Selecting Images From UAV Sequences For Photogrammetric Reconstruction

Photogrammetric reconstructions are becoming increasingly popular due to their precision, accessibility, and cost-efficiency, with drones (UAVs) being utilized more frequently for data collection. The method of systematic frame selection from video sequences sometimes fails to deliver accurate results due to the changing flight speeds of the device and alterations in movement angles. This study seeks to assess six methods for selecting frames from video sequences at non-uniform intervals and examine their efficacy in producing architectural models. Two sections of the Agricultural University of Krakow campus were selected as representative research sites, utilizing data from the Geoportal (2021) as reference material. Various frame choice approaches were examined, with the Tanimoto method demonstrating superior speed, yielding optimal reconstruction outcomes and the highest accuracy in land coverage.