Abstract
The goal of this paper is to focus on multi-image monoscopic digital photogrammetry, illustrating several types of applications used in a single case study chosen for its unique characteristics: Palazzo dei Capitani del Popolo in the main square in the old town centre of Ascoli Piceno. The description of this experimentation will be illustrated in the paper not only to assess the potential and limits of these systems, but also to place them in a scientific context and recall the theoretical fundamentals of this method, since we believe these in-depth studies to be increasingly necessary in order for these digital technologies to be used properly.
Top1. Introduction
The first procedural changes in the field of photogrammetry took place in the eighties thanks to developments in the computer world. However, only when digital photography was perfected and personal computers improved their processing power was there any significant change in the applications, which then became popular and widespread.
These purely technological improvements – calculus algorithms and digital photographic materials – made it possible to fully exploit the scientific, geometric, projective and analytical fundamentals already inherent in the method, which in fact remained practically unchanged.
In moving to the digital realm, the main geometrical/projection and analytical principles forming the basis of photogrammetry remain largely unchanged. In general, to attain a description of the photographic object, one still has to establish a spatial relationship between the points of the subject, the points represented on the plate/film/sensor, and the positions and inclinations with respect to the centres of projection and optical axes of the photographic equipment. The combination of such operations, called “orientations”, aims to identify, when known, the intrinsic parameters (principal distance, position of the centre of projection with respect to the plate/sensor, lens distortion parameters, etc.) and extrinsic parameters (position and orientation of the photographic apparatus with respect to the points of the photographed subject) that characterise the photogrammetry model (Carpiceci, 2012; Cundari, 1983; Docci, 1964; Docci & Maestri, 2012; Fondelli, 1992; Girelli, 2007; Ippoliti, 2000; Paris, 2014; Saint Aubin, 1992).
In fact, albeit with some differences, digital photogrammetry systems and software make substantial reference to theoretical principles, so it can thus be termed “classic”. It utilises, in particular, both the collinearity equations, which allow the photogrammetry system to be described analytically with known and unknown variables, and epipolar geometry, with which it is possible to reduce the space in which correspondences between homologous points are sought.
This is the reference framework of this paper which will focus on multi-image monoscopic digital photogrammetry systems and software, i.e., all applications which use restitution methods in which the photographic shots are digital, superior to one, and restitution (even when guided by a restitution operator) is achieved without the assistance of stereoscopic observation. There are two main groups of applications depending on whether or not a restitution operator is required.
The first group includes so-called photomodelling software where the operator is tasked with identifying homological points for orientation as well as the characteristics points on which to base the geometric and texturised modelling processed by the software itself (De Luca, 2011).
The second group includes structure from motion systems (developed even more recently than the ones in the first group) in which the elaboration of different ad hoc algorithms has led to the automation of several stages: from orientation, including self-calibration, to the extraction of more or less dense points clouds, according to specific criteria of precision, used to construct a model, including a texturised model (Sicuranza, 2013; Remondino & El-Hakim 2006; Remondino, Del Pizzo, Kersten & Troisi, 2012).
The structure from-motion systems make it possible to automate the restitution process but, unlike manual systems, the points are not chosen by the operator. The latter, however, is responsible for controlling the input data (photographic campaign, quality of the photos, checking the self-calibration) and output data (precision and density of the points clouds, presence of gaps and their treatment during conversion into a polygonal model, etc.).
Within this reference framework, the goal of this paper is to focus specifically on multi-image monoscopic digital photogrammetry, illustrating several types of applications used in a single case study deliberately chosen for its unique characteristics: Palazzo dei Capitani del Popolo in the square with the same name in the old town centre of Ascoli Piceno.
Key Terms in this Chapter
Epipolar Geometry: Epipolar geometry deals with the relationships and limits between two or more photographic images that frame a single subject. Such relationships and limits are then used in procedures to obtain photogrammetry.
