3D 공간 재구축을 위한 환경별 최적 카메라 선택 기준 연구

초록

Recent advancements in 3D Gaussian Splatting technology have enabled the reconstruction of spaces with photographic precision and realism. However, the process of generating 3D models still relies heavily on image-based learning, which poses limitations in diverse environmental conditions. To address this, this study analyzes the performance of various cameras for 3D spatial reconstruction and presents optimal camera selection criteria for different environments. In this study, mirrorless cameras, smartphone cameras, and omnidirectional cameras were utilized to conduct 3D Gaussian Splatting-based reconstruction experiments in both indoor and outdoor environments. Performance evaluation was carried out quantitatively using PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index Measure) metrics, along with qualitative analysis through visual comparisons. The experimental results showed that mirrorless cameras provided high resolution and detailed rendering, demonstrating stable performance in both indoor and outdoor environments. Smartphone cameras excelled in accessibility and ease of use but were sensitive to lighting variations and exhibited reduced detail reproduction in long-distance shots. Omnidirectional cameras offered a wide field of view, effectively capturing the overall spatial composition, but suffered from image distortion and resolution degradation, particularly in indoor settings. Based on these findings, this study establishes optimal equipment selection criteria for 3D spatial reconstruction in different environments. These insights are expected to facilitate more precise and optimized spatial reconstruction processes in future applications such as 3D photogrammetry and virtual reality (VR) content creation.

키워드