Earticle

본 논문에서는 기존 U-Net모델을 기반으로 skip-connection 과정에 cross-attention을 적용하여 retinal vessel segmentation 성능을 향상 시키는 모델을 제안한다. U-Net은 encoding과 decoding 과정 동안 발생한 edge information의 손실을 보완하기 위해 skip-connection 구조를 사용한다. 하지만 이러한 구조는 불필요한 information이 함께 전달되어 noise가 발생하는 문제점이 있다. 이를 해결하기 위해 skip-connection에 attention gate를 적용하여 필요한 성분만 추출하는 attention U-Net 모델이 제안되었다. 그러나 encoder의 상위 layer feature에는 context information이 부족하기 때문에 pixel classification 성능에 열화를 일으킬 수 있다는 문제점이 존재한다. 따라서 본 논문은 기본 U-Net의 skip-connection 과정에서 발생하는 edge와 context information 손실을 방지하기 위해 crossattention 구조를 사용한 새로운 network를 제안한다. 제안된 방법은 Drive dataset에서 96.89%의 분류 정확도를 보여, 기존 U-Net 모델과 Attention U-Net 모델 대비 가장 좋은 성능을 보인다.

This paper looks into student performance in the use of the segmentation strategy while doing sight translation. The paper finds that while students are using the segmentation strategy in their performances, as indicated by a simple comparison of the number of sentences in the original text with the target text transcription, there is room for improvement in a more accurate use of the falling and rising intonation to indicate that there is indeed a segmentation that took place in the target text. Also, the students should be encouraged to develop a larger pool of discourse markers that can be used in making the segmented parts to be logically connected while fully conveying the message of the original text. The experiment result of this study also finds the need for a more detailed definition of the terms used in the analysis, such as ‘sentence’: a term well-defined and accepted in the written text, but that needs to be redefined in the audio input of the SI data.

Traditional methods for plant phenotypic analysis, including attributes such as color, plant health status, and size, rely on expert manual analysis and judgment. Measurements are manually taken using tools, and all data is recorded by hand, which is highly inefficient. However, the development of artificial intelligence and deep learning has provided efficient solutions for plant phenotypic analysis. This study established a framework based on high-resolution images and utilized the SEGMENT ANYTHING MODEL (SAM) and Explainable Contrastive Language-Image Pretraining (ECLIP) for plant phenotypic analysis. Through the segmentation results of this model, the length and width of radishes, cucumbers, and pumpkins were measured. Since this method performed experiments without requiring model training or manual annotation of data, the framework demonstrated strong efficiency in the segmentation tasks prior to phenotypic analysis, significantly reducing manual labor costs. Moreover, the experimental results indicated that the mean absolute error (MAE) was below 0.05 for most test samples.

Detecting mass lesions not only helps reduce the cost of treating breast cancer but also enhances the lifespan of patients. Various computer-aided detection (CAD) systems have been developed to assist physicians in detecting mass in mammograms for early cancer screening. In this paper, a method for suspicious massive lesion segmentation in patches is proposed, which modified UNet with EfficientNet as the encoder. The proposed architectures are evaluated on publicly available dataset, namely the Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSM). The quantitative results show that the proposed architecture can achieve mass segmentation with segmentation ac- curacy, Dice and IoU scores of 95.23%, 92.56% and 88.81% respectively in patches extracted from CBIS-DDSM.

This study is to observe watching and purchasing characteristics of licensing product according to visiting frequency of female spectators to professional baseball game and to segment the female spectators based on the demographical characteristic, watching characteristic and purchasing characteristic of licensing product to provide basic data for professional baseball team’s marketing strategy. First, in the demographical characteristic of female spectators in professional baseball game, female in 20’s were dominated both in high and low watching groups and they were mainly unmarried female student who are currently enrolled in University or graduated. Second, in watching characteristics, most females in high and low watching groups visited sports stadium with friend or lover to cheer a team or player and they spend mainly for beverage and meal. Third, in purchasing characteristic of licensing product, there were lots of females who have bought in high and low watching groups and they purchased mainly clothes and caps. Also, in case of low watching group, there were many non buyers. Major purchasing type was to purchase cheering tool including clothes or caps.

