction on the first principal components. And also, days of tasseling (QN1), days of silking (QN2), plant height (QN4) and ear height (QN5) contributed in negative direction on the second principal component. Thus these morphological characters, which were greatly contributed in the first and second principal components, might be considered to be useful for discrimination among 156 normal inbred lines. Specifically, this study's assessment of morphological characteristics of 156 normal inbred lines will be helpful useful for normal maize breeding programs such activities as planning crosses for hybrid and line development at Maize Experiment Station, Gangwon Agricultural Research and Extension Services.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.8
• /
• pp.3779-3784
• /
• 2012

Real-time process monitoring and diagnosis of industrial processes is one of important operational tasks for quality and safety reasons. The objective of fault diagnosis or identification is to find process variables responsible for causing a specific fault in the process. This helps process operators to investigate root causes more effectively. This work assesses the applicability of combining a nonlinear statistical technique of kernel Fisher discriminant analysis with a preprocessing method as a tool of on-line fault identification. To compare its performance to existing linear principal component analysis (PCA) identification scheme, a case study on a benchmark process was performed to show that the fault identification scheme produced more reliable diagnosis results than linear method.

• Korean Journal of Agricultural Science
• /
• v.9 no.1
• /
• pp.396-450
• /
• 1982

To obtain basic information on the Korean local corn lines a total of 57 lines were selected from 1,000 Korean local collection at Chungnam National University, classified by principal component analysis, and genetic nature was investigated. The results are summarized as follows. 1. There were a great variation in mean values of plant characters of the lines. The mean values of plant characters except for density of kernels varied with types of crossing. All characters except. for tasselling dates were reduced in magnitude when selfed, while those characters were increased when topcrossed. 2. The correlation coefficients among characters studied ranged front 0.99 to -0.59. The correlation coefficients among characters were not greatly changed depending upon types of crosses. 3. In order to classify the lines more effectively, selected 12 plant characters were used to classify 57 local lines by principal component analysis. The first four component could explain 86.4%, 83.4% and 81.1% of the total variations in sibbed lines, selfed lines and topcrossed lines, respectively. 4. Contribution of characters to principal component was high at upper principal components and low at lower principal components. 5. Biological meaning of the principal component and plant types corresponding to the each principal component were explained clearly by the correlation coefficient between principal components and characters. The first principal component appeared to correspond to the size of plant and ear. The second principal component appeared to correspond to the degree of differentiation in organs and the duration of vegetative growing period. But biological meaning of the third and fourth principal components was not clear. 6. The lines were classified into 4 lineal groups by the taxonomic distance. Group I included 52 lines which was 91.2% of total lines, group II 3 lines, group III 1 lines and group IV I lines, respectively. Four groups could be characterized as follows : Group I : early maturity, short-culmed, medium height plant, small ears, medium kernels and medium yielding. Group II : late maturity, medium height plant, small ears, small kernels, prolific ears and higher yielding. Group III : medium maturity, tall-culmed, small ears, small kernels and low yielding. Group IV : medium maturity, tall-calmed, large ears, one ear plant and me yielding. 7. The inbreeding depression varied with plant characters and lines. The characters such as yield, kernel weight per ear, ear weight and plant height showed great degree of inbreeding depression. Group I showed high inbreeding depression in such characters as 100 kernel weight, leaf number, plant height and days to tasselling, while group II showed high inbreeding depression in other plant characters. 8. Heterosis of plant characters varied also with lines. The ear weight, kernel weight per ear, yield, 100 kernel weight, and plant height were some of the plant characters showing high heterosis. Group II showed high values of heterosis in such characters as ear length, ear diameter, ear weight, kernel weight per ear, 100 kernel weight, and leaf length, while group I was high in heterosis in other plant characters. 9. The degree of homozgosity was highest in ear weight (79.1%) and lowest in ear number per plant (-21%). Group II showed higher degree of homozygosity than group I. 10. Correlation coefficients between characters of ribbed and topcrossed lines were positive for all characters. Highly significant. correlation coefficients between ribbed and topcrossed lines were obtained especially for characters such as ear number per plant, plant height, leaf length and yield per plot.

