This project investigates human-centered interaction and user experience in XR environments, focusing on how immersive interfaces can provide safer, more comfortable, and engaging experiences. Our research includes adaptive XR interfaces for hazard awareness, gaze-responsive VR spectator views for reducing cybersickness, and multimodal interaction techniques for AR-based exercise and rehabilitation.

We further explore XR as an interactive platform for healthcare, training, and education, including virtual companions that enhance engagement and social presence during exercise and emotionally responsive virtual humans for clinical and nursing education. These studies aim to develop XR experiences that adapt to users' perceptual, physical, and emotional states.

Related Publications

A Color-Contrast-Based XR Interface Design Study: Focusing on AI-Driven Hazard Detection Scenarios | 2026 | International Conference on Human Interaction and Emerging Technologies 2026 (IHIET-AI) | Juhee Lee*, Sungnam Kim, Sunghee Ahn, and Jong-Il Park

Reducing Cybersickness for 2D VR Spectators Using a Gaze-Based Stabilized Third-Person View | 2026 | IEEE VR 2026 | Sungnam Kim* and Jong-Il Park

2D 관전자 환경에서의 시선 반응형 3인칭 VR 관전 시점 설계 | 2026 | 제 38 회 영상처리 및 이해에 관한 워크샵 (IPIU 2026) | 김성남*, 이윤상, 박종일

AR-Based Multi-sensory Human Interaction in Rehabilitation: Enhancing User Experience in Exercise | 2025 | HCII 2025 | Sangyun Lee* , Nahyeon Kong , Sunghee Ahn , and Jong-Il Park

AR Fitness Dog: The Effects of a User-Mimicking Interactive Virtual Pet on User Experience and Social Presence in Physical Exercise  | 2025 | IEEE VR 2025 | Hyeongil Nam*, Kisub Lee, Jong-Il Park, Kangsoo Kim

증강현실 속 대화형 가상 강아지를 활용한 피트니스 버디 시스템 개발 | 2024 | 2024년 한국방송미디어공학회 하계학술대회 | 이기섭*, 변우현, 김강수, 박종일

Virtual Reality Interaction Toward the Replacement of Real Clinical Nursing Education | 2022 | Lecture Notes in Computer Science | Chanhee Kim*, Hyeongil Nam, Ji-Young Yeo & Jong-Il Park

An Emotionally Responsive Virtual Parent for Pediatric Nursing Education: A Framework for Multimodal Momentary and Accumulated Interventions | 2022 | IEEE International Symposium on Mixed and Augmented Reality (ISMAR 2022) | Hyeongil Nam, Chanhee Kim, Kangsoo Kim, Ji-Young Yeo, Jong-Il Park

We develop an automated pipeline for converting architectural floorplans into structured 3D scenes. The system analyzes floorplans in a patch-wise manner, identifies structural elements such as walls, doors, and windows, and reconstructs their geometry through vectorization and geometric refinement. This enables reliable 3D reconstruction even for large-scale and complex architectural layouts.

The reconstructed models are further enriched with semantic information, including exterior/interior wall classification and automatic placement of doors and windows. Our recent work focuses on generating lightweight, game-ready 3D scenes that can be directly deployed in digital twins, VR/AR applications, and Unity/Unreal-based game environments.

Related Publications

F3GM: 경량 게임 배포를 위한 평면도 기반 3D Map 생성 | 2026 | 제38회 영상처리 및 이해에 관란 워크샵 (IPIU 2026) | 정주용*, 권태수, 박종일

Patch-Wise Analysis and Reconstruction of Large-Scale Floorplan for Digital Twin Modeling | 2025 | IEEE Access, Vol. 13, pp. 215031-215044 | Seunghan Paek*, Wonyoung Cho*, Juyong Jeong, Jong-Il Park

저용량 배포를 위한 평면도 기반 3D 씬 자동 생성 파이프라인 | 2025 | 한국방송·미디어공학회 추계학술대회 | 정주용*, 김성남, 백승한, 박종일

건축 Mesh 모델의 외벽·내벽 자동 분류 및 구조 요소 배치를 위한 파이프라인 |. 025 | 한국방송·미디어공학회 하계 학술대회 | 정주용*, 백승한, 조원영, 박종일

건축 평면도의 의미론적 분할을 위한 패치 기반 방법 | 2025 | 제37회 영상처리 및 이해에 관한 워크샵 (IPIU 2025) | 조원영*, 백승한, 이상윤, 박종일

건축 평면도 3D 변환을 위한 세선화 및 벡터화 알고리즘 | 2024 | 한국방송·미디어공학회 추계 학술대회 | 하승우*, 박종일

학습 기반 방법을 활용한 건축 평면도의 구성 요소 추출 | 2024 | 한국방송·미디어공학회 하계 학술대회 | 조원영*, 백승한, 박종일

건축 평면도 3D 모델링을 위한 패턴 기방 교점 검출 | 2024 | 제36회 영상처리 및 이해에 관한 워크샵 (IPIU 2024) | 하승우*, 백승한, 박종일

규칙 기반 레이블 부여 알고리즘을 이용한 건축 평면도 세선화 | 2024 | 한국방송·미디어공학회 하계학술대회 | 하승우*, 박종일

This project develops 3D Scene Graph representations for structured and efficient scene understanding. Our research extracts objects and semantic information from 3D scenes and represents their spatial and relational context as graphs. Recent work extends this direction through segmentation-aware object extraction and open-vocabulary relational scene graphs, enabling more flexible understanding of previously unseen objects and relationships.