영상 세션화는 기본적으로 이미지를 의미있는 구조로 나누는 것이다. 항상 특정 요구에 따라 이미지를 완전히 분할 한다. 하지만 결과가 원하지 않는 오버세션화가 이루어질 수 있다. 따라서 본 논문에서는 오버세션화를 줄이기 위해 가우시안 혼합 모델(GMM) 기반의 변형된 워터쉐드 알고리즘을 제안한다. 거리 변환 영상은 가우시안 혼합 모델에 의해 처리된다. 그리고 기대 최대화(EM) 알고리즘은 가우시안 혼합모델 데이터에 맞추는 데 사용된다. 이들 평균 의 평균을 최적의 임계값으로 선택하고 획득된 임계값에 기초하여 이진 영상으로 변환한다. 이렇게 얻은 이진 영상 은 적절한 구조 요소의 적용하여 침식화한다. 침식된 영상에서 수행된 워터쉐드 분할은 효과적인 분할 결과를 제공 한다. 제안된 방법이 사용될 때 옥수수, 쌀 및 밀과 같은 다른 커널이 효과적으로 분할된다. 실험 목적으로 38 개의 옥수수 커널, 95 개의 쌀 커널 및 32 개의 밀 커널을 사용한다. 분할 정확도는 옥수수, 쌀 및 밀에 대해 각각 94.7 %, 96.8 % 및 90.6 % 이다.

Image segmentation is basically the division of an image into meaningful structures. It always produces complete division of the image as the specific demand. However, it is prone to over-segmentation which makes the result unfavorable. Therefore, this paper proposes modified watershed algorithm based on Gaussian mixture model(GMM) to reduce the over-segmentation. Distance transformed image is processed by GMM. And expectation maximization(EM) algorithm is used to fit GMM to data. Average of these means is chosen as optimal threshold and the image is converted into binary image based on the obtained threshold value. Binary image thus obtained is eroded with the help of proper structuring element. Watershed segmentation thus carried out on the eroded image gives the effective segmentation results. Different kernels such as corn, rice, and wheat are effectively segmented when the proposed methodology is used. 38 corn kernels, 95 rice kernels, and 32 wheat kernels are used for the experimental purpose. The accuracy of segmentation is as high as 94.7%, 96.8%, and 90.6% for corn, rice, and wheat respectively.

본 논문에서는 복부 CT 영상에서 딥러닝을 이용한 정확하고 빠른 근육 분할 기법을 제안한다. 데이터 세트는 DICOM 태그의 리스케일 슬로프, 인터셉트 정보를 이용해 픽셀 값을 정규화하고 훈련 데이터 세트에는 유사 변환 을 이용해 이미지 증강을 수행했다. 증강된 훈련 데이터 세트에 U-net을 이용해 근육 영역 분할 학습을 수행했다. U-net은 생물의학 이미지 분할을 위해 고안된 fully convolutional networks 기반 모델이다. 분할된 근육 영역 에서 hounsfield unit 범위로 판별된 근육 외의 부분을 제거하여 근육 마스크를 생성한다. 제안된 모델의 정확도는 98%이고, 한 장의 복부 CT 영상에서 근육을 분할하는 데 평균 1초가 소요되었다.

In this paper, we propose an accurate and fast fully automated muscle segmentation technique using deep learning in abdominal CT images. The dataset normalized pixel values using rescale slope, intercept information in the DICOM tag, and image augmentation was performed using similarity transformation in the training dataset. We performed muscle region segmentation learning using U-net on the augmented training dataset. U-net is a fully convolutional networks based model designed for biomedical image segmentation. A muscle mask is generated by removing a portion other than the muscle determined by the hounsfield unit range from the divided muscle region. The accuracy of the proposed model was 98% and it took an average of 1 second to segment the muscles in one abdominal CT image.

