• 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.

• Proceedings of the Korean Society of Crop Science Conference
• /
• 2017.06a
• /
• pp.301-301
• /
• 2017

Subspecific difference of the percentage of white immature kernels (WIK) between japonica and indica rice cultivars was analyzed in relation to ripening temperature and yield characters. Thirty-three Chinese and 10 Japanese rice cultivars, including 32 japonica and 11 indica, were cultivated with three different cropping seasons for three years. The results were as follows: (1) Indica had less number of panicles, larger number of spikelets per panicle with higher yield, and longer and narrower kernels than japonica. In japonica, Chinese cultivars had less number of panicles and larger number of spikelets per panicle than Japanese cultivars. In addition, WIK was significantly higher in Chinese cultivars than in Japanese cultivars, because of the higher percentage of milky white kernels, even at similar temperature conditions during ripening. On the other hand, WIK in indica was not significantly different between the production areas and between the cropping seasons. (2) Regardless of subspecies, WIK in a large number of Chinese cultivars increased with increasing temperature during ripening within 20 days after heading, while this relation was uncommon in Japanese cultivars, showing the low temperature response. However, some Chinese cultivars had the low WIK with the low temperature response. (3) WIK in japonicawas positively correlated with 1000-kernel weight, spikelet density, kernel width and thickness, but negatively correlated with panicle length and grain filling percentage, while in indica it was positively correlated with panicle number per area, grain filling percentage, brown rice yield and kernel width, but negatively correlated with kernel length. These results indicated that WIK in both subspecies had a close relation to kernel size, and that WIK was high in japonica cultivars with wide and thick kernels and in indica cultivars with short and wide kernels.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.27 no.3
• /
• pp.326-333
• /
• 2003

In the measurement using DMA and CPC in series, there is some time delay for particles classified in DMA to detect in CPC. During this time, the DMA time-response changes due to the velocity profile of sampling tube and the diffusion of particles in the volume that exists between the DMA exit and the detector of ultra-fine CPC. This is called mixing effect. In the accelerated measurement methods like the TSI -SMPS, the size distribution is obtained from the correlation between the time-varying electrical potential of the DMA and the corresponding particle concentrations sampled in DMA. If the DMA time -response changes during this delay time, this can cause the error of a size distribution measured by this accelerated technique. The kernel function considering this mixing effect using the residence time distribution is proposed by Russell et al. In this study, we obtained a size distribution using this kernel to compare to the result obtained by the commercial accelerated measurement system, TSI -SMPS for verification and considered the errors that result from the mixing effect with the geometric mean diameters of originally sampled particles, using virtually calculated responses obtained with this kernel as input data.

• ETRI Journal
• /
• v.36 no.3
• /
• pp.333-342
• /
• 2014

Most hyper-ellipsoidal clustering (HEC) approaches use the Mahalanobis distance as a distance metric. It has been proven that HEC, under this condition, cannot be realized since the cost function of partitional clustering is a constant. We demonstrate that HEC with a modified Gaussian kernel metric can be interpreted as a problem of finding condensed ellipsoidal clusters (with respect to the volumes and densities of the clusters) and propose a practical HEC algorithm that is able to efficiently handle clusters that are ellipsoidal in shape and that are of different size and density. We then try to refine the HEC algorithm by utilizing ellipsoids defined on the kernel feature space to deal with more complex-shaped clusters. The proposed methods lead to a significant improvement in the clustering results over K-means algorithm, fuzzy C-means algorithm, GMM-EM algorithm, and HEC algorithm based on minimum-volume ellipsoids using Mahalanobis distance.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.31 no.2 s.257
• /
• pp.260-268
• /
• 2007

Demand for the use of 3D CAD DMU systems over the Internet environment has been increased. However, transmission of commercial 3D kernels has delayed the communication effectiveness due to the kernel size. Light weight CAD geometric kernel design methodology is required for rapid transmission in the distributed environment. In this paper, an assembly data structure suitable for the top-down and bottom-up assembly models has been constructed. Part features are stored without a hierarchy so that they are created and saved in no particular order. In particular, this paper proposes a new assembly representation model, called multi-level assembly representation (MAR), for the PDM based assembly DMU system. Since the geometric kernel retains assembly hierarchy and topological information, it is applied to the web-viewer for the PDM based DMU system. Effectiveness of the proposed geometric kernel is confirmed through various case studies.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.8
• /
• pp.4043-4060
• /
• 2017

Network Intrusion Detection (NID), an important topic in the field of information security, can be viewed as a pattern recognition problem. The existing pattern recognition methods can achieve a good performance when the number of training samples is large enough. However, modern network attacks are diverse and constantly updated, and the training samples have much smaller size. Furthermore, to improve the learning ability of SVM, the research of kernel functions mainly focus on the selection, construction and improvement of kernel functions. Nonetheless, in practice, there are no theories to solve the problem of the construction of kernel functions perfectly. In this paper, we effectively integrate the advantages of the radial basis function kernel and the polynomial kernel on the notion of the game theory and propose a novel kernel SVM algorithm with game theory for NID, called GTNID-SVM. The basic idea is to exploit the game theory in NID to get a SVM classifier with better learning ability and generalization performance. To the best of our knowledge, GTNID-SVM is the first algorithm that studies ensemble kernel function with game theory in NID. We conduct empirical studies on the DARPA dataset, and the results demonstrate that the proposed approach is feasible and more effective.

• Applied Biological Chemistry
• /
• v.31 no.1
• /
• pp.13-20
• /
• 1988

The hydration characteristics of four naked baley varieties by different kernel sizes were investigated. The predominant kernel size was 7 mesh, followed by 10 mesh kernel, of which comprised $86{\sim}94%$ of the total kernel. Diffusion coefficients of naked barleys at $40^{\circ}C$ increased as the kernel sizes decreased. The volume increases of naked barleys were linearly related to the moisture gain, regardless variety and kernel size.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.29-37
• /
• 2021

In this paper, we propose a deep learning-based retinal vessel segmentation model for handling multi-scale information of fundus images. we integrate the selective kernel convolution into U-Net-based convolutional neural network. The proposed model extracts and segment features information with various shapes and sizes of retinal blood vessels, which is important information for diagnosing eye-related diseases from fundus images. The proposed model consists of standard convolutions and selective kernel convolutions. While the standard convolutional layer extracts information through the same size kernel size, The selective kernel convolution extracts information from branches with various kernel sizes and combines them by adaptively adjusting them through split-attention. To evaluate the performance of the proposed model, we used the DRIVE and CHASE DB1 datasets and the proposed model showed F1 score of 82.91% and 81.71% on both datasets respectively, confirming that the proposed model is effective in segmenting retinal blood vessels.

• Journal of the Korean Statistical Society
• /
• v.31 no.2
• /
• pp.261-276
• /
• 2002

Probability density or its derivatives may have a discontinuity/change point at an unknown location. We propose a method of estimating the location and the jump size of the discontinuity point based on kernel type density or density derivatives estimators with one-sided equivalent kernels. The rates of convergence of the proposed estimators are derived, and the finite-sample performances of the methods are illustrated by simulated examples.

• East Asian mathematical journal
• /
• v.33 no.1
• /
• pp.37-44
• /
• 2017

We will prove size estimates of the Bergman kernel for the generalized Fock space ${\mathcal{F}}^2_{\varphi}$, where ${\varphi}$ belongs to the class $\mathcal{W} $. The main tool for the proof is to use the estimate on the canonical solution to the ${\bar{\partial}}$-equation. We use Delin's weighted $L^2$-estimate ([3], [6]) for it.