• Journal of Electrical Engineering and Technology
• /
• v.12 no.4
• /
• pp.1406-1416
• /
• 2017

In complex and large-scale industries, properly designed fault detection and identification (FDI) systems considerably improve safety, reliability and availability of target processes. In thermal power plants (TPPs), generating units operate under very dangerous conditions; system failures can cause severe loss of life and property. In this paper, we propose a bagged auto-associative kernel regression (AAKR)-based FDI approach for steam boilers in TPPs. AAKR estimates new query vectors by online local modeling, and is suitable for TPPs operating under various load levels. By combining the bagging method, more stable and reliable estimations can be achieved, since the effects of random fluctuations decrease because of ensemble averaging. To validate performance, the proposed method and comparison methods (i.e., a clustering-based method and principal component analysis) are applied to failure data due to water wall tube leakage gathered from a 250 MW coal-fired TPP. Experimental results show that the proposed method fulfills reasonable false alarm rates and, at the same time, achieves better fault detection performance than the comparison methods. After performing fault detection, contribution analysis is carried out to identify fault variables; this helps operators to confirm the types of faults and efficiently take preventive actions.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.9 no.3
• /
• pp.1173-1192
• /
• 2015

The Support Vector Data Description (SVDD) has achieved great success in anomaly detection, directly finding the optimal ball with a minimal radius and center, which contains most of the target data. The SVDD has some limited classification capability, because the hyper-sphere, even in feature space, can express only a limited region of the target class. This paper presents an anomaly detection algorithm for mitigating the limitations of the conventional SVDD by finding the minimum volume enclosing ellipsoid in the feature space. To evaluate the performance of the proposed approach, we tested it with intrusion detection applications. Experimental results show the prominence of the proposed approach for anomaly detection compared with the standard SVDD.

• The Korean Journal of Applied Statistics
• /
• v.36 no.2
• /
• pp.167-173
• /
• 2023

This paper investigates several methods of visualizing high-dimensional data in a low-dimensional space. At first, principal component analysis and multidimensional scaling are briefly introduced as linear approaches, and then kernel principal component analysis, self-organizing map, locally linear embedding, Isomap, Laplacian Eigenmaps, and local multidimensional scaling are introduced as nonlinear approaches. In particular, t-SNE, which is widely used but relatively unfamiliar in the field of statistics, is described in more detail. We also present a simple example for several methods, including t-SNE. Finally, we provide a review of several recent studies pointing out the limitations of t-SNE and discuss the future research problems presented.

• Korean Chemical Engineering Research
• /
• v.46 no.2
• /
• pp.233-247
• /
• 2008

Process monitoring technology is able to detect the faults and the process changes which occur in a process unpredictably, which makes it possible to find the reasons of the faults and get rid of them, resulting in a stable process operation, high-quality product. Statistical process monitoring method based on data set has a main merit to be a tool which can easily supervise a process with the statistics and can be used in the analysis of process data if a high quality of data is given. Because a real process has the inherent characteristics of nonlinearity, non-Gaussianity, multiple operation modes, sensor faults and process changes, however, the conventional multivariate statistical process monitoring method results in inefficient results, the degradation of the supervision performances, or often unreliable monitoring results. Because the conventional methods are not easy to properly supervise the process due to their disadvantages, several advanced monitoring methods are developed recently. This review introduces the theories and application results of several remarkable monitoring methods, which are a nonlinear monitoring with kernel principle component analysis (KPCA), an adaptive model for process change, a mixture model for multiple operation modes and a sensor fault detection and reconstruction, in order to tackle the weak points of the conventional methods.

• Structural Monitoring and Maintenance
• /
• v.9 no.2
• /
• pp.179-200
• /
• 2022

This paper proposes a data-driven methodology for online early damage identification under changing environmental conditions. The proposed method relies on two data analysis methods: feature-based method and hybrid principal component analysis (PCA) and kernel PCA to separate damage from environmental influences. First, spectral sub-band features, namely, spectral sub-band centroids (SSCs) and log spectral sub-band energies (LSSEs), are proposed as damage-sensitive features to extract damage information from measured structural responses. Second, hybrid modeling by integrating PCA and kernel PCA is performed on the spectral sub-band feature matrix for data normalization to extract both linear and nonlinear features for nonlinear procedure monitoring. After feature normalization, suppressing environmental effects, the control charts (Hotelling T² and SPE statistics) is implemented to novelty detection and distinguish damage in structures. The hybrid PCA-KPCA technique is compared to KPCA by applying support vector machine (SVM) to evaluate the effectiveness of its performance in detecting damage. The proposed method is verified through numerical and full-scale studies (a Bridge Health Monitoring (BHM) Benchmark Problem and a cable-stayed bridge in China). The results demonstrate that the proposed method can detect the structural damage accurately and reduce false alarms by suppressing the effects and interference of environmental variations.