This paper presents analysis of segmentation and syntactic conversionstrategies observed in KE-EK SI, to answer 1) what syntactic makeupis found in the ST's segmented meaning unit, and 2) how the ST'ssyntactic makeup is converted into the TT. The study analyzed 559segments, marked in the data from 16 files of authentic speeches andtheir SI performances by four different professional interpreters in realconference settings. 559 segments were reviewed and categorizeddepending on their syntactic makeup and the non-predicate basedsegments were analyzed to see how they were delivered in the TT. Insegmentation, the study found that the syntactic makeup of 70% of thesegments were predicate-based, indicating that interpreters rely mostlyon predicate-based expressions when segmenting the ST. In conversion,the study found that 50% of the KE and 66% of the EK non-predicatebased segments were converted into predicate based expressions in theTT. In doing so, the interpreters were observed to reflect the syntacticmakeup of neighboring segments when processing one segment,indicating that SI is not entirely a strict linear process linking segmentto segment, but also a loop-like process between and within the sourceand target texts.

With the introduction of the new "National Law on Land Planning and Use", Urban Management Zone Segmentation became an important and essential tool in urban planning. The final results obtained from studies conducted on Urban Management Zone Segmentation are being utilized as basic data for city management planning, and thus must be finished before starting the actual urban management zone planning. This study aims to improve the standards currently used for Urban Management Zone Segmentation. Moreover, this research focuses on alternative method for Urban Management Zone Segmentation by using the case study of Daejeon Metropolitan City.Based on the study, it was determined that there are three issues of institutional improvement that should be considered : 1) First, it is necessary to prepare a program for improving the reliability of the preliminary data ; 2) next, an operating system for integrating Land suitability Assessment and Urban Management Zone Segmentation was found to be able to resolve the challenges of Land suitability Assessment ; and 3) finally, to manage the Urban Management Zone Segmentation smoothly, it is necessary to prepare the well defined standard of segmentation for 3rd class land.

본 논문에서는 사람 얼굴의 표정을 인식하기 위하여 판단 요소를 선정하고 판단 요소의 변화 상태를 파악 하여 표정을 인식하는 방법을 제안한다. 판단 요소를 선정하기 위하여 이미지 영역 분할 방법을 사용하며, 판단 요 소의 변화율을 이용하여 표정을 판단한다. 표정을 판단하기 위하여 90명의 표정을 데이터베이스화하여 비교하였고, 4 개의 표정(웃음, 화남, 짜증, 슬픔)을 인식하는 방법을 제안한다. 제안한 방법은 시뮬레이션을 실시하여 판단 요소 검 출 성공률과 표정 인식률을 통해 검증한다.

In this paper, we propose a method to recognize the facial expressions by selecting face elements and finding its status. The face elements are selected by using image area segmentation method and the facial expression is decided by using the normal distribution of the change rate of the face elements. In order to recognize the proper facial expression, we have built database of facial expressions of 90 people and propose a method to decide one of the four expressions (happy, anger, stress, and sad). The proposed method has been simulated and verified by face element detection rate and facial expressions recognition rate.

