• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.7C
• /
• pp.428-434
• /
• 2011

Recently, subspace analysis has raised its performance to a higher level through the adoption of kernel-based nonlinearity. Especially, the radial basis function, based on its nonparametric nature, has shown promising results in face recognition. However, due to the endemic small sample size problem of face data, the conventional kernel-based feature extraction methods have difficulty in data representation. In this paper, we introduce a novel variant of the RBF kernel to alleviate this problem. By adopting the concept of the nearest feature line classifier, we show both effectiveness and generalizability of the proposed method, particularly regarding the small sample size issue.

• Korean Journal of Agricultural Science
• /
• v.13 no.1
• /
• pp.1-16
• /
• 1986

This experiment was carried out to obtain genetic information for future corn breeding. The materials used for the study were obtained from the nationwide collection of Korean local corn lines. A total of 262 lines were used for the study of morphological characters and for the classification of lines. Results obtained are summarized as follows; 1. The days to flowering of lines ranged from 57 days to 87 days. Most lines had an average of 67 days of flowering days. 2. The number of tillers of lines showed a lot of variation among lines with 49.2% of coefficient of variation. 3. The coefficients of variation computed based on the phenotypic observation or measurement of each line were 36.1%, 27.2%, 20.0%, 16.4% and 16.3% for kernel weight per ear, 100 kernel weight, ear height, plant height and ear length, respectively. 4. Ear height, ear length, ear diameter, tiller number and days to flowering were highly and positively correlation with the plant height. Kernel size, ear size, and plant height were highly correlated with 100 kernel weight and kernel weight per ear. 5. The 262 corn lines were possibly classified into four major groups by the Euclidean distance. Group I comprised 110 lines, group II 74 lines, group III 66 lines and group IV 12 lines, respectively. Group I was characterized as having early maturity, medium plant height large kernel size and large ear size. Group II had medium maturity, short plant height, medium kernel size and small ear size. Group III had medium maturity, medium plant height, large kernel size and medium ear size. Group IV had late maturity, long plant height, small kernel size, small ear size and many tillering. 6. The plant height showed significant difference between group I and II, II and III, and II and IV group. No statistical differences were observed between group III and IV. The ear size of group I was significantly different from those of group II, III and IV. Also difference of ear size between group II and III was significant. The kernel size, 100 kernel weight and kernel weight per ear were all significantly different among all groups classified. The row number was different between group I and II. The row number of lines in group IV was significantly different with group I, II, III respectively. The number of tillers and flowering days of lines in group IV were greatly different from those of group I, II and III. 7. The corn lines collected from northwest plain regions and middle hilly regions in Korea had medium maturity, medium plant height, large ear and large kernels. The corn lines from middle eastern hilly regions had medium size of ear kernels. The corn lines from middle southern hilly regions had late maturity, small kernel size and many tillers. The corn lines from southwest plain areas had late maturity, long plant height and many tillers.

• Journal of the Institute of Convergence Signal Processing
• /
• v.13 no.3
• /
• pp.130-135
• /
• 2012

In this paper, we propose a technique to detect the size and location of the small target in images by using Gaussian kernel repeatedly. In order to detect the size and location of the small target, we find the local maximum value by applying the facet model and then use the $3{\times}3$ Gaussian kernel repeatedly. we determine the size of small target by comparing the local maximum value $D_2$ according to the number of iteration. To reduce the computational complexity, we use the Gaussian pyramid when using the kernel repeatedly. Through the experiment, we verified that the size and location of the small target is detected by the number of iterations and results show improvements from conventional methods.

• Korean Journal of Remote Sensing
• /
• v.32 no.1
• /
• pp.13-24
• /
• 2016

This paper presents a method to evaluate the performance of subpixel localization operators using target image data. Subpixel localization of edges is important to extract the precise shape of objects from images. In this study, each target image was designed to provide reference lines and edges to which the localization operators can be applied. We selected two types of moment-based operators: Gray-level Moment (GM) operator and Spatial Moment (SM) operator for comparison. The original edge localization operators with kernel size 5 are tested and their extended versions with kernel size 7 are also tested. Target images were collected with varying Camera-to-Object Distance (COD). From the target images, reference lines are estimated and edge profiles along the estimated reference lines are accumulated. Then, evaluation of the performance of edge localization operators was performed by comparing the locations calculated by each operator and by superimposing them on edge profiles. Also, enhancement of edge localization by increasing the kernel size was also quantified. The experimental result shows that the SM operator whose kernel size is 7 provides higher accuracy than other operators implemented in this study.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.1 no.1
• /
• pp.59-65
• /
• 1993

