Pemodelan 3D Candi Banyunibo Berbasis Integrasi Terrestrial Laser Scanner dan Fotogrametri UAV untuk Dokumentasi Digital

Farizi Hibatul Hakim, Ir. Rochmad Muryamto, M.Eng.Sc.

2026 | Tugas Akhir | D4 TEKNOLOGI SURVEI DAN PEMETAAN DASAR

Candi Banyunibo merupakan salah satu cagar budaya di Daerah Istimewa Yogyakarta yang terletak di kawasan rawan gempa akibat aktivitas Sesar Opak. Pengalaman pascagempa Yogyakarta 2006 menunjukkan bahwa ketiadaan dokumentasi geometrik eksisting menyulitkan proses pemugaran bangunan cagar budaya. Kondisi serupa berpotensi terjadi pada Candi Banyunibo mengingat belum tersedianya dokumentasi geometrik digital yang detail dan terukur. Dokumentasi digital as-built yang akurat diperlukan sebagai arsip spasial jangka panjang guna mengantisipasi risiko kehilangan data spasial. Proyek akhir ini merekonstruksi model 3D parametrik Candi Banyunibo berbasis fusi data Terrestrial Laser Scanner (TLS) dan fotogrametri Unmanned Aerial Vehicle (UAV). Model dikembangkan pada klasifikasi Level of Detail (LOD) 4, diikatkan pada Sistem Referensi Geospasial Indonesia (SRGI) 2013, serta divalidasi ketelitian posisi dan dimensinya.

Akuisisi data dilaksanakan di Kompleks Candi Banyunibo menggunakan instrumen TLS Leica RTC360 pada 19 stasiun pengamatan dan UAV DJI Mavic Mini yang menghasilkan 302 foto udara. Pengukuran titik kontrol geodesi menggunakan GNSS Rapid Static untuk 4 titik Ground Control Point (GCP) dan RTK Radio untuk 30 titik uji independen. Kedua metode diikatkan pada stasiun Continuously Operating Reference Station (CORS) Badan Informasi Geospasial (BIG). Data TLS dan UAV diintegrasikan menggunakan algoritma Iterative Closest Point di CloudCompare. Rekonstruksi mesh dan pemetaan tekstur hibrida dilakukan di Agisoft Metashape, sedangkan pemodelan parametrik as-built dikerjakan di Autodesk Revit. Validasi kualitas model mencakup analisis Cloud-to-Cloud (C2C), Multiscale Model-to-Model Cloud Comparison (M3C2), dan Cloud-to-Mesh (C2M), serta uji akurasi dimensi linier dengan parameter RMSE, dan uji-t berpasangan. Visualisasi akhir model dipublikasikan melalui platform Cesium Ion.

Perataan jaring GCP menghasilkan standard error horizontal sebesar 0,0028 m dan vertikal sebesar 0,0048 m. Titik uji independen menghasilkan ketelitian horizontal sebesar 0,0282 m dan vertikal sebesar 0,0295 m. Pengujian akurasi posisi absolut model mesh menghasilkan RMSE tiga dimensi sebesar 1,46 cm, yang berada dalam batas ketidakpastian titik uji independen sebesar ±3 cm berdasarkan standar ASPRS/NSSDA. Pengujian akurasi dimensi linier pada 43 pasangan data menghasilkan rata-rata besaran perbedaan sebesar 2,5 mm dan RMSE sebesar 3,4 mm. Nilai RMSE tersebut memenuhi ambang batas Level of Accuracy (LOA) 40 USIBD, yaitu kurang dari 5 mm, serta ambang ketelitian LOD 4. Uji-t berpasangan menghasilkan t-hitung sebesar 1,0966, lebih kecil daripada t-tabel sebesar 2,0181. Dengan demikian, H0 gagal ditolak dan belum terdapat bukti statistik yang cukup untuk menyatakan bahwa rata-rata ukuran dimensi model berbeda secara signifikan dari hasil pengukuran lapangan. Analisis C2M menghasilkan rata-rata simpangan sebesar 2 mm, yang menunjukkan bahwa rata-rata perbedaan permukaan model relatif kecil.

Banyunibo Temple is one of the cultural heritage sites in the Special Region of Yogyakarta located in an earthquake-prone area due to the Opak Fault activity. The 2006 Yogyakarta earthquake demonstrated that the absence of geometric documentation significantly hampered heritage restoration efforts. A similar risk exists for Banyunibo Temple, as detailed and measurable digital geometric documentation has not yet been established. Accurate digital as-built documentation is therefore necessary as a long-term spatial archive to mitigate the risk of future spatial data loss. This final project reconstructs a parametric 3D model of Banyunibo Temple based on the fusion of Terrestrial Laser Scanner (TLS) and Unmanned Aerial Vehicle (UAV) photogrammetry data. The model was developed at Level of Detail (LOD) 4 classification, referenced to the Indonesian Geospatial Reference System (SRGI) 2013, and validated for positional and dimensional accuracy.

Data acquisition was conducted using a Leica RTC360 TLS instrument at 19 scanning stations and a DJI Mavic Mini UAV producing 302 aerial images. Geodetic control point measurements used GNSS Rapid Static for 4 Ground Control Points (GCPs) and RTK Radio for 30 independent check points. Both methods were referenced to the Continuously Operating Reference Station (CORS) network of the Badan Informasi Geospasial (BIG). TLS and UAV data were integrated using the Iterative Closest Point algorithm in CloudCompare. Mesh reconstruction and hybrid texture mapping were performed in Agisoft Metashape, while parametric as-built modelling was carried out in Autodesk Revit. Model quality validation included Cloudto-Cloud (C2C), Multiscale Model-to-Model Cloud Comparison (M3C2), and Cloudto-Mesh (C2M) analyses, as well as linear dimensional accuracy assessment using a distometer with RMSE, and paired t-test parameters. The final model was published through the Cesium Ion platform.

The GCP network adjustment yielded horizontal and vertical standard errors of 0.0028 m and 0.0047 m, respectively. The independent check points produced horizontal and vertical precisions of 0.0282 m and 0.0295 m, respectively. The absolute positional accuracy assessment of the mesh model produced a threedimensional RMSE of 1.46 cm, which was within the ±3 cm uncertainty of the independent check points based on the ASPRS/NSSDA standards. The linear dimensional accuracy assessment of 43 paired measurements produced a mean difference magnitude of 2.5 mm and an RMSE of 3.4 mm. The RMSE satisfied the USIBD Level of Accuracy (LOA) 40 threshold of less than 5 mm and the LOD 4 accuracy requirement. The paired t-test produced a calculated t-value of 1.0966, which was lower than the critical t-value of 2.0181. Therefore, the null hypothesis was not rejected, and there was insufficient statistical evidence to conclude that the mean model dimensions differed significantly from the field measurements. The C2M analysis produced a mean deviation of 2 mm, indicating a relatively small average surface difference. 

Kata Kunci : terrestrial laser scanner, fotogrametri UAV, fusi spasial, model 3D parametrik, dokumentasi as-built, cagar budaya, level of accuracy