이미지 처리 및 영상 분석 기법을 통한 건축 모델링 데이터 정량적 평가 기법에 대한 연구

다국어 초록 (Multilingual Abstract)

This paper proposes an improved framework for evaluating large-scale architectural 3D models by addressing the limitations of traditional methods, which often demand high computational loads or rely on qualitative assessments. To overcome these challenges, the study utilizes normalized cross correlation(NCC) and structural similarity index(SSIM) for quantitatively assessing texture alignment and visual quality. Results show that photogrammetry and multi-view stereo(MVS) modeling excelled in visual quality through SSIM and NCC scores. Additionally, chamfer distance and hausdorff distance were used to evaluate geometric accuracy, revealing that both MVS and photogrammetry effectively reflected structural similarity. These findings highlight that NCC and SSIM are valuable for visual consistency, while chamfer and hausdorff distances effectively address geometric accuracy, offering improvements over conventional methods.