Rapid, real-time detection of anomalies and locate structural defects during earthquakes is critical for ensuring safety and enabling timely decision-making. Although deep learning-based structural health monitoring (SHM) has shown considerable promise, conventional supervised models are often impractical because labeled damage data from real-world structures are extremely scarce. To address this challenge, this paper proposes a Multi-Class Deep Support Vector Data Description (SVDD) framework for structural defect detection. The proposed Multi-Class Deep SVDD approach learns the boundary of normal data using only normal seismic acceleration responses. When new data are recorded, the system infers both the occurrence and location of defects by evaluating whether the responses fall within or deviate from the learned normal boundary. The framework is validated using the Los Alamos National Laboratory 3-story bookshelf structure benchmark dataset. Experimental results show that the proposed model achieves a peak average accuracy of 87.12% in a 4-dimensional latent space, substantially outperforming traditional baseline methods, including Kernel Density Estimation (KDE), SVDD, and One-Class Deep SVDD. These findings indicate that the Multi-Class Deep SVDD framework provides a robust and objective metric for rapid post-earthquake safety assessment without requiring prior exposure to faulty datasets.

This study presents the development of an AI-based real-time on-device segmentation system designed to support recyclable waste sorting. A lightweight semantic segmentation model was implemented by combining the MobileViT-x-small backbone with the DeepLabV3 architecture, enabling pixel-level classification of recyclable items and intuitive visualization on a smartphone screen. A total of 200 real-world images were collected, with 150 used for training and 50 for testing. To enhance generalization under limited data conditions, the training set was expanded to 750 images through geometric and color-based augmentation techniques. The trained model was subsequently converted into ONNX format and deployed within a Flutter-based mobile application, allowing real-time inference directly on the device without reliance on external servers. The proposed system overlays semi-transparent masks and class labels onto the live camera feed, thereby reducing sorting errors and promoting active user participation in everyday recycling practices.

Background: Real-time ergonomic risk assessment in manufacturing environments is challenged by severe class imbalance in high-risk postures and the need for deployment-efficient models. Conventional oversampling techniques may violate biomechanical constraints, limiting their suitability for human motion data. Objectives: This study aimed to compare multiple machine learning models for real-time ergonomic risk assessment while addressing data imbalance using biomechanically appropriate learning strategies and evaluating both predictive performance and deployment efficiency. Design: Comparative study. Methods: A large-scale workplace safety dataset comprising image-based skeletal keypoints was analyzed. To mitigate class imbalance without generating biomechanically implausible samples, cost-sensitive learning and focal loss were employed instead of synthetic oversampling. Subject-wise data splitting was applied to prevent data leakage. Five model families, including Random Forest, convolutional neural networks, and a lightweight graphbased network, were evaluated using accuracy, F1-score, area under the receiver operating characteristic curve (AUC), and high-risk recall. Statistical significance was assessed using bootstrap confidence intervals and McNemar and DeLong tests. Results: The lightweight graph-based model demonstrated competitive classification performance while maintaining reduced computational complexity. Although none of the models achieved the predefined high-risk recall threshold, statistically significant performance differences were observed across model families. Conclusion: The findings suggest that biomechanically informed imbalance handling improves methodological validity in ergonomic risk assessment. While deployment feasibility appears promising, further empirical validation on edge hardware is required.

