• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.11C
• /
• pp.1518-1526
• /
• 2004

In this paper, an algorithm using wavelet transform for detecting a cut that is a radical scene transition point, and fade and dissolve that are gradual scene transition points is proposed. The conventional methods using wavelet transform for this purpose is using features in both spatial and frequency domain. But in the proposed algorithm, the color space of an input image is converted to YUV and then luminance component Y is transformed in frequency domain using 2-level lifting. Then, the histogram of only low frequency subband that may contain some spatial domain features is compared with the previous one. Edges obtained from other higher bands can be divided into global, semi-global and local regions and the histogram of each edge region is compared. The experimental results show the performance improvement of about 17% in recall and 18% in precision and also show a good performance in fade and dissolve detection.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.41 no.9
• /
• pp.1120-1131
• /
• 2016

In this paper, we propose a smoker recognition algorithm, detecting smokers in a video sequence in order to prevent fire accidents. We use description-based method in hierarchical approaches to recognize smoker's activity, the algorithm consists of background subtraction, object detection, event search, event judgement. Background subtraction generates slow-motion and fast-motion foreground image from input image using Gaussian mixture model with two different learning-rate. Then, it extracts object locations in the slow-motion image using chain-rule based contour detection. For each object, face is detected by using Haar-like feature and smoke is detected by reflecting frequency and direction of smoke in fast-motion foreground. Hand movements are detected by motion estimation. The algorithm examines the features in a certain interval and infers that whether the object is a smoker. It robustly can detect a smoker among different objects while achieving real-time performance.

• Journal of KIISE:Databases
• /
• v.29 no.2
• /
• pp.99-109
• /
• 2002

ROLAP(Relational Online Analytical Processing) is a process and methodology for a multidimensional data analysis that is essential to extract desired data and to derive value-added information from an enterprise data warehouse. In order to speed up query processing, most ROLAP systems pre-compute summary tables. This process is called 'cube generation' and it mostly involves intensive table sorting stages. (1) showed that it is much faster to generate ROLAP summary tables indirectly using a MOLAP(multidimensional OLAP) cube generation algorithm. In this paper, we present such an indirect ROLAP cube generation algorithm that is fast and scalable. High memory utilization is achieved by slicing the input fact table along one or more dimensions before generating summary tables. High speed is achieved by producing summary tables from their smallest parents. We showed the efficiency of our algorithm through experiments.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.1 no.3
• /
• pp.97-103
• /
• 2008

In this paper, $256{\times}256$ input image is classified into a validity block and an edge block of $8{\times}8$ block for image compression. DCT(Discrete Cosine Transform) is executed only for the DC coefficient that is validity coefficients for a validity block. Predict the position where a quantization coefficient becomes 0 for an edge block, I propose new algorithm to execute DCT in the reduced region. Not only this algorithm that I proposed reduces computational complexity of FDCT(Forward DCT) and IDCT(Inverse DCT) and decreases encoding time and decoding time. I let compressibility increase by accomplishing other stability verticality zigzag scan by the block size that was classified for each block at the time of huffman encoding each. In addition, the algorithm that I suggested reduces Run-Length by accomplishing the level verticality zigzag scan that is good for a classified block characteristic and, I offer the compressibility that improved thereby.

• Journal of the Korea Society for Simulation
• /
• v.26 no.3
• /
• pp.55-63
• /
• 2017

In R&D of artillery system, error budget method is used to predict artillery firing error without field firing test. The error budget method for artillery has been consistently developed but apply for practical R&D of the weapon system has been avoided because of lacks of error budget source information. The error budget source is composed of every detailed error components which affect total distance and deflection error of artillery, and most of them are difficult to be calculated or measured. Also with the inaccuracy of source information, simulated error result dose not reflect real firing error. To resolve that problem, an optimization algorithm is adopted to figure out error budget sources from existing filed firing test. The method of finding input parameter estimation which is commonly used in aerodynamics was applied. As an optimization algorithm, CMA-ES is used and presented in the paper. The error budget sources which are figured out by the presented method can be applied to compute ROC of new weapon systems and may contribute to an improvement of accuracy in artillery.

