Image matching for 3D photogrammetric reconstruction

Ivan Detchev, Ayman Habib, Jiann Yeou Rau

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Image matching or the identification of conjugate points appearing in two or more photographs is an integral processing task in photogrammetry. More specifically, three-dimensional photogrammetric reconstruction, i.e. computing the object space coordinates of certain points of interest, would be impossible without having these points matched in the image space of at least two photographs. Observing a point of interest in only one photograph is simply not enough, because the range from the image point to the object point is inherently unknown. Thus, image matching is applied to satellite scenes and aerial photography for the generation of digital terrain models or building models, and also in close range photogrammetry for the reconstruction of objects and surfaces. The human brain has incredible capabilities and manual image matching, i.e. one performed by a human operator, is very accurate and reliable; however, it is labour intensive, time consuming, and can become quite tedious very quickly. This is why the topic of fully automated image matching, i.e. one performed solely by a computer, has been addressed in photogrammetric research for a long time. While the problem of automatically matching signalized targets has been satisfactorily solved, the matching of natural or non-signalized features still requires some human interaction. This paper will attempt to summarize the methodology of performing image matching of ordered photographs in a controlled environment, i.e. the interior and the exterior orientation parameters for the involved cameras and photographs are known. Example results from using a sophisticated commercial software package called CLOse RAnge MAtching (CLORAMA) and from a simple in-house matching program will be shown. The former method is based on least squares matching and requires very high resolution imagery when homogenous texture is present, while the latter method is based on the straightforward normalized cross-correlation matching and requires a pattern to be projected in order to create artificial texture.

Original languageEnglish
Title of host publication32nd Asian Conference on Remote Sensing 2011, ACRS 2011
Pages247-252
Number of pages6
Publication statusPublished - 2011 Dec 1
Event32nd Asian Conference on Remote Sensing 2011, ACRS 2011 - Tapei, Taiwan
Duration: 2011 Oct 32011 Oct 7

Publication series

Name32nd Asian Conference on Remote Sensing 2011, ACRS 2011
Volume1

Other

Other32nd Asian Conference on Remote Sensing 2011, ACRS 2011
CountryTaiwan
CityTapei
Period11-10-0311-10-07

All Science Journal Classification (ASJC) codes

  • Computer Networks and Communications

Fingerprint Dive into the research topics of 'Image matching for 3D photogrammetric reconstruction'. Together they form a unique fingerprint.

Cite this