• The Journal of the Petrological Society of Korea
• /
• v.17 no.3
• /
• pp.166-178
• /
• 2008

The modal analyses on the phenocryst phases and the normative mineralogies from the bulk chemical analyses record that the volcanic rocks related to the Yangsan caldera might been derived from compositionally zoned magma. The volcanic rocks show linearly continuous mineralogical gradients, not only totally in the relations between $SiO_2$ and proportion of phenocryst content, but also within each rock unit in the relations between total phenocryst content and the proportion, Q-Ab-Or and Q-An(Ab-Or) diagrams. The roughly gradational modal variations of the phenocryst phases are shown upward within each rock unit. However, the contents and proportion of the phenocrysts in the Yangsan Tuff and the Hwajeri Formation represent the zigzaggedly undulatory variations. The continuous mineralogical gradients without large gaps define a large zoned magma system in the pre-eruptive, later precollapse and postcollapse magma chambers respectively. The zigzagged variations reflect the intermittent eruptive pulses representing any time gaps.

• The Bulletin of The Korean Astronomical Society
• /
• v.38 no.1
• /
• pp.30.1-30.1
• /
• 2013

The sizes of galaxies are known to be closely related with their masses, luminosities, redshifts and morphologies. However, when we fix these quantities and morphology, we still find large dispersions in the galaxy size distribution. We investigate the origin of these dispersions for red early-type galaxies, using two SDSS-based catalogs. We find that the sizes of faint galaxies (${\log}(M_{dyn}/M_{\odot})$ < 10.3 or $M_r$ >-19.5) are affected more significantly by luminosity, while the sizes of bright galaxies (${\log}(M_{dyn}/M_{\odot})$ > 11.4 or $M_r$ <-21.4) are by dynamical mass. At fixed mass and luminosity, the sizes of low-mass galaxies (${\log}(M_{dyn}/M_{\odot})$ ~ 10.45 and $M_r$~-19.8) are relatively less sensitive to their colors, color gradients and axis ratios. On the other hand, the sizes of intermediate-mass (${\log}(M_{dyn}/M_{\odot})$ ~ 10.85 and $M_r$~-20.4) and high-mass (${\log}(M_{dyn}/M_{\odot})$ ~ 11.25 and $M_r$~-21.0) galaxies significantly depend on those parameters, in the sense that larger red early-type galaxies have bluer colors, more negative color gradients (bluer outskirts) and smaller axis ratios. These results indicate that the sizes of intermediate- and high-mass red early-type galaxies are significantly affected by their recent minor mergers or rotations. Major dry mergers also may have influenced on the size growth of high-mass red early-type galaxies.

• The Korean Journal of Ecology
• /
• v.18 no.1
• /
• pp.63-87
• /
• 1995

The present study was undertaken to classify and describe the Quercus mongolica forest on Mt. Sobaek (36°50'∼ 37°07'N, 128°25'∼ 128°35'E), Korea by a reciprocal averaging methol (RA ordination) and the methodology of the ZM school of phytosociology. Atemporal vegetation tabel was prepared by the RA ordination. As a result it showed that the RA ordination on the first axis is very applicable for selecting the character and differential species of vegetation units of the Quercus mongolica forest. The phytosociological investigations were carried out in 34 stands on the mountain. The field data obtained were classified by the tabel comparison method. For the establishment of new vegetation units, the floristic composition of the Quercus mongolica forest was compared with those of the other vegetation units of the Mt. Sobaek, Considering the candidates of the character and differential species derived from RA ordination. As the result, the forest vegetation was classified into the following vegetation units; Fagetea crenatae Miyawaki et al. 1968 Acero-Quercetalia mongolicas Song 1988 em. Takeda et al.1994 Lindero-Quercetalia mongolicae Kim 1990 1. Lindero-Quercetum mongolicae assoc. nov. 1-1. quercetosum serratae subassoc. nov 1-2. typicum subassoc. nov. Rhododendro-Qurcion mongolicas Song 1988 em. Takeda et al. 1994. It is inferred that the Lindero-Quercion mongolicae and the Rhododendro-Quercion mongolicae corresponded to the cool-temperate forests of central and northern areas, respectively, in Korean Peninsula. Syntaxonomical synonyms were also rearranged based on the result of the present study. On the other hand, the RA ordination of stands and species on the first and second axes revealed some environmental gradients such as altitute, human impact and species richness between the vegetation units. However, the RA ordinations is considered to be less indicative for the gradients of the lower vegetation units than association. With the present study. some phytosociological problems in Korea were discussed here in detail from the floristic viewpoint.