K-방산의 성공시대를 이어가기 위해서는 관련 법령의 정비와 외교·정책적 지원, 시스템적 지 원 등도 필요하지만, 여기에 마케팅 전략이 추가된다면 보다 큰 시너지 효과가 발휘될 것이다. 본 연구는 ‘K-방산’을 ‘STP 마케팅 전략’ 관점에서 방산수출 사례를 분석하여 우리에게 주는 시사점은 무엇인지 제시하는 데 목적이 있다. 방산수출의 특성과 대한민국이 방산수출에서 가지는 강점을 마케팅 전략으로 어떻게 투영 시킬 것인지에 대해 ‘STP 마케팅 전략’ 관점에서 ‘세분화(Segmentation) - 표적화 (Targeting) - 위치잡이(Positioning)’ 단계별로 ‘K-방산’의 성공과 실패 사례를 분석하여 시사점을 제시하였다. 세분화(Segmentation)에서는 방산수출 시장별 세분화 요인 정립과 마케팅 전략 전담팀 구성, 네트워크 구축 및 유지를 통하여 국가별 맞춤형 ‘K-방산’ 수출 전 략 수립의 필요성을 제시하였고, 표적화(Targeting)를 위해서는 방산 구조 고도화 추진과 무기체계 공동개발 등을 위한 무관 역할 증대를 제시하였다. 위치잡이(Positioning)에서는 ‘K-방산’ 브랜드화를 위해 ‘방위산업발전 기본계획’에 마케팅 전략 추가, 군(軍) 시범 운용제도 시행, 적정수준의 작전요구성능(ROC) 제시와 시험평가 기준 합리화를 제시하였다. 본 연구는 K-방산수출 관련 기업과 정부의 단편적 정보와 언론보도로 연구되었고, 관련된 선행연구가 없어서 연구의 한계점이 있으나, 마케팅 전략의 필요성과 시사점을 제시하는 수준에서 시사점을 도출하였다.

In order to continue the success era of the K-defense industry, it is necessary to maintain related laws, diplomatic and policy support, and system support, but if marketing strategies are added to this, greater synergy will be exerted. The purpose of this study is to analyze defense export cases from the perspective of STP marketing strategy of K-defense industry and to suggest implications for us. In terms of STP marketing strategy, we analyzed the success and failure cases of K-defense industry by segmentation - Targeting - Positioning stage and suggested implications for how to reflect the characteristics of defense industry exports and the strengths of Korea in defense industry exports. Segmentation suggested the necessity of establishing a customized K-defense export strategy for each country through the establishment of segmentation factors for each defense industry export market, the formation of a dedicated marketing strategy team, and the establishment and maintenance of a network. In order to target, I suggested an increase in the role of non-military for the promotion of defense industry structure and joint development of weapons systems. In order to make the K-defense industry brand, Positioning added marketing strategy to the defense industry development basic plan, implemented the military pilot operation system, presented the appropriate level of operational requirement performance and rationalized the test evaluation standard. This study was conducted with fragmentary information and media reports from companies and governments related to export of the K-defense industry, and although there are limitations in the study due to the lack of related previous studies, the implications were derived at the level of suggesting the necessity and implications of marketing strategies.

완공된 건축물에서 나타나는 대다수의 균열은 구조적, 설계적, 시공적 결함의 문제로 나타나며, 안전을 위해 균열에 대한 탐지와 점검이 이루어져야 한다. 특히 조적벽 건물의 균열은 일반적인 평평한 벽면에서 발생하는 균열과 다르게, 울퉁불퉁한 벽면에서 발생하는 조적벽 균열은 일반적인 점검을 어렵게 하는 요소들이 존재한다. 이는 많은 인력과 자금 및 시간이 있어야 하는 문제를 가지고 있다. 즉, 객관적인 점검 결과를 제시해줄 수 있고, 점검에 필요한 자원을 절약해줄 수 있는 연구가 필요하다. 초기에는 영상 처리(Image Processing)기술을 이용한 균열 탐지 방법이 연구되었으나, 노이즈에 취약하고, 영상의 형상 별로 별도의 필터를 모델링해야 한다. 최근에는 딥러닝(Deep Learning)을 활용한 방법으로 균열 탐지 연구가 활발히 진행되고 있다. 본 연구에서는 U-Net 딥러닝 모델을 이용하여 딥러닝 기반 조적벽 균열 탐지 모델을 제안하였다.

