• Proceedings of the Korea Information Processing Society Conference
• /
• 2024.05a
• /
• pp.727-730
• /
• 2024

최근 딥러닝 기술이 급속히 발전하며 현대 사회의 다양한 응용분야에서 빠르게 적용되고 있다. 특히 영상 기반의 딥러닝 기술은 자연어 처리와 함께 인공지능 기술의 핵심 연구 분야로 많은 연구가 진행되고 있다. 논문에서는 최근 많은 연구가 진행되고 있는 영상의 의미적 분할 (Semantic Segmentation) 성능을 향상하기 위한 연구를 진행한다. 특히 모델에서 고정밀의 의미적 분할을 수행할 수 있도록 추가적인 정보로써 의사 깊이맵 (Pseudo Depth-Map)을 활용하는 방법을 제안하였다. 더불어, 의사 깊이맵을 모델 상에서 효과적으로 학습시키기 위하여 다중 디코더 모델과 학습 효율을 높이는 학습 스케줄링 전략을 제안한다. 의사 깊이맵과 다중 디코더 모델 기반의 제안 모델은 기존 의미적 분할 모델과 비교하여 iIoU 기준 2%의 성능 향상을 보였다.

• IEIE Transactions on Smart Processing and Computing
• /
• v.2 no.2
• /
• pp.45-56
• /
• 2013

The precise identification of vehicle positions, known as the vehicle localization problem, is an important requirement for building intelligent vehicle ad-hoc networks (VANETs). To solve this problem, two categories of solutions are proposed: stand-alone and data fusion approaches. Compared to stand-alone approaches, which use single information including the global positioning system (GPS) and sensor-based navigation systems with differential corrections, data fusion approaches analyze the position information of several vehicles from GPS and sensor-based navigation systems, etc. Therefore, data fusion approaches show high accuracy. With the position information on a set of vehicles in the preprocessing stage, data fusion approaches is used to estimate the precise vehicular location in the local map building stage. This paper proposes an efficient local map building scheme, which increases the accuracy of the estimated vehicle positions via V2V communications. Even under the low ratio of vehicles with communication modules on the road, the proposed local map building scheme showed high accuracy when estimating the vehicle positions. From the experimental results based on the parameters of the practical vehicular environments, the accuracy of the proposed localization system approached the single lane-level.

• The KIPS Transactions:PartD
• /
• v.8D no.4
• /
• pp.421-426
• /
• 2001

A formal development method is used to develop software rigorously and systematically. In a formal development method, we specify system by a formal specification language and gradually develop the system more concretely until we can implement the system. VDM is one of formal specification languages. VDM uses mathematical data structures such as sets, sequences, and maps to specify the system, but most programming languages do not have such data structures. Therefore, these data structures should be converted. We can convert mathematical data structures in VDM to a linked list, a data structure in a programming language. In this article, we propose a method to convert a set, a sequence, and a map in VDM to a linked list in a programming language and prove the correctness of this conversion mathematically.

• Proceedings of the Korean Society for Bioinformatics Conference
• /
• 2003.10a
• /
• pp.170-177
• /
• 2003

Gene expression data are the quantitative measurements of expression levels and ratios of numberous genes in different situations based on microarray image analysis results. The process to draw meaningful information related to genomic diseases and various biological activities from gene expression data is known as gene expression data analysis. In this paper, we present a hierarchical clustering method of gene expression data based on self organizing map which can analyze the clustering result of gene expression data more efficiently. Using our proposed method, we could eliminate the uncertainty of cluster boundary which is the inherited disadvantage of self organizing map and use the visualization function of hierarchical clustering. And, we could process massive data using fast processing speed of self organizing map and interpret the clustering result of self organizing map more efficiently and user-friendly. To verify the efficiency of our proposed algorithm, we performed tests with following 3 data sets, animal feature data set, yeast gene expression data and leukemia gene expression data set. The result demonstrated the feasibility and utility of the proposed clustering algorithm.

• New & Renewable Energy
• /
• v.1 no.1 s.1
• /
• pp.44-53
• /
• 2005

To understand general status of the national wind environment and to distinguish potential areas to be developed as a largescale wind farm, a synoptic wind map of the Korean Peninsula is established by processing remote sensing data of the satellite, NASA QuikSCAT which Is deployed for the SeaWinds Project since 1999. According to the validation results obtained by comparing with the measurement data of marine buoys of KMA(Korea Meteorological Administration), the cross-correlation factor Is greatly Improved up to 0.87 by blending the sea-surface dat3 of QuikSCAT with NCEP/NCAR CDAS data. It is found from the established synoptic wind map that the wind speed in winter is prominent temporally and the South Sea shows high energy density up to the wind class 6 spatially. The reason is deduced that the northwest winds through the yellow Sea and the northeast winds through the East Sea derived by the low-pressure developed in Japan are accelerated passing through the Korea Channel and formed high wind energy region in the South Sea; the same trends are confirmed by the statistical analysis of meteorological observation data of KMA.

