Precise 3d Modelling Using Combination Depth Maps Processing and Ground Control Point

Abstract

The use of Unmanned Aerial Vehicles (UAV) has become increasingly popular compared to terrestrial mapping technologies such as terrestrial laser scanners (TLS) and Global Navigation Satellite Systems (GNSS), which are costly and have limited coverage. UAV offers quicker, wider coverage, and more detailed result. Typically, UAVs are used in precision agriculture, disaster mapping, and 3D building modeling. However, the accuracy of UAV 3D models often has issues. This study aims to improve the accuracy of 3D building facade models using a combination of depth maps and GCPs and to compare the results with direct measurements using a digital meter. The study was conducted at Kehati Building, Cibinong, Integrated Science Area Complex Area, BRIN. The study began with the acquisition of photo data using UAVs and the measurement of Ground Control Points (GCP). Subsequently, the aerial photos were processed into orthophotos with four processing scenarios based on the use of GCP and input mesh. Scenario 1 and 2 without using GCP, the difference is only in the input mesh. Scenario 1 uses dense cloud, while scenario 2 uses depth maps. On the other hand, scenarios 3 and 4 use GCP with the difference in the input mesh used. Finally, the geometric accuracy of the 3D building model produced from the four scenarios was tested. Accuracy tests showed that scenarios using GCP, as demonstrated in scenarios 3 and 4, had better RMSE which are 0.082 m and 0.064 m respectively. However, identification in scenario 3 was more difficult due to the imperfectly formed facade. Therefore, it can be concluded that the use of GCP significantly improves the dimensional accuracy of the facade.