The research object of linguistics is the spoken characters, or the spoken language, which follow the principle of linearity. However, Chinese is an exception, with its characters as the second language. This is why Saussure concludes that Chinese is absolutely diverse. His second conclusion, then, is that Chinese is a language for critical thinking. This paper explores the Chinese language system, starting from segmenting the acoustic chain, where the issue of segmentation comes naturally. How can we segment Chinese characters? We find that the components of characters imply the way of segmentation, as is shown in the number of Seven and Eight. Why Seven and Eight? We attempt to find the reasons through two myths about the origin of Chinese character writing: He-tu and Luo-shu. This paper also studies the structure of symbols of Chinese characters with the support of the first Chinese dictionary, Shuowen Jiezi, and confirms Saussure's assertion of Chinese. The exploration of the segmentation approach also verifies Saussure's conclusion that linguistic symbols are a pure value system, and that the critical nature of Chinese is fully realized in the exploration of value systems. The Chinese language system and Saussure’s theory fit each other well.

L’objet de la linguistique, c’est le mot parlé, c’est la langue parlée. Le principe de la linéalité s’impose. Pourtant, le chinois est unique en son genre, son écriture est une seconde langue. Saussure le considère comme une diversité absolue et radicale, et il conclue que le chinois est de l’ordre spéculatif. Nous commençons par le découpage de la chaîne acoustique, il apparaît la manière de couper en chaîne acoustique. Le découpage en français ne montre pas la manière, or le chinois, son écriture est mise en évidence, qie fen comportent les nombre sept et huit. Pourquoi sept et huit,non d’autres nombres ? Nous cherchons sa raison d’être par les deux mythes, qui sont même l’origine de la création de l’écriture. L’examen se fait au recours à un dictionnaire, le premier de la Chine. En étudiant la structure des caractères et l’ensemble de ce système, l’auteur a découvert que, d’une part, la langue est un système des valeurs pures, d’autre part, la spécularité du chinois s’incarne dans l’examen de ce système unique en son genre. La pensée de Saussure et le chinois font ressortir les mérites l’un et l’autre.

The paper analyzes linguistic features of translation productsperformed on Computer-aided translation(CAT) tools, while comparing them with translations that are not translated with CATtools. This study aims to find linguistic grounds of review criterionfor CAT-based collaborative translations by selectively examiningreadability indicators. The analysis focuses on morphologicaldifferences on sentence(segment) level, employing quantitativemethodologies that are used to validate translation universals aswell as to analyze corpus stylistics and translation readability suchas sentence length, demonstrative pronoun and conjunctive adverb. The analysis result can be considered as review criteria whileevaluating and training collaborative translation.

본 논문은 공유형 전동킥보드 시스템 운영 시, 관리 상 발생할 수 있는 주차 문제를 해결하기 위해 반납 인증사진으로 주 차, 비주차 구역을 판단하는 AI모델을 제시한다. 본 연구에서는 주차/비주차 구역 배경 관련 객체를 판별하기 위해 ADE20K 에 Pre_trained된 Segfomer_b0 모델과 점자블록, 전동킥보드에 Fine_tuning한 Segfomer_b0 모델을 통해 주차/비주차에 관련 된 객체의 Segmentation map을 추출하고, Swin 모델을 통해 주차/비주차 구역을 이진 분류하는 방법을 제시하였다. 최종적으 로 총 1,689장을 직접 라벨링한 후 진행한 Fine_tuning SegFomer 모델은 mAP가 81.26% 수준으로 전동킥보드와 점자블록을 인식하였으며, 총 2,817장을 훈련한 Classification 모델은 92.11%의 정확도와 91.50%의 F1-Score로 주차구역과 비주차 구역을 분류하는 것이 가능하였다.

This paper proposes an AI model that determines parking and non-parking zones based on return authentication photos to address parking issues that may arise in shared electric scooter systems. In this study, we used a pre-trained Segformer_b0 model on ADE20K and fine-tuned it on tactile blocks and electric scooters to extract segmentation maps of objects related to parking and non-parking areas. We also presented a method to perform binary classification of parking and non-parking zones using the Swin model. Finally, after labeling a total of 1,689 images and fine-tuning the SegFomer model, it achieved an mAP of 81.26%, recognizing electric scooters and tactile blocks. The classification model, trained on a total of 2,817 images, achieved an accuracy of 92.11% and an F1-Score of 91.50% for classifying parking and non-parking areas.

