<![CDATA[Application of Unmanned Aerial Vehicles (UAV) for images acquisiton has been widely applied in survey and mapping. One of non-metric camera as the sensor that can be mounted on the UAV is fish-eye lenses. Fish-eye lenses camera provides images with wide range coverage. However these images are distorted and make them more difficult to use for mapping or 3D modelling. This research is aimed to make a 3D surface model by images reconstruction and to estimate the geolocation accuracy of the model generated by UAV images processing. As the approach of the method, combines the automation of computer vision technique with the photogrammetric grade accuracy. The complete photogrammetric workflow implemented in Pix4D Mapper. Meanwhile, UAV platform used is DJI Phantom 2 Vision+. Sample location in this research is an area of Geospatial Laboratorium in Parangtritis, Yogyakarta. The covered area in this research is 3.934 Ha. From the results of 186 images obtained 2.47 cm value of average Ground Sampling Distance (GSD). Moreover the numbers of 3D points for Bundle Block Images Adjustment are 243,373 points with 0.4348 value of Mean Reprojection Error (pixels). The results of 3D Densified Points are 6,207,780 and 101.04 points of average density per-m3. Generally, geolocation acuracy of the model produced by using this method is between 2.47 - 4.94 cm. Thus, it can be concluded that UAV with fish-eye lenses camera can be used to reconstruct 3D surface model. However, images correction and calibration should be required to produce an accurate 3D model.]]>
Copyrights © 2018