목적: 본 연구의 목적은 고화질로 인쇄된 사진 형태의 모형안을 이용하여 실시간 영상 기반 안구운동 측정 장비를 개발하고, 반복 측정 실험을 통해 동공 중심 검출 알고리즘의 안정성과 신뢰도를 정량적으로 평가하는 것이다. 기존 상용 eye-tracking 시스템에 비해 저비용 하드웨어와 오픈소스 소프트웨어만으로 구축 가능한 장비의 초기 성능을 검증하고자 하였다. 방법 : XIMEA 고속 카메라를 기반으로 적외선 조명 및 실시간 영상 처리 알고리즘을 구성하여 동공 영역을 검 출하고 중심 좌표(x, y)를 추적하였다. 모형안을 고정된 거리에서 촬영한 후, 동일한 환경에서 10회 반복 측정을 수행하였다. 각 반복 측정은 900프레임으로 구성되었으며, 총 9,000프레임의 동공 영상 데이터를 수집하였다. 동 공 중심 검출 성공률을 산출하였으며, 반복 측정 간 중심 좌표의 변동성을 표준편차로 정량화하여 알고리즘의 안정 성을 평가하였다. 결과 : 총 9,000프레임 중 동공 중심 검출 성공률은 평균 97.3%를 나타냈다. 반복 측정 간 중심 좌표의 변동성 은 x축 표준편차 0.46±0.05 pixel, y축 표준편차 0.52±0.04 pixel로 측정되었으며, 모든 조건에서 중심 좌표의 표준편차가 1 pixel 미만을 유지하였다. 시간 분포 시계열 분석 결과, 중심 좌표는 특정 방향으로의 점진적인 위치 편향이 거의 관찰되지 않았으며, 중심 주변에 밀집된 분포를 보였다. 결론 : 본 연구에서 개발한 실시간 안구운동 측정 장비는 모형안 기반 반복 측정 실험에서 높은 동공 검출 성공 률과 우수한 반복 측정 안정성을 보여주었다. 저비용 장비 구성과 자유로운 알고리즘 수정 가능성은 연구 단계의 eye-tracking 시스템 개발에 유리한 장점을 제공하며, 향후 사람 대상 연구 이전의 초기 장비 검증 모델로 활용 가능하다. 또한 동공 중심뿐만 아니라 동공 지름 변화, 홱보기 검출 등 다양한 시기능 분석 지표로 확장할 수 있는 기술적 기반을 마련하였다.

This study proposes a lightweight algorithm for real-time front-vehicle detection using low-resolution camera footage under various driving conditions. The proposed method first extracts driving lanes using Canny edge detection and the Hough transform, thus enabling efficient lane detection. A forward region of interest (ROI) is delineated based on the extracted lane geometry. Subsequently, YOLOv11 is employed to detect vehicles within each frame, where only those located inside the defined ROI are classified as preceding vehicles. To evaluate the applicability of the proposed method in diverse environments, its performance was assessed across six driving scenarios: normal driving, traffic congestion, complex structural environments, nighttime, tunnel sections, and sharp curves. Experimental results show that the proposed approach maintains a stable detection accuracy across different conditions while offering a low computational cost and a high processing speed. Compared with segmentation-based deep-learning lane-detection models, the proposed method demonstrates superior real-time capability and can operate using only a built-in monocular camera without relying on expensive sensors such as LiDAR, radar, or artificial markers. This study serves as a foundation for vision-based ADASs, front-vehicle-following control, and road-hazard detection systems.

This research presents a single-walled carbon nanotube (SWCNT)-enabled real-time monitoring system to optimize postcuring conditions (temperature and duration) for epoxy resin. This method can serve as an alternative to traditional methods like Differential Scanning Calorimetry (DSC), which is effective in measuring the degree of cure in polymers during industrial curing (manufacturer-recommended cure cycle). Two different programs using SWCNTs were employed to design the cure cycles for investigating the development of mechanical properties: Program A as the comparison of effects of varied duration of high-temperature curing and Program B as high-temperature curing followed by the varied duration of low-temperature post-curing. By correlating variation in the electrical resistance of SWCNT with curing stages, we illustrate that extending post-curing at 100 °C for 24 h after an initial 3-h cure at 130 °C increases (i) tensile strength by 60% and ultimate tensile elongation by 101% and (ii) shear strength by 14% and ultimate shear elongation by 16% compared to industry standards. This approach not only improves mechanical performance but also enables precise, non-destructive cure-state detection, offering a scalable solution for high-performance composites in the aerospace and automotive sectors.

This study aims to explore the creative and technological significance of applying real-time motion capture data to XR (Extended Reality)-based multidisciplinary performances. By analyzing the case of the performance All About Error, which integrated real-time captured movements of dancers with audiovisual content delivered on a media wall (LED screen), the research investigates both the potential and the limitations of creating nonlinear, interactive stage environments. The methodology combines a review of prior XR production cases with an in-depth analysis of the actual production process of the performance. The findings demonstrate that the convergence of technology and art in multidisciplinary performances advances beyond traditional unidirectional and linear formats, fostering bidirectional and multidimensional performances that respond in real time to the performers’ movements. This evolution promotes expanded visual communication and discourse between performers and audiences, illustrating the creative potential to redefine the boundaries of live art. Utilizing real-time motion data on stage not only maximizes audience immersion and active participation but also suggests that real-time, interactive technologies in digital media art can expand into a variety of fields, including performing arts and games. This trend points to new directions and growth opportunities in artistic creation and provides important implications for future research in performing arts and interactive media art.

