• Journal of the Korea Convergence Society
• /
• v.12 no.11
• /
• pp.45-51
• /
• 2021

With the development of deep learning, semantic segmentation methods are being studied in various fields. There is a problem that segmenation accuracy drops in fields that require accuracy such as medical image analysis. In this paper, we improved PSPNet, which is a deep learning based segmentation method to minimized the loss of features during semantic segmentation. Conventional deep learning based segmentation methods result in lower resolution and loss of object features during feature extraction and compression. Due to these losses, the edge and the internal information of the object are lost, and there is a problem that the accuracy at the time of object segmentation is lowered. To solve these problems, we improved PSPNet, which is a semantic segmentation model. The multi-scale attention proposed to the conventional PSPNet was added to prevent feature loss of objects. The feature purification process was performed by applying the attention method to the conventional PPM module. By suppressing unnecessary feature information, eadg and texture information was improved. The proposed method trained on the Cityscapes dataset and use the segmentation index MIoU for quantitative evaluation. As a result of the experiment, the segmentation accuracy was improved by about 1.5% compared to the conventional PSPNet.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.13 no.1
• /
• pp.77-83
• /
• 1995

This paper is concentrated on map generalization in digital environment for automated multi-scale map pro-duction using conventional hardcopy maps. Line generalization is urgently required process to prepare small scale digital map database when large scale map databases are available. This paper outlines a new approach to the line generalization when preparing small scale map on the basis of existing large scale distal map. Line generalizations are conducted based on zero-crossing algorithm using six sheets of 115,000 scale YEOSU area which produced by National Geographic Institute. The results are compared to Douglas-Peucker algorithm and manual method. The study gives full details of the data reduction rates and alternatives based on the proposed algorithm.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2009.05b
• /
• pp.189-193
• /
• 2009

As large-scale buildings, skyscrapers, and multi-purpose buildings recently increase in numbers dramatically the internal space of such buildings becomes more and more large and complicated accordingly. The Multi-purpose buildings are increasing rapidly in number, and concerns over the evacuation process in case of fire emergency are also increasing. However, A controller who is responsible for visitor's safety makes a decision about measures for visitor safety in human based decision making process. Many potential accidents that are caused by human error lurk in results of the process. The accidents can be decreased by changing the decision making process from human-based into technology-based. Through this study, the way to improve current disasters prevention system is presented and it can also improve shelter inducement methods effectively. SICS(Spatial Information Control System) which is being studied in other research is used as the tool for improvement.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.38 no.4
• /
• pp.335-343
• /
• 2001

The conventional wavelet-based fractal image coding has the disadvantage that the encoding takes a long time, since each range block finds the best domain in the image. In this Paper, we propose wavelet-based fractal image coding using SAS(Self Affine System) method and multi-scale factor. It consists of the range and domain blocks in DWT(discrete wavelet transform) region. Using SAS method, the proposed method is that the searching process of the domain block is not required, and the range block selects the domain which is relatively located the same position in the upper level. The proposed method can perform a fast encoding by reducing the computational complexity in the encoding process. In order to improve the disadvantage of SAS method which is reduced image qualify, the proposed method is improved image qualify using the different scale factors for each level. As a result, there is not influence on an image quality, the proposed method is enhanced the encoding time and compression ratio, and it is able to the progressive transmission.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.14 no.6
• /
• pp.77-83
• /
• 2015

This paper is a study of the functions required for the system to quantify the closest target weight by combining several random weights such as chips, snacks, fruits, and vegetables. The multi-head weigher is designed for high-performance applications requiring increased production rates and tight accuracy tolerances. This combination system has 12 heads considered in the form of a rectangular array of $2{\times}6$ or $3{\times}4$. Channel combination can usually occur between 1 and n, and the frequency was the highest with two or three combinations. Experimental result of a combination system for a total target weight was measured at the range from 100g to 500g by increments of 50g, and the average success rate was about 70%. The average elapsed time was about 1.7 seconds, which means it can be used for the packaging of agricultural products with a variety of items.