• Journal of the Korea Computer Graphics Society
• /
• v.18 no.3
• /
• pp.1-7
• /
• 2012

In this paper, we propose a method to compute the interface between protein molecules. When a molecules is represented as a set of spheres with van der Waals radii, the distance from a spatial point p to the molecule corresponds to the distance from p to the closet sphere. The molecular interface is composed of equi-distant points from two molecules. Our algorithm decomposes the space into a set of voxels, and then constructs a voxel map by storing the information of spheres intersecting each voxel. By using the voxel map, we compute the distance between a point and the molecule. We also use GPU for the parallel processing, and efficiently approximate the interface of a pair of molecules.

• Korean Journal of Computational Design and Engineering
• /
• v.7 no.4
• /
• pp.219-232
• /
• 2002

In numerically controlled(NC) machining simulation, a Z-map has been used frequently for representing a workpiece. Since the Z-map is usually represented by a set of Z-axis aligned vectors, the machining process can be simulated through calculating the intersection points between the vectors and the surface swept by a machining tool. In this paper, we present an efficient method to calculate those intersection points when an APT-type tool moves along a linear tool path. Each of the intersection points can be expressed as the solution of a system of non-linear equations. We transform this system of equations into a single-variable equation, and calculate the candidate interval in which the unique solution exists. We prove the existence of a solution and its uniqueness in this candidate interval. Based on these characteristics, we can effectively apply numerical methods to finally calculate the solution of the non-linear equations within a given precision. The whole process of NC simulation can be achieved by updating the Z-map properly. Our method can provide more accurate results with a little more processing time, in comparison with the previous closed-form solution.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2015.10a
• /
• pp.331-333
• /
• 2015

Data is increasing explosively with the spread of networks and mobile devices and there are problems in effectively processing the rapidly increasing data using existing recommendation techniques. Therefore, researches are being conducted on how to solve the scalability problem of the collaborative filtering technique. In this paper applies MapReduce, which is a distributed parallel process framework, to the collaborative filtering technique to reduce the scalability problem and heighten accuracy. The proposed technique applies MapReduce and the index technique to a user-based collaborative filtering technique and as a method which improves neighbor numbers which are used in similarity calculations and neighbor suitability, scalability and accuracy improvement effects can be expected.

• Journal of Korea Multimedia Society
• /
• v.14 no.5
• /
• pp.595-606
• /
• 2011

This paper proposes a robust watermarking scheme for GIS digital map by using the geometric properties of polyline and polygon, which are the fundamental components in vector data structure. In the proposed scheme, we calculate the lengths and the perimeters of all polylines and polygons in a map and cluster them to a number of groups. Then we embed the binary watermark by changing the mean of lengths and perimeters in an embedding group. For improving the safety and robustness, we permute the binary watermark through PRNS(pseudo-random number sequence) processing and embed it repeatedly in a model. Experimental results verified that our scheme has a good invisibility, safety and robustness to various geometric attacks and also our scheme needs not the original map in the extracting process of watermark.

• Journal of the Korean Association of Geographic Information Studies
• /
• v.19 no.3
• /
• pp.29-42
• /
• 2016

Remote sensing allows acquisition of information across a large area without contacting objects, and has thus been rapidly developed by application to different areas. Thus, with the development of remote sensing, satellites are able to rapidly advance in terms of their image resolution. As a result, satellites that use remote sensing have been applied to conduct research across many areas of the world. However, while research on remote sensing is being implemented across various areas, research on data processing is presently insufficient; that is, as satellite resources are further developed, data processing continues to lag behind. Accordingly, this paper discusses plans to maximize the performance of satellite image processing by utilizing the CUDA(Compute Unified Device Architecture) Library of NVIDIA, a parallel processing technique. The discussion in this paper proceeds as follows. First, standard KOMPSAT(Korea Multi-Purpose Satellite) images of various sizes are subdivided into five types. NDVI(Normalized Difference Vegetation Index) is implemented to the subdivided images. Next, ArcMap and the two techniques, each based on CPU or GPU, are used to implement NDVI. The histograms of each image are then compared after each implementation to analyze the different processing speeds when using CPU and GPU. The results indicate that both the CPU version and GPU version images are equal with the ArcMap images, and after the histogram comparison, the NDVI code was correctly implemented. In terms of the processing speed, GPU showed 5 times faster results than CPU. Accordingly, this research shows that a parallel processing technique using CUDA Library can enhance the data processing speed of satellites images, and that this data processing benefits from multiple advanced remote sensing techniques as compared to a simple pixel computation like NDVI.