• The KIPS Transactions:PartB
• /
• v.15B no.6
• /
• pp.533-542
• /
• 2008

Background subtraction from video data is one of the most important task in various realtime machine vision applications. In this paper, a new scheme for background subtraction based on the hexagonal pixel sampling is proposed. Generally it has been found that hexagonal spatial sampling yields smaller quantization errors and remarkably improves the understanding of connectivity. We try to apply the hexagonally sampled image to the LBP based non-parametric background subtraction algorithm. Our scheme makes it possible to omit the bilinear pixel interpolation step during the local binary pattern generation process, and, consequently, can reduce the computation time. Experimental results revealed that our approach based on hexagonal spatial sampling is very efficient and can be utilized in various background subtraction applications.

• Journal of Korean Institute of Industrial Engineers
• /
• v.40 no.4
• /
• pp.404-414
• /
• 2014

Yield prediction is important to manage semiconductor quality. Many researches with machine learning algorithms such as SVM (support vector machine) are conducted to predict yield precisely. However, yield prediction using SVM is hard because extremely imbalanced and big data are generated by final test procedure in semiconductor manufacturing process. Using SVM algorithm with imbalanced data sometimes cause unnecessary support vectors from major class because of unselected support vectors from minor class. So, decision boundary at target class can be overwhelmed by effect of observations in major class. For this reason, we propose a under-sampling method with minor class based SVM (MCSVM) which overcomes the limitations of ordinary SVM algorithm. MCSVM constructs the model that fixes some of data from minor class as support vectors, and they can be good samples representing the nature of target class. Several experimental studies with using the data sets from UCI and real manufacturing process represent that our proposed method performs better than existing sampling methods.

• Journal of the Institute of Electronics and Information Engineers
• /
• v.54 no.4
• /
• pp.68-74
• /
• 2017

This paper presents a joint demosaicking and arbitrary-ratio down sampling algorithm for color filter array (CFA) images. Color demosaiking is a necessary part of image signal processing pipeline for many types of digital image recording system using single sensor. Also, such as smart phone, obtained high resolution image from image sensor has to be down-sampled to be displayed on the screen. The conventional solution is "Demosaicking first and down sampling later". However, this scheme requires a significant amount of memory and computational cost. Also, artifacts can be introduced or details get damaged during demosaicking and down sampling process. In this paper, we propose a method in which demosaicking and down sampling are working simultaneously. We use inverse mapping of Bayer CFA and then joint demosaicking and down sampling with arbitrary-ratio scheme based on signal decomposition of high and low frequency component in input data. Experimental results show that our proposed algorithm has better image quality performance and much less computational cost than those of conventional solution.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.51 no.10
• /
• pp.495-501
• /
• 2002

A difference based current transformer (CT) saturation detection algorithm uses the third difference of a secondary current to detect the instants of the beginning/end of saturation. The third difference of a secondary current contains high frequency components when a CT is saturated. Thus, an effect of an anti-aliasing lowpass filter implemented in digital protection relays on the detection algorithm should be studied. This paper describes performance analysis of a lowpass filter on the CT saturation detection algorithm. The cutoff frequency of the lowpass filter is normally set to be half of a sampling frequency. In this Paper, two sampling frequencies of 3,840 (Hz) corresponding to 64 sample/cycle (s/c) and 1,920 (Hz) corresponding to 32 (s/c) are studied; the cutoff frequencies of the lowpass filters are set to be 1,920 (Hz), 960 (Hz) and 960(Hz), 480(Hz), respectively. And the proposed algorithm is verified by experiment. A 2nd order Butterworth filter is designed as a lowpass filter. The test results and experiment results clearly indicate that the saturation detection algorithm successfully detects the instants of the beginning/end of saturation even though a secondary current is filtered by the designed lowpass filters.

• Journal of Biomedical Engineering Research
• /
• v.17 no.4
• /
• pp.469-478
• /
• 1996

In the conventional digital ultrasound scanner, the reflected signal is sampled either in polar coordinates of R-$\theta$ method, or in Cartesian coordinates of uniform ladder algorithm (ULA). The R-$\theta$ scan method necessitates a coordinate transform process which makes hardware complex in comparison with ULA scan mrthoA In spite of this complexity, R-$\theta$ method has a good resolution in ultrasonographic (US) image, since scan direction of the US imaging is a radial direction. In this paper, a new digital scan converter is proposed, which is named the radius uniform ladder algorithm (RULA). The RULA has the rome scan direction as the US scanning in the radial direction and as the display space in the $\theta$ direction. In tllis new approach, sampled points we uniformly distributed in each horizontal line i.n well as in each radial ray so that the data are displayed in the Cartesian coordinates by the 1-D interpolation process. The propped algorithm has an uniform resolution in the periphery and the center field in comparison with equi-angle ULA and equi-interval ULA. To extend the scan angle, concentric square raster sampling (CSRS) is adopted with reduction of discontinuities on the junctions between horizontal scan and vertical scan. The discontinuities are reduced by using the hmction filtering along the $\theta$ direction.

