Ubaid, Muhammad Jawadil (2026) Analisis Pemodelan 3D Struktur Eksterior Tower 3 ITS Berbasis Data Aerial Laser Scanner Menggunakan Realityscan dan Cloudcompare Terhadap As-Built Drawing. Other thesis, Institut Teknologi Sepuluh Nopember.
![[thumbnail of 5016221091-Undergraduate_Thesis.pdf]](http://repository.its.ac.id/style/images/fileicons/text.png) |
Text
5016221091-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (14MB) | Request a copy |
Abstract
RealityScan dan CloudCompare merupakan dua perangkat lunak pengolahan point cloud yang digunakan dalam pemodelan tiga dimensi bangunan berbasis data Aerial Laser Scanner (ALS). Penelitian ini membandingkan kualitas pemodelan 3D struktur eksterior Tower 3 ITS dari kedua perangkat lunak melalui analisis Cloud-to-Cloud Distance, Surface Density, durasi pemrosesan, kualitas model 3D, pencapaian Level of Detail (LoD) 3, serta akurasi geometrik dan kualitas visual terhadap as-built drawing. Akuisisi data menggunakan DJI Matrice 350 RTK dengan sensor LiDAR Zenmuse L2 melalui tiga flight plan (oblique 45°, oblique 60°, dan manual vertikal), menghasilkan 209.927.308 titik point cloud. Model 3D dibangun dengan Autodesk Revit metode scan-to-BIM dan dievaluasi berdasarkan LoD 3 OGC CityGML 2012. CloudCompare menghasilkan 143.314.392 titik dengan kerapatan 1.774–258.297 pts/m² dan durasi impor 40 detik, sedangkan RealityScan hanya 2.686.658 titik dengan kerapatan 1.174–11.036 pts/m² dan durasi impor 12 menit 20 detik akibat reduksi otomatis 98,13%. Cloud-to-Cloud Distance berkisar 0,000–3,250 m, terbesar pada balok lantai 4–10 dan kisi-kisi sisi barat. CloudCompare merepresentasikan elemen eksterior lebih detail dibanding RealityScan yang kehilangan detail akibat reduksi otomatis. Model CloudCompare memenuhi seluruh kriteria LoD 3 dengan RMSE rata-rata 0,032 m, sedangkan RealityScan mendekati LoD 3 dengan RMSE 0,413 m karena koordinat X melampaui toleransi 0,5 m. Evaluasi confusion matrix menunjukkan CloudCompare merekonstruksi seluruh 71 objek referensi (recall dan F1-Score 100%), sedangkan RealityScan hanya 60 objek (recall 84,72%, F1-Score 91,73%). Pengujian akurasi geometrik dengan pengukuran langsung menggunakan total station dan roll meter serta pengukuran di masing-masing perangkat lunak menunjukkan selisih RealityScan −0,255 m hingga +0,137 m dan CloudCompare -0,021 m hingga +0,120 m, dengan validasi roll meter -0,325 m hingga +0,224 m. Seluruh hasil memenuhi toleransi SNI 2847:2019, sehingga CloudCompare menghasilkan akurasi geometrik dan kualitas visual lebih baik sebagai rujukan pemilihan perangkat lunak pengolah data LiDAR untuk pemeliharaan bangunan tinggi.
================================================================================================================================
RealityScan and CloudCompare are point cloud processing software packages used for three-dimensional (3D) building modeling based on Aerial Laser Scanner (ALS) data. This study compares the quality of 3D modeling of the exterior structure of Tower 3 ITS using both software through Cloud-to-Cloud Distance, Surface Density, processing time, 3D model quality, Level of Detail (LoD) 3 achievement, and geometric accuracy and visual quality against as-built drawings. Data acquisition was conducted using a DJI Matrice 350 RTK equipped with a Zenmuse L2 LiDAR sensor through three flight plans (45° oblique, 60° oblique, and manual vertical flights), producing 209,927,308 point cloud points. The 3D models were developed in Autodesk Revit using the scan-to-BIM method and evaluated based on the LoD 3 criteria of OGC CityGML 2012. CloudCompare produced 143,314,392 points with a density of 1,774–258,297 pts/m² and an import duration of 40 seconds, while RealityScan generated 2,686,658 points with a density of 1,174–11,036 pts/m² and required 12 minutes 20 seconds due to an automatic reduction of 98.13%. The Cloud-to-Cloud Distance ranged from 0.000 to 3.250 m, with the largest differences observed on the floor 4–10 beams and west-side façade grilles. CloudCompare represented exterior elements more completely and in greater detail than RealityScan. The CloudCompare model satisfied all LoD 3 requirements with an average RMSE of 0.032 m, whereas RealityScan achieved an average RMSE of 0.413 m. Confusion matrix evaluation showed that CloudCompare reconstructed all 71 reference objects (Recall and F1-Score of 100%), while RealityScan reconstructed 60 objects (Recall of 84.51% and F1-Score of 91,73%). Geometric accuracy assessment indicated deviations of −0.021 m to +0.120 m for CloudCompare and −0.255 m to +0.137 m for RealityScan. All results satisfied the tolerance requirements of SNI 2847:2019. Therefore, CloudCompare demonstrated superior geometric accuracy, object completeness, and visual quality for high-rise building maintenance applications.
Actions (login required)
 |
View Item |