• Electrical & Electronic Materials
• /
• v.7 no.2
• /
• pp.145-151
• /
• 1994

The multi-dielectric layer SiOz/Si3N4/SiO2(ONO) is used to scale down the memory device. In this paper, the change of composition in ONO layer due to the process condition and the conduction mechanism are observed. The composition of the oxide film grown through the oxidation of nitride film is analyzed using auger electron spectroscopy(AES). AES results show that oxygen concentration increases at the interface between oxide and nitride layers as the thickness -of the top oxide layer increases. Results of I-V measurement show that the insulating properties improve as the thickness of the top oxide layer increases. But when the thickness of the nitride layer decreases below 63.angs, insulating peoperties of film 28.angs. of top oxide and film 35.angs. turn over showing that insulating property of film 28.angs. of top oxide is better than that of film 35.angs. of top oxide. This phenomenon of turn over is thought as the result of generation of surface state due to oxygen flow into nitride during oxidation process. As the thickness of the top oxide and nitride increases, the electrical breakdown field increases, but when the thickness of top oxide reaches 35.angs, the same phenomenon of turn over occurs. Optimum film thickness for scaled multi-layer dielectric of memory device SONOS is estimated to be 63.angs. of nitride layer and 28.angs. of top oxide layer. In this case, maximum electrical breakdown field and leakage current are 18.5[MV/cm] and $8{\times}{10^-12}$[A], respectively.

• Journal of Electrochemical Science and Technology
• /
• v.2 no.3
• /
• pp.157-162
• /
• 2011

We report a simple route for synthesizing multi-dimensional structured silicon anode materials from commercially available bulk silicon powders via metal-assisted chemical etching process. In the first step, silver catalyst was deposited onto the surface of bulk silicon via a galvanic displacement reaction. Next, the silver-decorated silicon particles were chemically etched in a mixture of hydrofluoric acid and hydrogen peroxide to make multi-dimensional silicon consisting of one-dimensional silicon nanowires and micro-scale silicon cores. As-synthesized silicon particles were coated with a carbon via thermal decomposition of acetylene gas. The carbon-coated multi-dimensional silicon anodes exhibited excellent electrochemical properties, including a high specific capacity (1800 mAh/g), a stable cycling retention (cycling retention of 89% after 20 cycles), and a high rate capability (71% at 3 C rate, compared to 0.1 C rate). This process is a simple and mass-productive (yield of 40-50%), thus opens up an effective route to make a high-performance silicon anode materials for lithiumion batteries.

• Korean Journal of Remote Sensing
• /
• v.17 no.3
• /
• pp.189-197
• /
• 2001

Generally, remote sensing is useful to obtain the quantitative and qualitative information of a wide area. For monitoring earth resources and environment, land cover classification of remotely sensed data are needed over increasingly larger area. The objective this study is to propose the process for land cover classification method over a wide area using multi-scene satellite data. Land cover of Korean peninsula was extracted from a Landsat TM and ETM+ mosaic created from 23 scenes at 100-meter resolution. Well-known techniques that used to general image processing and classification are applied to this wide area classification. It is expected that these process is very useful to promptly and efficiently grasp of small scale spatial information such as national territorial information.

• Membrane Journal
• /
• v.17 no.3
• /
• pp.197-209
• /
• 2007

In this study, a multi-staged pilot-scale membrane plant was constructed and operated for the separation of $CO_2$ from LNG-fired boiler flue gas of 1,000 $Nm^3/day$. The target purity and recovery ratio of $CO_2$ required for the pilot plant were 99% and 90%, respectively. For this purpose, we previously developed the asymmetric polyethersulfone hollow fibers and evaluated the effects of operating pressure and feed concentration of $CO_2$ on separation performance[1,2]. The permeation data obtained were also analyzed in relation with the numerical simulation data using counter-current flow model[3,4]. Based on these results, we designed and prepared the demonstration plant consisting of dehumidification process and four-staged membrane process. The operation results using this plant were compared with the numerical simulation results on multi-staged membrane process. The experimental results matched well with the numerical simulation data. The concentration and the recovery ratio of $CO_2$ in the final stage permeate stream were ranged from $95{\sim}99%$ and $70{\sim}95%$, respectively, depending on the operating conditions. This study demonstrated the applicability of the membrane-based pilot plant for $CO_2$ recovery from flue gas.

• Journal of Korean Society of Industrial and Systems Engineering
• /
• v.25 no.4
• /
• pp.22-31
• /
• 2002

A high-speed rail system represents a typical example of large-scale multi-disciplinary systems, consisting of subsystems such as train, electrical hardware, electronics, control, information, communication, civil technology etc. The system design and acquisition data of the large-scale system must be the subject under strict configuration control and management. Not only the requirements of the large-scale system dictate the contracts with the suppliers but also become the basis for the development process, project execution, system integration, and testing. The requirements database provide the system design specification of all development activities. Using the RDD-100, a systems engineering tool, the Korea next-generation high-speed rail program can establish requirements traceability and development process management in performing the enabling train technology development projects. This paper presents the results from a computer-aided systems engineering application to the Korea next-generation high-speed railway project. Especially, the focus of the study was on requirement management and PBS(Product Breakdown Structure) management.