• Korean Journal of Remote Sensing
• /
• v.37 no.1
• /
• pp.111-122
• /
• 2021

Building footprint extraction is an active topic in the domain of remote sensing, since buildings are a fundamental unit of urban areas. Deep convolutional neural networks successfully perform footprint extraction from optical satellite images. However, semantic segmentation produces coarse results in the output, such as blurred and rounded boundaries, which are caused by the use of convolutional layers with large receptive fields and pooling layers. The objective of this study is to generate visually enhanced building objects by directly extracting the vertices of individual buildings by combining instance segmentation and keypoint detection. The target keypoints in building extraction are defined as points of interest based on the local image gradient direction, that is, the vertices of a building polygon. The proposed framework follows a two-stage, top-down approach that is divided into object detection and keypoint estimation. Keypoints between instances are distinguished by merging the rough segmentation masks and the local features of regions of interest. A building polygon is created by grouping the predicted keypoints through a simple geometric method. Our model achieved an F1-score of 0.650 with an mIoU of 62.6 for building footprint extraction using the OpenCitesAI dataset. The results demonstrated that the proposed framework using keypoint estimation exhibited better segmentation performance when compared with Mask R-CNN in terms of both qualitative and quantitative results.

• Journal of Korean Library and Information Science Society
• /
• v.48 no.3
• /
• pp.233-257
• /
• 2017

In this paper, we propose an effective method to convert and construct the MEDLINE, the world's largest biomedical bibliographic database, into linked data. To do this, we first derive the appropriate RDF schema by analyzing the MEDLINE record structure in detail, and convert each record into a valid RDF file in the derived schema. We apply the dual batch registration method to streamline the subject URI duplication checking procedure when merging all RDF files in the converted record unit and storing it in a single RDF triple storage. By applying this method, the number of RDF triple storage accesses for the subject URI duplication is reduced from 26,597,850 to 2,400, compared with the sequential configuration of linked data in units of RDF files. Therefore, it is expected that the result of this study will provide an important opportunity to eliminate the inefficiency in converting large volume bibliographic record sets into linked data, and to secure promptness and timeliness.

• Journal of Korean Society on Water Environment
• /
• v.26 no.1
• /
• pp.104-110
• /
• 2010

Eutrophication is a well-known phenomenon in lentic habitats, however it is receiving increasing attention in shallow streams of Korea due to the increase of periphyton to a nuisance level. In this study temporal and spatial variation in periphyton standing crop and nutrient concentrations were surveyed in the upper reach of the Han River (the Inbuk Stream) that used to be a pristine rural stream until 1980s. Chlorophyll-a concentration per unit surface area of bottom substrate was examined monthly for one year period at nine sites along the Inbuk Stream together with environmental factors such as phosphorus, nitrogen, and water velocity. The standing crop of periphytic algae ranged from $4{\sim}242mgChl.a/m^2$ with a median of $55mgChl.a/m^2$, often exceeding the nuisance level criterion. Along the stream periphyton increased significantly from $39{\pm}48mgChl.a/m^2$ to $94{\pm}49mgChl.a/m^2$ after merging of a tributary in an intensive agricultural basin with high phosphorus concentrations. Seasonally periphyton biomass was highest in autumn (median $171{\pm}76mgChl.a/m^2$) from October through December when water flow velocity was low (median $0.4{\pm}0.3m/s$), while it was higher in flood season (median $1.2{\pm}0.4m/s$) and freezing season (median $0.2{\pm}0.3m/s$) was lower. The result shows that this rural stream often shows characteristics of eutrophication according to periphyton standing crop and it may be regulated by phosphorus and water velocity.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.659-663
• /
• 2023