• KIPS Transactions on Software and Data Engineering
• /
• v.3 no.9
• /
• pp.361-368
• /
• 2014

The LPR(License plate recognition) system has been developed to efficient control for complex traffic environment and currently be used in many places. However, because of light, noise, background changes, environmental changes, damaged plate, it only works limited environment, so it is difficult to use in real-time. This paper presents a heuristic segmentation algorithm for robust to noise and illumination changes and introduce a real-time license plate recognition system using it. In first step, We detect the plate utilized Haar-like feature and Adaboost. This method is possible to rapid detection used integral image and cascade structure. Second step, we determine the type of license plate with adaptive histogram equalization, bilateral filtering for denoise and segment accurate character based on adaptive threshold, pixel projection and associated with the prior knowledge. The last step is character recognition that used histogram of oriented gradients (HOG) and multi-layer perceptron(MLP) for number recognition and support vector machine(SVM) for number and Korean character classifier respectively. The experimental results show license plate detection rate of 94.29%, license plate false alarm rate of 2.94%. In character segmentation method, character hit rate is 97.23% and character false alarm rate is 1.37%. And in character recognition, the average character recognition rate is 98.38%. Total average running time in our proposed method is 140ms. It is possible to be real-time system with efficiency and robustness.

• Korean Journal of Environmental Biology
• /
• v.31 no.4
• /
• pp.419-428
• /
• 2013

The objectives of this research were to determine ecological characteristics and chemical gradients in two different loach populations-Misgurnus anguillicaudatus ($M_a$) and Koreocobitis rotundicaudata ($K_r$). Eight chemical parameters were compared to test the chemical preferences between the two species. The population of Kr occurred more in better water quality, compared to the $M_a$-population. The $M_a$-population co-occurred more with tolerant species (TS) and omnivores, whereas the $K_r$-population co-occurred more with sensitive species (SS) and insectivores. The differences in the tolerance and trophic composition matched well with the result of stream health assessment, based on the multi-metric model of Index of Biological Integrity (IBI). The IBI value in habitat with the $M_a$-population was "19.6" (range: 8~40) indicating a fair condition, and the IBI values of $K_r$-population was 29.4 (range: 12~40) indicating good condition. Similar results were found in physical habitat health analysis (QHEI). Overall, data analysis showed that the $M_a$-population belong to the tolerant species and $K_r$-population belong to sensitive species in the tolerance classification.

• Journal of radiological science and technology
• /
• v.21 no.1
• /
• pp.35-39
• /
• 1998

We analyzed image factors to determine the characteristic factors that need for intelligent replenishment system of the auto film processor. We processed the serial 300 sheets of radiographic films of chest phantom without replenishment of developing and fixation replenisher. We took the digital data by using film digitizer which scaned the films and automatically summed up the pixel values of the films. We analyzed characteristic curves, average gradients and relative speeds of individual film using densitometer and step densitometry. We also evaluated the pH of developer, fixer, and washer fluid with digital pH meter. Fixer residual rate and washing effect were measured by densitometer using the reagent methods. There was no significant reduction of the digital density numbers of the serial films without replenishment of developer and fixer. The average gradients were gradually decreased by 0.02 and relative speeds were also gradually decreased by 6.96% relative to initial standard step-densitometric measurement. The pHs of developer and fixer were reflected the inactivation of each fluid. The fixer residual rates and washing effects after processing each 25 sheets of films were in the normal range. We suggest that the digital data are not reliable due to limitation of the hardware and software of the film digitizer. We conclude that average gradient and relative speed which mean the film's contrast and sensitivity respectively are reliable factors for determining the need for the replenishment of the auto film processor. We need more study of simpler equations and programming for more intelligent replenishment system of the auto film processor.

• 한국해양학회지
• /
• v.26 no.3
• /
• pp.223-241
• /
• 1991

In order to study the formation and structure of tidal fronts and their influence on the distribution and productivity of phytoplankton in the outer of Kyonggi Bay, analyses on the water temperature data from 1977 to 1986 and 3 surveys from 1981 to 1986 were carried out in the mid0eastern coast of the Yellow Sea. Temperature gradients and dissolved oxygen gradients were implied that the tidal fronts are formed at the outer of the Kyonggi Bay along the western side of Tae-An peninsula from spring to summer. the formations of tidal fronts in this study area influence the distribution of phytoplankton and primary productivity. The standing stocks, chlorophyll concentrations and primary productivity of phytoplankton in the frontal area are higher than those of the outer stratified waters and the inner coastal mixed waters. These high production in the frontal area are resulted from good light condition and rich nutrient within the water columns. With a boundary of frontal area, there are relatively high chlorophyll concentrations and primary productivity in the coastal mixed waters while there are low chlorophyll concentrations and relatively high primary productivity in the stratified waters. These relatively high primary productivity in the outer area are resulted from the high potential production by nanoplankton in the surface layer and the high production of tychopelagic diatoms under the thermocline with the deep transparency.

