The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XXXVIII-5/W16
10 Sep 2012
 | 10 Sep 2012


R. Roncella, C. Re, and G. Forlani

Keywords: Orientation, Automation, Reconstruction, Close Range, Photogrammetry, Software, Comparison, Estimation

Abstract. Automatic orientation of image sequences in close range photogrammetry is becoming more and more important, not least to maintain a degree of competitiveness with other survey techniques, such as laser scanning. The objective of this paper is to compare two Structure from Motion (SFM) strategies. The previous strategy has been used at our Department for some years already in a wide range of projects and is based on the Harris operator and the fundamental matrix plus the trifocal tensor estimation to filter out the outliers. While it has in most cases performed satisfactorily, the percentage of accepted matches is generally smaller than expected; sometimes this leads to failure of the successful estimation of the trifocal tensor. The second one has only recently been implemented and is still under testing; it is based on the SURF operator and the 5-point relative orientation algorithm. The paper will show a comparison between the two strategies on a series of test cases.