Stereoscopic Vision (Natural and Artificial Stereoscopic Vision): As is known, human vision is capable of perceiving differences in distances to points in space. That is, human vision can perceive relief—it is stereoscopic. In fact, in seeing two different points, observers cannot evaluate the real distances between themselves and the points; rather, they perceive the difference between the two distances, i.e., they can assess which of two points is closer and which is further away. The physiological process of vision is very complex, but for our purposes it is important to recall that the perception of each individual point is realised by forming two separate retinal images, one in each eye. Using the principles of natural binocular vision, it is therefore possible to artificially reproduce stereoscopic vision. When simultaneously observing two photographic images of the same subject taken from two different points of view, instead of perceiving two distinct images, the observer sees a three-dimensional image of the subject. However, the appropriate optical tricks are necessary. In particular, the right-hand image should be seen only by the right eye and the left image only by the left eye. Under these conditions, two different retinal images are formed and sent simultaneously to the brain, which processes them and combines them to form a single virtual three-dimensional model. In theory, it is possible that whatever the photography conditions are, the two images always constitute a virtual three-dimensional model. While this is true of geometrical projection, to produce the effect in reality, the images should be made by simulating the conditions of natural binocular vision, thus taking into account the parameters of the inter-pupil distance and the object distance, with the associated reciprocal variations.
Perspective and Photography (Identity Relationship between Perspective and Photography): The condition that allows the principles of projection to be applied to photography comes from the identity relationship between photographic images and central projection. In fact, the process of forming central projection can be seen in the ideal bundle of projection rays, with their centre in the objective lens, which hit the points of the object and fix the relative image in the photographic plane. According to this identity relationship, the centre of the objective lens is the point of view of central projection; the photographic image plane is the perspective plane on which the image forms; and the focus of the objective lens is the principal distance. If each photo is considered as a perspective, it is possible to render the geometric configuration of the subject photographed if the elements defining the perspective system—i.e., the principal point, the horizon, and the distance circle—are known or can be defined. It is also possible to obtain measurements (and not just proportions) if the photography conditions or at least one real measurement relative to either the subject is known.
Monoscopic Multi-Image (Monoscopic, Multi-Image Photogrammetry): The term “monoscopic, multi-image photogrammetry” is used when multiple photographic images of the same subject are used but the procedures do not make use of stereoscopic observation.
Calibration: This is a procedure to determine the deformation parameters of the photographic image caused by the device/objective lens in the photographic equipment.
Focal Length: The distance from the optical centre of the objective lens and the photosensitive support, expressed in mm.
Photogrammetry: Photogrammetry is a surveying method through which precise rigorous information is obtained regarding the geometry, form, measurement, and material quality of the subject photographed. Given the identity relationship between photographic image and central projection, the operational processes allowing measurements to be obtained from the images, and therefore the metrical/geometrical contents of the objects to be investigated, are completely based on applying the principles of projection geometry to the photographic images. In general, describing the subject photographed requires establishing a spatial relationship among the object points, the points represented on the plate/film/sensor, their positions and inclinations with respect to the centres of projection and the optical axes of the photographic equipment. The combination of such operations, collectively called “orientations”, aims to identify, when known, the intrinsic (principal distance, position of the centre of projection with respect to the plate/sensor, lens distortion parameters, etc.) and extrinsic parameters (position and orientation of the photographic equipment with respect to the points of the subject) that characterise the photogrammetric model. The theoretical principles of photogrammetry refer to the equations of collinearity (with which the phtogrammetric system in the form of known and unknown variables can be analytically described) and epipolar geometry (with which it is possible to reduce the area in which one looks for relationships between homologous points).
Resolution: This defines the quality of the photographic image. It expresses the density of photo-sensitive points in an analogue or digital support and is measured by the number of points per length (generally inches). DPI is the acronym for “dots per inch”.
Structure-from-Motion (Structure-from-Motion Systems for Photogrammetry): In the field of monoscopic, multi-image digital photogrammetry, a structure-from-motion system is the set of applications that use rendering methods of more than one native digital snapshot and most of the rendering is done with automatised procedures. In these systems, the development of algorithms, computer vision in particular, has allowed the different phases to be automatised: from the orientation phase, including auto-calibration, to the more or less dense point-cloud-extraction phase according to determined accuracy criteria, from which a texturised model can also be obtained.
Digital Photogrammetry: Digital photogrammetry uses native digital photographic materials automatised procedures to manage the different phases of the process. The development of digital photogrammetry is due to the most recent technological progress, which has led to the production of particularly high-performance computers at low prices and to the development of powerful calculation algorithms.