차선이탈경고시스템(LDWS)은 첨단운전자지원시스템(ADAS)의 핵심 기능으로, 비의도적 차선 이탈을 탐지하여 사고를 예방한다. 그러나 기존의 에지 검출이나 허프 변환 기반 기법은 조명 변화, 차선 마모, 복잡한 도로 조건에서 한계가 있다. 본 연구는 딥러닝 기반 차선 세그멘테이션과 역원근 투영(IPM) 변환을 통합하여 LDWS 성능을 개선하는 방법을 제안한다. CamVid 데이터셋을 활용해 SegNet과 U-Net을 학습시켰으며, 결과를 탑뷰 영상으로 변환한 뒤 ResNet18 분류기를 통해 전처리 시나리오를 비교하였다. 실험 결과, U-Net이 SegNet보다 우수한 세그멘테이션 성능을 보였고, IPM 적용 시 분류 정확도가 크게 향상되었다. 특히 세그멘테이션 ROI와 IPM을 결합했을 때 최고 F1-score를 달성하여 균형 잡힌 검출 성능을 입증하였다. 따라서 제안 방법은 LDWS의 정확도와 강 인성을 향상시키는 효과적인 접근임을 확인하였다.

The Lane Departure Warning System (LDWS), a core function of Advanced Driver Assistance Systems (ADAS), prevents accidents by detecting unintended lane departures. Traditional methods based on edge detection or Hough transform are limited to illumination changes, lane fading, and complex roads. This study proposes integrating deep learning–based lane segmentation with Inverse Perspective Mapping (IPM) to improve LDWS performance. A SegNet and U-Net segmentation model was trained using the CamVid dataset, and the outputs were converted into top-view images, which were then evaluated with a ResNet18 classification model. The results show that U-Net achieved a higher level of segmentation accuracy, and IPM significantly improved classification accuracy. The combination of segmentation ROI and IPM yielded the best F1-score, confirming the effectiveness of the proposed approach for enhancing LDWS accuracy and robustness.

The Paris Agreement signatories have committed to limit global average temperature increase above pre-industrial levels to below 2℃. Reporting of the greenhouse gas (GHG) inventory is regulated by the United Nations Framework Convention on Climate Change. Currently, countries are transitioning from the Measurement, Reporting, and Verification reporting system to the Enhanced Transparency Framework (ETF) reporting system. Under the ETF, countries are required to use the 2006 guidelines (GL). This study explored how replacing the 1996 GL with the 2006 GL or the 2019 Refinement impacts the overall GHG inventory from the livestock sector, with Korea as a case study. The result affirmed that changes in guideline led to changes in total estimated emissions. Moving from the 1996 GL to the 2019 Refinement resulted in more significant differences in estimated emissions than moving to the 2006 GL in terms of source-based emissions, annual inventory, or trend. Notably, guidelines’ changes also impacted the proportion of each source’s contribution to total estimated emissions. While applying the most recent guidelines is expected to produce more accurate estimations, consistency with the previous inventory calculated with previously used guidelines should be maintained. Additionally, the changes in the contribution of each source clarifies that although enteric fermentation is the largest contributor of GHGs, relevant mitigations are likely less feasible compared to those related to manure management. This is because of naturally occurring biological processes. Thus, mitigations in manure management are suggested.

Gabor filters well support computer vision tasks due to its steerable properties, that are useful for handling image transformation such as? image rotation and image scaling. In this paper, the Gabor-based UNet (GSAUNet) model is proposed where the Gabor filters replace the few default filters in the UNet. Besides, it has adopted a squeeze-and-excitation module to increase channel interdependencies and attention mechanism to focus on an important part of the network. The performance of GSAUNet is assessed on the CamVid and Cityscape datasets that attain improvement for the semantic segmentation tasks.