Digital Surface Models (DSMs) have large relevance in many engineering, land planning and environmental applications for a long time. At present, the data required for the DSMs generation can be acquired by several sensors/techniques, among which airborne LiDAR, aerial photogrammetry, optical and radar spaceborne sensors play the major role. In this respect, the availability of new high resolution optical spaceborne sensors offers new interesting potentialities for DSMs generation, among which low cost, speed of data acquisition and processing and relaxed logistic requirements, quite important for the areas where the organization of aerial flights can be difficult for several motivations. Thanks to the very high resolution and the good radiometric quality of the images, it seems possible to extract DSMs comparable to middle scale aerial products; anyway, it has to be underlined that the DSM accuracy level is strictly related both to the quality of the stereo image orientation and to the effectiveness of the matching strategy. Two different types of orientation models are usually adopted: the physical sensor models (also called rigorous models or geometric reconstruction) and the generalized sensor models. The first one is based on a standard photogrammetric approach, where the image and the ground coordinates are linked through the collinearity equations, so that the involved parameters have a physical meaning. On the contrary, the generalized models are usually based on the Rational Polynomial Functions (RPFs), which link image and terrain coordinates through the Rational Polynomial Coefficients (RPCs) and eventual additional transformation parameters [Tao and Hu, 2001, 2002; Fraser and Hanley, 2003; Hanley and Fraser, 2004; Crespi et Al., 2009]. As regards the matching, it is well known that many different approaches have been developed in recent years. In all methods, the main step is to define the matching entity, that is a primitive chosen in the master image to be looked for in the slave image(s); basically, we can distinguish two techniques, the Area Based Matching (ABM) and the Feature Based Matching (FBM). In ABM methods, a small image window represents the matching primitive and the main strategies to assess similarity are cross-correlation and Least Squares Matching (LSM). FBM methods use, as main class of matching, basic features that are typically the easily distinguishable primitives in the input images, like corners, edges, lines, etc. [Gruen A. W. 1985; Jacobsen, 2006; Nascetti, 2009; Tang L. et al.,2002]. In addition, new matching strategies where ABM is used together with dynamic programming techniques were proposed during last decade [Birchfield S. and Tomasi C., 1998, 1999]; recently the quite promising technique of semi-global matching was proposed and applied to aerial imagery [Hirschmüller, 2008; Hirschmüller and Scharstein, 2009]. In this paper we present and discuss some results obtained with a new proprietary matching strategy for DSMs generation, which is implemented into the SISAR software developed at the Area di Geodesia e Geomatica – Università di Roma "La Sapienza". In order to assess the accuracy of the new strategy, some tests were carried out, using a stereo pair of Augusta coastal zone (Sicily, South Italy) acquired from WorldView-1 and one of the first available GeoEye-1 stereo pair, which was acquired over Rome. The results show that an accuracy at the level of about 2 m is achievable in open areas with both WorldWiew-1 and GeoEye-1 stereo pairs, whereas higher errors are displayed in urban areas. For WorldWiew-1 the results are still acceptable, being the accuracy at the level of 3 meters, but for GeoEye-1 the DSM extracted over a very dense urban area are much worse, with an accuracy at the level of 8-10 meters. Nonetheless, the new matching strategy has been proven effective, performing always better if compared with the one implemented into a well known and largely used software as PCI-Geomatics. In order to evaluate the potentiality of the new matching strategy and the accuracy of the extracted DSMs, some tests were carried out. In details, two stereo pairs acquired byWorldView-1 and GeoEye-1 satellites have been used to compare the DSMs generated with the new strategy to those derived using the well known commercial software PCI Geomatics v.10.2 (OrthoEngine).
Scheda prodotto non validato
Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo
|Titolo:||Premio Nazionale del 2010 Associazione Italiana di Cartografia (Secondo Classificato)|
|Data di pubblicazione:||2010|
|Appartiene alla tipologia:||14a Premio o riconoscimento scientifico|