We also investigate hierarchical and topology-based scene representations to reduce the complexity of large-scale environments. By organizing conceptual, spatial, and topological information at multiple levels, the resulting scene graphs can support efficient scene retrieval, localization, loop-closure detection, and various spatial computing applications.

Related Publications

SAGE: Segmentation-Aware 3D Object Extraction from Single Imaegs | 2026 | International Workshop on Advanced Image Technology (IWAIT 2026) | Juyong Jeong*, Sung-rok Kwon*, Hajeong Lee, Jong-Il Park

Open-vocabulary Relational Scene Graph Generation for Large-scale Scene | 2026 | 제 38 회 영상처리 및 이해에 관한 워크샵 (IPIU 2026) | 이상윤*, 박종일

효율적인 장면 이해를 위한 개념 정보 기반 다단계 장면 표현 방법 | 2025 | 2025년 한국방송미디어공학회 하계학술대회 | 변우현*, 이상윤, 박종일

루프클로저 탐지를 위한 의미론적 토폴로지 그래프 | 2024 | 2024년 한국방송미디어공학회 추계학술대회 | 변우현*, 박종일

This project develops a Visual Positioning System (VPS) for estimating camera pose in environments where GPS is unreliable. Our early work, SkyPose, estimates camera pose in mountainous and mining areas by matching image skylines with 3D terrain geometry, and extends this positioning result to AR-based visualization of mining terrain information.

We further expand this approach to large-scale positioning by combining street-view panoramas, satellite-based BEV imagery, and skylines generated from 3D models. The goal is to build a robust visual positioning framework that can operate across both urban and non-urban environments.

Related Publications

SkyPose: Real-time camera pose estimation by skyline matching in mountainous terrain | 2026 | International Workshop on Advanced Image Technology (IWAIT 2026) |Sangyun Lee*, Juyong Jeong*, Jong-Il Park

증강현실을 이용한 광산지형정보 3차원 가시화 기술 | 2026 | 방송공학회 논문지 제31권 제 1호, 138-151 | 이상윤*, 정주용*, 김형민, 김수로, 박종일

GPS 의존도를 완화한 시각 정보 기반 모바일 맨홀 ID 식별 시스템 | 2026 | 제38회 영상처리 및 이해에 관한 워크샵 (IPIU 2026) | 권성록*, 박종일

모바일 환경에서의 인스턴스 분할 및 위치정보 기반 맨홀 식별 시스템 | 2025 | 2025 한국방송·미디어공학회 하계 학술대회 | 권성록*, 백승한, 이승훈, 박종일

This project investigates 3D reconstruction techniques for recovering accurate geometry from diverse visual inputs and challenging imaging conditions. Our research covers stereo endoscopic reconstruction, NeRF-based object reconstruction under occlusion, floorplan-to-3D modeling, and reconstruction under refractive environments, extending 3D modeling to medical, indoor, and industrial applications.

Recent work further explores Gaussian Splatting-based surface reconstruction, robust image capture strategies for indoor reconstruction, and RGB-D-based modeling of underground structures. These studies aim to improve reconstruction quality and stability while adapting 3D reconstruction pipelines to different sensors, scene scales, and real-world conditions.

Related Publications

실내 3차원 재구성 안정성에 대한 입력 촬영 조건 분석 | 2026 | 2026년 한국방송미디어공학회 하계학술대회 | 최낙민*, 정주용, 백승한, 박종일

표면 재구성을 위한 적응형 차원 학습 가우시안 스플래팅 | 2026 | 제 38 회 영상처리 및 이해에 관한 워크샵 (IPIU 2026) | 이하정*, 박종일

3D Modeling for Slope Angles and Volume of Undereground Conduits Using RGB-D Camera | 2025 | IEEE Access | Seunghan Paek*, SangHwa Lee, SangHyun Joo and Jong-Il Park

Patch-wise Analysis and Reconstruction of Large-scale Floorplan for Digital Twin Modeling | 2025 | IEEE Access | Seunghan Paek*, Wonyoung Cho*, Juyong Jeong and Jong-Il Park

3D Voxel Reconstruction from Occlusion Image Using NeRF-Based Method | 2023 | The 14th International Conference on 3D Systems and Applications (3DSA 2023) | Minsu Kyeon*, Hyeongil Nam, Jong-Il Park

Stereo Endoscope-Based 3D Surface Reconstruction for Image-Guided System | 2013 | Proceedings of Asian Conference on Computer Aided Surgery (ACCAS’13) | Junyeong Choi*, Byung-Kuk Seo, Hanhoon Park, and Jong-Il Park

We develop AI-assisted AR systems for transforming clinical records into intuitive, patient-specific risk visualizations. The pipeline extracts structured information from physical medical documents, predicts disease progression using competing-risk models, and maps the results into interactive time-spatial AR narratives.

Our recent work focuses on combining document understanding, prognostic modeling, and grounded interaction to improve patient understanding and clinical risk communication.