• Korean Journal of Remote Sensing
• /
• v.5 no.1
• /
• pp.57-67
• /
• 1989

This study on the southern coastal waters of Korea has been made by analysis of NOAA image and oceanographic observation data from October 1987 to August 1988. The results obtained from the study are as follow: Horizontal distributions of water temperature in different layers in winter ranged from 6.07 to 18.62$^{\circ}C$ at 0m layer, 6.02 to 18.54$^{\circ}C$ at 30m layer and 7.19 to 18.69$^{\circ}C$ at 50m layer. Consequently its vertical distribution showed homogeneity. Horizontal water temperature gradients were 0.28$^{\circ}C$/mile between the coastal waters and Tsushima warm waters. In summer, its horizontal distribution varied from 19.37 to 29.92$^{\circ}C$ at 0m layer, 13.26 to 27.11$^{\circ}C$ at 30m layer and 7.36 to 26.6$0^{\circ}C$ at 50m layer, and its vertical profile showed stratified structure. Vertical water temperature gradients were 0.44$^{\circ}C$/m between 30 and 50m layers. It was remarkable that distribution of southern coastal water system analysed by NOAA image coincided with relatively the oceanographic observation data but SST from NOAA image seemed to be 2-4$^{\circ}C$ lower in winter and 4-6$^{\circ}C$ lower in summer than the oceanographic data.

• Korean Journal of Soil Science and Fertilizer
• /
• v.40 no.6
• /
• pp.488-494
• /
• 2007

Highland regions for farming are generally located in slopes higher than 7%, where alpine farming systems rely on highly input agriculture management with great amounts of chemical fertilizer and/or compost. Most of the uplands is thus needed to maintain environmentally friendly soil management due to its impact on soil erosion and runoff during heavy rainfall season. Therefore, the objective of this research is to evaluate the effect of reduction of soil erosion by applying four wild edible greens (fatsia, goat beard, leopard plant, and aster). The lysimeter experiment of slope gradients of 15, 30, and 45% was conducted in an alpine region of Hoengkye, Kangwon, in 2005 and 2006. In 2005, both amounts of soil loss from the experiment plots cultivated with goat beard and aster were lower than one with Chinese cabbage by about 50%. The amounts of runoff of goat beard and aster plots were also lower than those of the others. An increase in the slope gradients was accompanied with an increase in runoff. Of the plots of slope gradient of 15, 30, and 45%, S of goat beard plots was 52.50, 108.33, and 171.50 kg, respectively. Soil loss of Chinese cabbage was 2 to 3 times as high as those of goat beard plots. These results suggest that goat beard and aster plants with minimum tillage reduce soil erosion compared to Chinese cabbage cultivation.

• Journal of Internet Computing and Services
• /
• v.20 no.4
• /
• pp.91-102
• /
• 2019

In autonomous driving systems, the ability to classify pedestrians in images captured by cameras is very important for pedestrian safety. In the past, after extracting features of pedestrians with HOG(Histogram of Oriented Gradients) or SIFT(Scale-Invariant Feature Transform), people classified them using SVM(Support Vector Machine). However, extracting pedestrian characteristics in such a handcrafted manner has many limitations. Therefore, this paper proposes a method to classify pedestrians reliably and effectively using CNN's(Convolutional Neural Network) deep features and transfer learning. We have experimented with both the fixed feature extractor and the fine-tuning methods, which are two representative transfer learning techniques. Particularly, in the fine-tuning method, we have added a new scheme, called M-Fine(Modified Fine-tuning), which divideslayers into transferred parts and non-transferred parts in three different sizes, and adjusts weights only for layers belonging to non-transferred parts. Experiments on INRIA Person data set with five CNN models(VGGNet, DenseNet, Inception V3, Xception, and MobileNet) showed that CNN's deep features perform better than handcrafted features such as HOG and SIFT, and that the accuracy of Xception (threshold = 0.5) isthe highest at 99.61%. MobileNet, which achieved similar performance to Xception and learned 80% fewer parameters, was the best in terms of efficiency. Among the three transfer learning schemes tested above, the performance of the fine-tuning method was the best. The performance of the M-Fine method was comparable to or slightly lower than that of the fine-tuningmethod, but higher than that of the fixed feature extractor method.