In a spark ignition engine, in order to make research on flame propagation, attentive concentration should be paid on initial combustion stage about the formation and development of flame. In addition, the initial stage of combustion governs overall combustion period in a spark ignition engine. With the increase of the size of flame kernel, it could reach initial flame stage easily, and the mixture could proceed to the combustion of stabilized state. Therefore, we must study the theoretical calculation of minimum flame kernel radius which effects on the formation and development of kernel. To calculate the minimum flame kernel radius, we must know the thermal conductivity, flame temperature, laminar burning velocity and etc. The thermal conductivity is derived from the molecular kinetic theory, the flame temperature from the chemical reaction equations and the laminar burning velocity from the D.K.Kuehl's formula. In order to estimate the correctness of the theoretically calculated minimum flame kernel radius, the researcheres compared it with the RMaly's experimental values.

• Food Science and Biotechnology
• /
• v.16 no.6
• /
• pp.918-923
• /
• 2007

To investigate the relationship of wheat single kernel characteristics with end-use properties, 183 soft wheat cultivars and lines were evaluated for milling quality characteristics (kernel hardness, kernel and flour protein, flour ash), and end-use properties (i.e., as ingredients in sugar-snap cookies, sponge cake). Significant positive correlations occurred among wheat hardness parameters including near-infrared reflectance (NIR) score and single kernel characterization system (SKCS). The SKCS characteristics were also significantly correlated with conventional wheat quality parameters such as kernel size, wheat protein content, and straight-grade flour yield. The cookie diameter and cake volume were negatively correlated with NIR and SKCS hardness, and there was an inverse relationship between flour protein contents and kernel weights or sizes. Sugar-snap cookie diameter was positively correlated with sponge cake volume.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.13 no.4
• /
• pp.254-268
• /
• 2013

For real-world clustering tasks, the input data is typically not easily separable due to the highly complex data structure or when clusters vary in size, density and shape. Kernel-based clustering has proven to be an effective approach to partition such data. In this paper, we provide an overview of several fuzzy kernel clustering algorithms. We focus on methods that optimize an fuzzy C-mean-type objective function. We highlight the advantages and disadvantages of each method. In addition to the completely unsupervised algorithms, we also provide an overview of some semi-supervised fuzzy kernel clustering algorithms. These algorithms use partial supervision information to guide the optimization process and avoid local minima. We also provide an overview of the different approaches that have been used to extend kernel clustering to handle very large data sets.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.36 no.1A
• /
• pp.35-40
• /
• 2011

The complex-valued blind algorithm based on a set of randomly generated symbols and Euclidean distance can take advantage of information theoretic learning and cope with the channel phase rotation problems. On the algorithm, in this paper, the effect of kernel size has been studied and a kernel-modified version of the algorithm that rearranges the forces between the information potentials by kernel-modification has been proposed. In simulation results for 16 QAM and complex-channel models, the proposed algorithm show significantly enhanced performance of symbol-point concentration and no phase rotation problems caused by the complex channel models.

• Journal of Mechanical Science and Technology
• /
• v.18 no.5
• /
• pp.838-846
• /
• 2004

Minimum ignition energies of hydrogen/air and methane/air mixtures have been investigated numerically by solving unsteady one-dimensional conservation equations with detailed chemical kinetic mechanisms. Initial kernel size needed for numerical calculation is a sensitive function of initial pressure of a mixture and should be estimated properly to obtain quantitative agreement with experimental results. A simple macroscopic model to determine minimum ignition energy has been proposed, where the initial kernel size is correlated with the quenching distance of a mixture and evaluated from the quenching distance determined from experiment. The simulation predicts minimum ignition energies of two sample mixtures successfully which are in a good agreement with the experimental data for the ranges of pressure and equivalence ratio.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.13 no.6
• /
• pp.1257-1262
• /
• 2018

In this paper, a deep learning method is proposed to detect the fire effectively by using video of surveillance camera. Based on AlexNet model, classification performance is compared according to kernel size and stride of convolution layer. Dataset for learning and interfering are classified into two classes such as normal and fire. Normal images include clouds, and foggy images, and fire images include smoke and flames images, respectively. As results of simulations, it is shown that the larger kernel size and smaller stride shows better performance.