OCENA DOKŁADNOŚCI MODELU BUDYNKU Z BARDZO GĘSTEJ CHMURY PUNKTÓW POZYSKANEJ Z INTEGRACJI ZDJĘĆ O RÓŻNEJ GEOMETRII
Rafał Drzewicki, Aleksandra Bujakiewicz · Archives of Photogrammetry Cartography and Remote Sensing · 2018
Accuracy for reconstruction of 3D models of buildings, depends mainly upon density of point clouds, which are used for their creation. The methods for creation of the very dense points clouds on base of automatic measurement of the multi images have been successfully developed. In this paper, the example for automatic reconstruction of 3D model of building of quite complicated shape with use the dense points cloud from non-metric photographs of different geometry, is presented. For this purpose, using the BSP (DJI Phantom) from three height levels - 200 photographs with overlap of about 90%, were acquired. In addition, 46 photographs from ground stations with the camera (FUJIFILM X-S1),were taken. The entire reconstruction process of 3D model of the building, was executed with Agisoft PhotoScan programe. Because of different cameras specification for photographs taken from BSP and from ground stations and various approaches for cameras calibration, the two groups of photographs were located in two classes (clasters), for which the exterior orientation parameters (EO) were separately determined by aero and /terra triangulation, referenced to the same ground coordinate system. The automatic measurement of very large numer of image object points on all photographs and the use of their interior and exterior orientation parameters, have enabled to generate one common very dense points cloud (about 6 millions), which was used to produce the final 3D building model in a few forms. The accuracy of reconstruction of the building model shape was estimated on base of comparison of the model and real data (measures on the building) and also the RSE for the control and check points. The overall accuracy of 0.01 – 0.03 meters was obtained, which have confirmed the high potentiality for integration of different geometry non-metric photographs for the reconstruction of good quality 3D model