This study proposes a real-time content design pipeline optimized for Unreal Engine, integrating generative AI-based image creation with AI-assisted 3D modeling tools. The pipeline aims to streamline the production of high-quality assets for real-time applications, including games and simulations. Two types of subjects were selected: a bust combining organic character features, and a stone slab characterized by planar and symmetrical structure. Multi-angle image data were first synthesized using advanced generative AI models to simulate diverse viewpoints. These were then processed using AI-enhanced photogrammetry and modeling tools to reconstruct detailed 3D meshes and extract base textures. Post-processing steps, including mesh decimation, UV unwrapping, and texture baking, were performed to ensure compatibility with Physically Based Rendering (PBR) workflows used in Unreal Engine. The final assets were successfully imported into Unreal Engine, demonstrating visual fidelity and performance suitability in a real-time environment. The study confirms the pipeline’s potential for accelerating asset development and suggests promising future directions in AI-driven digital content creation.

해상운송에서 충돌사고는 인명과 재산에 막대한 피해를 끼치는 중요한 안전 문제로써, 국제해사기구(International Maritime Organization, IMO)가 제정한 국제해상충돌예방 규칙(International Regulations for Preventing Collisions at Sea, COLREGs)의 철저한 준수가 권장 된다. 그러나 복잡한 해상 환경과 인간의 인지적 한계로 인해, 항해사가 실시간으로 최적의 충돌회피 의사결정을 내리기란 쉽지 않다. 본 연구는 대형 언어 모델(Large Language Model, LLM)인 GPT를 활용하여, 유인선 항해사가 COLREGs 규칙에 부합하는 충돌회피 판단을 신속 하고 정확하게 내릴 수 있도록 지원하는 단계별 프롬프트 설계안을 제시한다. 특히 4단계 충돌회피 과정을 확장하여, 항해사가 GPT와 자 연어로 상호작용할 때 사용할 표준화된 프롬프트를 구체화하였다. 가상의 시나리오 적용 결과, 항해사는 GPT의 조언을 통해 주변 상황 인식부터 회피경로 선정, 실행 단계까지 일관적으로 보고받을 수 있었으며, COLREGs의 준수와 충돌위험지수(Collision Risk Index, CRI) 계 산 등의 복잡한 작업을 AI가 보완함으로써 인적 오류를 줄일 가능성을 보였다. 이러한 결과는 자율운항선 뿐만 아니라 현행 유인선 운항 에서도 AI-항해사 협업을 통한 안전성 향상을 도모할 수 있을 것으로 기대한다.

식중독균은 식품의 생산, 가공 및 유통 과정에서 확산 될 수 있으며, 이는 대규모 식중독 사고로 이어질 수 있 다. ‘Farm to table’ 전 과정에서의 식품 안전을 확보하기 위해서는 신속하고 정확한 검출 기술이 필수적이다. 그러 나 기존 PCR 시스템은 실험실 환경에 제한되어 있어 현 장 적용이 어렵다. 이를 해결하기 위해 현장형 PCR 기기 가 개발되었으며, 마이크로유체칩(microfluidic chip)은 고속 처리, 비용 효율성 및 다중 검출 기능을 갖춘 기술로 주목 받고 있다. 특히, 반응 구획이 분리된 다중 반응 챔버를 활용하면 여러 병원체를 동시에 검출할 수 있다. 본 연구에 서는 현장형 실시간 PCR 장비와 마이크로유체칩을 통합한 Lab-on-a-chip 시스템을 개발하고, 이를 이용한 식중독균의 신속한 현장 검출법을 검증하였다. 본 시스템은 swab 분석 을 이용한 DNA 추출법과 현장형 실시간 PCR을 결합하여 E. coli O157:H7, Salmonella spp., L. monocytogenes, S. aureus의 DNA를 식품 및 환경 시료에서 효과적으로 추출 하고 분석할 수 있었다. GENECHECKER® UF-300 실시간 PCR 시스템을 활용한 검출 결과, 30분 이내에 105-101 CFU/ mL (cm2) 수준의 검출 한계를 나타내며, 신속하고 민감한 다중 병원체 검출이 가능함을 확인하였다. 본 연구 결과 는 마이크로유체칩을 활용한 현장형 실시간 PCR 시스템 이 식품 안전 모니터링 및 현장 진단에 효과적으로 활용 될 가능성을 보여준다. 현장형 다중 검출 시스템을 통해 식중독균을 보다 신속하게 검출할 수 있어, 식중독 예방 및 감시 체계에 중요한 역할을 할 것으로 기대된다.