期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2010
卷号：XXXVIII - Part 5
页码：228-233
出版社：Copernicus Publications
摘要：The metric documentation of architecture and cultural heritage with multi-image spherical panorama has already achieved good results, accurate and reliable (Fangi 2008, 2009). Panoramic images are generally acquired with expensive linear array rotating panoramic cameras having very high metric performances (Luhmann, Maas, 2003, 2004, and 2006). On the contrary the multi- image spherical panorama photogrammetry consists in the acquisition of partly-overlapping images taken from a unique point of view with a common digital camera and then projected on a virtual sphere with commercial stitching software; the sphere is mapped on the cartographic plane according to the equirectangular projection, producing the so-called spherical panorama. From the panorama image coordinates, the direction angles –horizontal and vertical – of an arbitrary object point are derived and can be used in a normal topographic adjustment procedure for the 3d evaluation, obviously provided more than one panorama taken from different points of view. Two angular corrections are necessary to compensate for the non perfect verticality of the sphere axes that are the two rotation angles around the horizontal axes. Differently from tachymetry where all the stations are normally linked by direct measurements along the traverse legs, in spherical photogrammetry the station points are connected together by coplanarity, observing a minimum of common points, as it is usually done in photogrammetry. Finally the block bundle adjustment estimates the coordinates of the tie points and the six orientation parameters per panorama, i.e. the three sphere center coordinates and the three direction angles. This metric documentation procedure, particularly suitable for architecture recording, is very quick and inexpensive. Low cost, easiness, completeness, light simple equipment are the main advantages. The interior orientation is skipped, consisting only in the estimation of the radius of the sphere expressed in pixel. Of the basic functions of the spherical photogrammetry, recently some new features have been added (Fangi, 2009). Geometrical constraints - like verticality, or horizontality, belonging to a plane, having equal X or equal Y - can help in the orientation and in the plotting phase to get better results. Once the object to be drawn is laying on a known surface, one can intersect this surface with the projective rays coming from one panorama only, enabling complex details to be plotted in absence of stereoscopy (monoplotting). It is possible to combine together the adjustment of spherical panos and non metric images. To try to solve the problem of the lack of stereoscopy, some procedure has been successfully applied like the photomodeling, in cad environment, using the back projection of the spherical panorama over the rough object model. Finally the algorithms of the spherical photogrammetry have been successfully experimented in the mobile mapping where the panoramas have been obtained with a six lens polycamera, Ladybug 3 (Fangi, 2009). Also the automation in the tie point detection is under experimentation (Barazzetti et al., 2010). From the photographic point of view there are still some problems remaining: the same object is imaged in the different panoramas with very large differences in the image scale, beyond the different illumination conditions, thus making the automatic procedure for tie point identification rather difficult. Normally a panorama is taken with wide-angle focal lens to reduce the amount of images, to get a more stable geometry, and for easiness. On the other end it is good to have the same image scale in the different panos of the same object region, even when they have been taken from different distances. In general the size of the panorama being the limiting factor, one wants to get the maximum resolution according to the interested FOV of the panorama: when FOV= 360° , wide angle is the most suitable, when FOV< 180° normal angle, and finally when FOV < 10° narrow angle is preferable. The local equalization of the image scale of the interested panoramas can be obtained using different focal lenses, according to the camera-object distance. It happens that the photographic scale of a panorama is not everywhere suitable for plotting, normally too small for the farther points. To solve the problem of the remote points, one can use long focal lenses with very narrow field of view, like the one of 2.5°, used for the experiment. But long focal lenses have a weak geometry, making the orientation difficult to be achieved. For long distances, a small orientation error results in large displacement of the plotted object. To help the orientation, the described approach has been followed. From the same camera point, panoramas both with normal and narrow angle lens are taken. The first panorama is used to establish the station coordinates and some control points. The camera station coordinates can then be imposed in the orientation of the narrow FOV panorama by bundle block adjustment. The imposition of the camera station point coordinates gives a good stability to the bundle geometry and ensures a better accuracy to the plotting. The study case is the survey of the top of the bell tower of the Shrine of the Holy House, Loreto, Italy and the Saint Yves at La Sapienza , Rome. The two bell towers have been designed by the famous architects F. Borromini (half of the 17 th century ) and L.Vanvitelli, one century later). These richly decorated bell towers are very high, 60-70 m above ground level. It was impossible to get close to the object for practical reasons. The shape and geometry of this architecture are complex and required a good resolution for the detail plotting. P anoramas have been realized with images acquired with a normal SRL digital camera using 28, 50, 200, and