이 논문은 한국어 음성인식 성능을 개선하고자 기존 음성인식 과정을 자모단위 음성인식 모델과 신경망 기반 자모 병합 모델 총 두 단계로 구성하였다. 한국어는 조합어 특성상 음성 인식에 필요한 음절 단위가 약 2900자에 이른다. 이는 학습 데이터셋에 자주 등장하지 않는 음절에 대해서 음성인식 성능을 저하시키고, 학습 비용을 높이는 단점이 있다. 이를 개선하고자 음절 단위의 인식이 아닌 51가지 자모 단위(ㄱ-ㅎ, ㅏ-ㅞ)의 음성인식을 수행한 후 자모 단위 인식 결과를 음절단위의 한글로 병합하는 과정을 수행할 수 있다[1]. 자모단위 인식결과는 초성, 중성, 종성을 고려하면 규칙 기반의 병합이 가능하다. 하지만 음성인식 결과에 잘못인식된 자모가 포함되어 있다면 최종 병합 결과에 오류를 생성하고 만다. 이를 해결하고자 신경망 기반의 자모 병합 모델을 제시한다. 자모 병합 모델은 분리되어 있는 자모단위의 입력을 완성된 한글 문장으로 변환하는 작업을 수행하고, 이 과정에서 음성인식 결과로 잘못인식된 자모에 대해서도 올바른 한글 문장으로 변환하는 오류 수정이 가능하다. 본 연구는 한국어 음성인식 말뭉치 KsponSpeech를 활용하여 실험을 진행하였고, 음성인식 모델로 Wav2Vec2.0 모델을 활용하였다. 기존 규칙 기반의 자모 병합 방법에 비해 제시하는 자모 병합 모델이 상대적 음절단위오류율(Character Error Rate, CER) 17.2% 와 단어단위오류율(Word Error Rate, WER) 13.1% 향상을 확인할 수 있었다.

• Journal of Korea Multimedia Society
• /
• v.12 no.2
• /
• pp.209-218
• /
• 2009

3D backprojection is a kind of reconstruction algorithm to generate volume data consisting of tomographic images, which provides spatial information of the original 3D data from hundreds of 2D projections. The computational time of backprojection increases in proportion to the size of volume data and the number of projection images since the value of every voxel in volume data is calculated by considering corresponding pixels from hundreds of projections. For the reduction of computational time, fast GPU based 3D backprojection methods have been studied recently and the performance of them has been improved significantly. This paper presents two multiple GPU based methods to maximize the parallelism of GPU and compares the efficiencies of two methods by considering both the number of projections and the size of volume data. The first method is to generate partial volume data independently for all projections after allocating a half size of volume data on each GPU. The second method is to acquire the entire volume data by merging the incomplete volume data of each GPU on CPU. The in-complete volume data is generated using the half size of projections after allocating the full size of volume data on each GPU. In experimental results, the first method performed better than the second method when the entire volume data can be allocated on GPU. Otherwise, the second method was efficient than the first one.

• The KIPS Transactions:PartA
• /
• v.18A no.4
• /
• pp.135-142
• /
• 2011

This research is to design an effective buffer structure and its management for flash memory based high performance SSDs (Solid State Disks). Specifically conventional SSDs tend to show asymmetrical performance in read and /write operations, in addition to a limited number of erase operations. To minimize the number of erase operations and write latency, the degree of interleaving levels over multiple flash memory chips should be maximized. Thus, to increase the interleaving effect, an effective buffer structure is proposed for the SSD with a hybrid address mapping scheme and super-block management. The proposed buffer operation is designed to provide performance improvement and enhanced flash memory life cycle. Also its management is based on a new selection scheme to determine random and sequential accesses, depending on execution characteristics, and a method to enhance the size of sequential access unit by aggressive merging. Experiments show that a newly developed mapping table under the MBA is more efficient than the basic simple management in terms of maintenance and performance. The overall performance is increased by around 35% in comparison with the basic simple management.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.15 no.5
• /
• pp.71-82
• /
• 2016

Information System for rear-end collision avoidance is a unit service of C-ITS pilot project. Road environment that the number of vehicles at the same driving high-speed has a possibility that the communication delay or failure caused by heavy load of vehicle to vehicle communication. In this study, effects of the road network about a communication failure rate of information system for rear-end collision avoidance was analyzed quantitatively with micro traffic simulation. The simulation was carried out in situation that crash of two vehicles are occurred at merging area with speed limit 80km/h and information of collision is prvoided to the rear vehicle. From simulation results, it can confirm the trend of the increasing 14% of potential conflict according to 10% increasing of the communication failure rate. C-ITS service has a goal of increasing safety. The coommunication failure rate increases due to heavy load of vehicle causes a fatal result in road safety administrator position. For the success of C-ITS project, a communication system developers side should perform the effort to reduce the communication failure rate.