• Journal of Information Processing Systems
• /
• v.14 no.6
• /
• pp.1464-1479
• /
• 2018

In face recognition, sometimes the number of available training samples for single category is insufficient. Therefore, the performances of models trained by convolutional neural network are not ideal. The small sample face recognition algorithm based on novel Siamese network is proposed in this paper, which doesn't need rich samples for training. The algorithm designs and realizes a new Siamese network model, SiameseFacel, which uses pairs of face images as inputs and maps them to target space so that the $L_2$ norm distance in target space can represent the semantic distance in input space. The mapping is represented by the neural network in supervised learning. Moreover, a more lightweight Siamese network model, SiameseFace2, is designed to reduce the network parameters without losing accuracy. We also present a new method to generate training data and expand the number of training samples for single category in AR and labeled faces in the wild (LFW) datasets, which improves the recognition accuracy of the models. Four loss functions are adopted to carry out experiments on AR and LFW datasets. The results show that the contrastive loss function combined with new Siamese network model in this paper can effectively improve the accuracy of face recognition.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.6
• /
• pp.675-681
• /
• 2019

Noise removal is a pre-requisite procedure in image processing, and various methods have been studied depending on the type of noise and the environment of the image. However, for image processing with high-frequency components, conventional additive white Gaussian noise (AWGN) removal techniques are rather lacking in performance because of the blurring phenomenon induced thereby. In this paper, we propose an algorithm to minimize the blurring in AWGN removal processes. The proposed algorithm sets the high-frequency and the low-frequency component filters, respectively, depending on the pixel properties in the mask, consequently calculating the output of each filter with the addition or subtraction of the input image to the reference. The final output image is obtained by adding the weighted data calculated using the standard deviations and the Gaussian distribution with the output of the two filters. The proposed algorithm shows improved AWGN removal performance compared to the existing method, which was verified by simulation.

• The Journal of Korean Institute of Information Technology
• /
• v.17 no.9
• /
• pp.19-30
• /
• 2019

Photovoltaic power generation requires MPPT algorithm to track stable and efficient maximum power output power point according to external changes such as solar radiation and temperature. This study implemented a model that could track MPP more quickly than original MPPT algorithm using artificial neural network. The proposed model finds the current and voltage of MPP using the original MPPT algorithm for various combinations of insolation and temperature for training data of artificial neural networks. The acquired MPP data was learned using the input node as insolation and temperature and the output node as the current and voltage. The Experiment results show tracking time of the original algorithms P&O, InC and Fuzzy were respectively 0.428t, 0.49t and 0.4076t for the 0t~0.3t range, and MPP tracking time of the proposed model was 0.32511t and it is 0.1t faster than the original algorithms.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.24 no.6
• /
• pp.722-728
• /
• 2020

Digital image processing has been applied in a wide range of fields due to the development of IoT technology and plays an important role in data processing. Various techniques have been proposed to remove such noise, but the conventional impulse noise canceling methods are insufficient to remove noise of edge components of an image, and have a disadvantage of being greatly affected by random impulse noise. Therefore, in this paper, we propose an algorithm that effectively removes edge component noise in random impulse noise environment. The proposed algorithm calculates the threshold value by determining the noise level and switches the filtering process by comparing the reference value with the input pixel value. The proposed algorithm shows good performance in the existing method, and the simulation results show that the noise is effectively removed from the edge of the image.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.6
• /
• pp.807-812
• /
• 2021

In order to be able to interact with the user to experience it as if it were real in virtual reality contents, input/output devices that make them feel the five senses of humans are required . In virtual reality (VR), devices that stimulate sight and hearing are the most representative. For a more realistic experience, suits and gloves that stimulate the sense of touch have recently been released, but there are not many cases applied to actual contents due to the limitation of device . In this paper, we analyze a virtual reality glove that can detect hand movement and touch in a virtual world. Based on the analysis, we propose an algorithm that can sense the intensity of collision with a VR object by tactile sense by improving the UI/UX using the vibration of the feedback method used in the existing virtual reality glove. In addition, the system implemented by the algorithm is applied to an actual case.