• Journal of Information Processing Systems
• /
• v.18 no.5
• /
• pp.677-687
• /
• 2022

We propose a detection algorithm based on tree-structured convolutional neural networks (TsCNNs) that finds pornography, propaganda, or other inappropriate content on a social media network. The algorithm sequentially applies the typical convolutional neural network (CNN) algorithm in a tree-like structure to minimize classification errors in similar classes, and thus improves accuracy. We implemented the detection system and conducted experiments on a data set comprised of 6 ordinary classes and 11 inappropriate classes collected from the Korean military social network. Each model of the proposed algorithm was trained, and the performance was then evaluated according to the images and videos identified. Experimental results with 20,005 new images showed that the overall accuracy in image identification achieved a high-performance level of 99.51%, and the effectiveness of the algorithm reduced identification errors by the typical CNN algorithm by 64.87 %. By reducing false alarms in video identification from the domain, the TsCNNs achieved optimal performance of 98.11% when using 10 minutes frame-sampling intervals. This indicates that classification through proper sampling contributes to the reduction of computational burden and false alarms.

• Journal of Korean Society for Quality Management
• /
• v.30 no.1
• /
• pp.74-96
• /
• 2002

The objectives of this study is to suggest the rectifying sampling inspection plan considering quality cost. Limiting quality level(LQL) plans(also called LTPD plans) and outgoing quality(OQ) plans are considered. The Hald's linear cost model is discussed with and without a beta prior for the distribution of the fraction of nonconforming items in a lot. It is assumed that the sampling inspection is error free. We consider the design of reliability acceptance sampling plan (RASP) for failure rate level qualification at selected confidence level. The lifetime distribution of products is assumed to be exponential. MIL-STD-690C and K C 6032 standards provide this procedures. But these procedures have some questions to apply in the field. The cost of test and confidence level(1-$\beta$ risk) are the problem between supplier and user. So, we suggest that the optimal life test sampling inspection plans using simple linear cost model considering product cost, capability of environment chamber, environmental test cost, and etc. Especially, we consider a reliability of lots that contain some nonconforming items. In this case we assumed that a nonconforming item fail after environmental life test. Finally, we develope the algorithm of the optimal sampling inspection plan based on minimum costs for rectifying inspection and RASP. And computer application programs are developed So, it is shown how the desired sampling plan can be easily found.

• Journal of Korean Society for Quality Management
• /
• v.43 no.4
• /
• pp.471-488
• /
• 2015

Purpose: Double sampling $T^2$ chart is a useful tool for detecting a relatively small shift in process mean when the process is controlled by multiple variables. This paper finds the optimal design of the double sampling $T^2$ chart in both economical and statistical sense under Weibull failure model. Methods: The expected cost function is mathematically derived using recursive equation approach. The optimal designs are found using a genetic algorithm for numerical examples and compared to those of single sampling $T^2$ chart. Sensitivity analysis is performed to see the parameter effects. Results: The proposed design outperforms the optimal design of the single sampling $T^2$ chart in terms of the expected cost per unit time and Type-I error rate for all the numerical examples considered. Conclusion: Double sampling $T^2$ chart can be designed to satisfy both economic and statistical requirements under Weibull failure model and the resulting design is better than the single sampling counterpart.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2001.05a
• /
• pp.7-14
• /
• 2001

In this study, we characterized the variation of sampling errors using the Waymire-Gupta-rodriguez-Iturbe multi-dimensional rainfall model (WGR model). The parameters used for this study are those derived by Jung et al. (2000) for the Han River Basin using a genetic algorithm technique. The sampling error problems considering in this study are those far using raingauge network, satellite observation and also for both combined. The characterization of sampling errors was done for each month and also for the downstream plain area and the upstream mountain area, separately. As results of the study we conclude: (1) The pattern of sampling errors estimated are obviously different from the seasonal pattern of mentally rainfall amounts. This result may be understood from the fact that the sampling error is estimated not simply by considering the rainfall amounts, but by considering all the mechanisms controlling the rainfall propagation along with its generation and decay. As the major mechanism of moisture source to the Korean Peninsula is obviously different each month, it seems rather norma1 to provide different pattern of sampling errors from that of monthly rainfall amounts. (2) The sampling errors estimated for the upstream mountain area is about twice higher than those for the down stream plain area. It is believed to be because of the higher variability of rainfall in the upstream mountain area than in the down stream plain area.