• Journal of the Korea Society for Simulation
• /
• v.19 no.1
• /
• pp.143-151
• /
• 2010

Fingerprint classification is a step to increase the efficiency of an 1:N fingerprint recognition system and plays a role to reduce the matching time of fingerprint and to increase accuracy of recognition. It is difficult to classify fingerprints, because the ridge pattern of each fingerprint class has an overlapping characteristic with more than one class, fingerprint images may include a lot of noise and an input condition is an exceptional case. In this paper, we propose a novel approach to design a stochastic model and to accomplish fingerprint classification using a directional characteristic of fingerprints for an effective classification of various qualities. We compute the directional value by searching a fingerprint ridge pixel by pixel and extract a directional characteristic by merging a computed directional value by fixed pixels unit. The modified Markov model of each fingerprint class is generated using Markov model which is a stochastic information extraction and a recognition method by extracted directional characteristic. The weight list of classification model of each class is decided by analyzing the state transition matrixes of the generated Markov model of each class and the optimized value which improves the performance of fingerprint classification using GA (Genetic Algorithm) is estimated. The performance of the optimized classification model by GA is superior to the model before the optimization by the experiment result of applying the fingerprint database of various qualities to the optimized model by GA. And the proposed method effectively achieved fingerprint classification to exceptional input conditions because this approach is independent of the existence and nonexistence of singular points by the result of analyzing the fingerprint database which is used to the experiments.

• Asia Marketing Journal
• /
• v.12 no.4
• /
• pp.183-201
• /
• 2011

In case of Aju capital, it adopted a strategy to use a single brand not two separate brands after M&A was completed. In order to implement this strategy, it has endeavored to effectively process the work of shifting existing marketing infrastructure of DAEWOO capital, the mergee, spending enough transition time for the brand migration. In the process of merging, Aju capital picked the strategy to use the brand of mergee first, which is the Daewoo Capital brand, and then took a transition time for a while to converge to the single brand of Aju capital. Putting another way, even if the M&A deal was completed back in 2005, it maximized the effect of launching its final brand "Aju capital" by capitalizing on the positive image of "Daewoo" during the transition time and changing its name just in the right moment. In a bid to implement this strategy successfully, it established a cautious but sophisticated brand migration strategy. 1) "Brand bridge" strategy through reinforcing brand power of "Naegeron", which is an individual product brand of Daewoo Capital 2) Establishing a good brand image through reinforcing customer satisfaction 3) It implemented and completed its brand transition initiative by going through the step of Aju Capital brand unification (from Sept 09 to present) Currently, the sales unit of Aju Capital is realizing quality growth through specialization. It's strategy is to construct a systematic sales portfolio in terms of both quality and quantity through product-by-product specialization where the existing practice was selling a variety of products in a single branch. Back in 2009, it opened a branch that specialize in imported cars and expanded its used car business to 6 specialized locations. Besides, the specialized locations for personal loan named "Naegeron" was expanded from 3 to 11 locations. Recently, it is expected that it will inject vigor to retail and corporate financing business alongside with its core business, which is auto financing.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.11
• /
• pp.140-149
• /
• 1996

Cursive characters are written on an economical principle to reduce the motion of a pen in the limit of distinction between characters. That is, the pen is not lifted up to move for writing a next stroke, the pen is not moved at all, or connected two strokes chance their shapes to a similar and simple shape which is easy to be written. For these reasons, strokes and korean alphabets are not only easy to be changed, but also difficult to be splitted. In this paper, we propose a curvature stroke modeling method for splitting and matching by using a structural primitive. A curvature stroke is defined as a substroke which does not change its curvanture. Input strokes handwritten in a cursive style are splitted into a sequence of curvature strokes by segmenting the points which change the direction of rotation, which occur a sudden change of direction, and which occur an excessive rotation Each reference of korean alphabets is handwritten in a printed style and is saved as a sequence of curvature strikes which is generated by splitting process. And merging process is used to generate various sequences of curvature strikes for matching. Here, it is also considered that imaginary strokes can be written or omitted. By using a curvature stroke as a unit of recognition, redundant splitting points in input characters are effectively reduced and exact matching is possible by generating a reference curvature stroke, which consists of the parts of adjacent two korean alphasbets, even when the connecting points between korean alphabets are not splitted. The results showed 83.6% as recognition rate of the first candidate and 0.99sec./character (CPU clock:66MHz) as processing time.