MURAKKAB GEOMETRIC OBYEKTLAR SIRTINI TIKLASHDA 3D GAUSSIAN SPLATTING ASOSIDA GEOMETRIK ANIQLIKNI OSHIRISH USULLARI
Аннотация
3D Gaussian Splatting (3DGS) so‘nggi yillarda radiance field rekonstruksiyasi sohasida inqilob qilib, NeRFga nisbatan real vaqtda yuqori sifatli yangi ko‘rinishlarni sintez qilish imkonini berdi. Biroq, 3DGS ning asosiy kamchiliklaridan biri – Gauss primitivlari ko‘pincha “yumshoq bulut” holatida qolib, aniq sirt geometriyasini tiklashda muammolar tug‘diradi. Ayniqsa, murakkab me’moriy obyektlar (o‘ymakorlik, yupqa va ichi bo‘sh elementlar, nozik teksturali sirtlar) sirtini yuqori aniqlikda tiklashda ushbu kamchilik keskin namoyon bo‘ladi. Ushbu tezisda murakkab geometrik obyektlar sirtini yuqori aniqlikda tiklash uchun 3DGSga asoslangan takomillashtirilgan usul taklif etiladi. Taklif etilayotgan yechim uchta asosiy yangilikni o‘z ichiga oladi: (1) Multi-View Stereo (MVS) asosida Gauss primitivlarini boshlang‘ich joylashtirish – bu monokulyar chuqurlik baholashga nisbatan mustahkam geometrik priyorni ta’minlaydi; (2) gradient norma penaltisi asosidagi geometrik takomillashtirish regulyarizatsiyasi – bu yuqori chastotali tebranishlarni bostirib, yupqa va ichi bo‘sh strukturalarni saqlashga yordam beradi; (3) Gauss primitivlarini tekislangan skalali tekisliklarga aylantirish orqali mahalliy sirtni ifodalash sodiqligini oshirish.
Ключевые слова:
Библиографические ссылки
Mildenhall, B., Srinivasan, P. P., Tancik, M., Barron, J. T., Ramamoorthi, R., & Ng, R. (2021). NeRF: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1), 99-106.
Kerbl, B., Kopanas, G., Leimkühler, T., & Drettakis, G. (2023). 3D Gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4), 1-14.
Yang, Y., et al. (2025). Generalizable 3D Gaussian splatting via multi-view stereo and consistency constraints. Neurocomputing, 658, 131696.
SuGaR: Guédon, A., & Lepetit, V. (2024). SuGaR: Surface-aligned Gaussian splatting for efficient 3D mesh reconstruction and high-quality mesh rendering. arXiv preprint arXiv:2411.12345.
2DGS: Huang, B., Yu, Z., Chen, A., Geiger, A., & Gao, S. (2024). 2D Gaussian splatting for geometrically accurate radiance fields. arXiv preprint arXiv:2403.17888.
Hou, Y., Wang, T., & Wang, X. (2025). GAGS: Gradient-guided adaptive Gaussian splatting for efficient and geometry-regularized surface reconstruction. ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, X-1/W2-2025, 59.
География читателей
Загрузки
Опубликован
Выпуск
Раздел
Лицензия
Это произведение доступно по лицензии Creative Commons «Attribution» («Атрибуция») 4.0 Всемирная.
Как цитировать
