Image understanding and feature extraction for applications in industry and mapping

Simple item page
dc.contributor.advisorRüther, Heinzen_ZA
dc.contributor.authorCalitz, Michaelangelo Francoen_ZA
dc.date.accessioned2015-12-28T05:58:19Z
dc.date.available2015-12-28T05:58:19Z
dc.date.issued1995en_ZA
dc.descriptionBibliography: p. 212-220.en_ZA
dc.description.abstractThe aim of digital photogrammetry is the automated extraction and classification of the three dimensional information of a scene from a number of images. Existing photogrammetric systems are semi-automatic requiring manual editing and control, and have very limited domains of application so that image understanding capabilities are left to the user. Among the most important steps in a fully integrated system are the extraction of features suitable for matching, the establishment of the correspondence between matching points and object classification. The following study attempts to explore the applicability of pattern recognition concepts in conjunction with existing area-based methods, feature-based techniques and other approaches used in computer vision in order to increase the level of automation and as a general alternative and addition to existing methods. As an illustration of the pattern recognition approach examples of industrial applications are given. The underlying method is then extended to the identification of objects in aerial images of urban scenes and to the location of targets in close-range photogrammetric applications. Various moment-based techniques are considered as pattern classifiers including geometric invariant moments, Legendre moments, Zernike moments and pseudo-Zernike moments. Two-dimensional Fourier transforms are also considered as pattern classifiers. The suitability of these techniques is assessed. These are then applied as object locators and as feature extractors or interest operators. Additionally the use of fractal dimension to segment natural scenes for regional classification in order to limit the search space for particular objects is considered. The pattern recognition techniques require considerable preprocessing of images. The various image processing techniques required are explained where needed. Extracted feature points are matched using relaxation based techniques in conjunction with area-based methods to 'obtain subpixel accuracy. A subpixel pattern recognition based method is also proposed and an investigation into improved area-based subpixel matching methods is undertaken. An algorithm for determining relative orientation parameters incorporating the epipolar line constraint is investigated and compared with a standard relative orientation algorithm. In conclusion a basic system that can be automated based on some novel techniques in conjunction with existing methods is described and implemented in a mapping application. This system could be largely automated with suitably powerful computers.en_ZA
dc.identifier.apacitationCalitz, M. F. (1995). <i>Image understanding and feature extraction for applications in industry and mapping</i>. (Thesis). University of Cape Town ,Faculty of Engineering & the Built Environment ,Division of Geomatics. Retrieved from http://hdl.handle.net/11427/15942en_ZA
dc.identifier.chicagocitationCalitz, Michaelangelo Franco. <i>"Image understanding and feature extraction for applications in industry and mapping."</i> Thesis., University of Cape Town ,Faculty of Engineering & the Built Environment ,Division of Geomatics, 1995. http://hdl.handle.net/11427/15942en_ZA
dc.identifier.citationCalitz, M. 1995. Image understanding and feature extraction for applications in industry and mapping. University of Cape Town.en_ZA
dc.identifier.ris TY - Thesis / Dissertation AU - Calitz, Michaelangelo Franco AB - The aim of digital photogrammetry is the automated extraction and classification of the three dimensional information of a scene from a number of images. Existing photogrammetric systems are semi-automatic requiring manual editing and control, and have very limited domains of application so that image understanding capabilities are left to the user. Among the most important steps in a fully integrated system are the extraction of features suitable for matching, the establishment of the correspondence between matching points and object classification. The following study attempts to explore the applicability of pattern recognition concepts in conjunction with existing area-based methods, feature-based techniques and other approaches used in computer vision in order to increase the level of automation and as a general alternative and addition to existing methods. As an illustration of the pattern recognition approach examples of industrial applications are given. The underlying method is then extended to the identification of objects in aerial images of urban scenes and to the location of targets in close-range photogrammetric applications. Various moment-based techniques are considered as pattern classifiers including geometric invariant moments, Legendre moments, Zernike moments and pseudo-Zernike moments. Two-dimensional Fourier transforms are also considered as pattern classifiers. The suitability of these techniques is assessed. These are then applied as object locators and as feature extractors or interest operators. Additionally the use of fractal dimension to segment natural scenes for regional classification in order to limit the search space for particular objects is considered. The pattern recognition techniques require considerable preprocessing of images. The various image processing techniques required are explained where needed. Extracted feature points are matched using relaxation based techniques in conjunction with area-based methods to 'obtain subpixel accuracy. A subpixel pattern recognition based method is also proposed and an investigation into improved area-based subpixel matching methods is undertaken. An algorithm for determining relative orientation parameters incorporating the epipolar line constraint is investigated and compared with a standard relative orientation algorithm. In conclusion a basic system that can be automated based on some novel techniques in conjunction with existing methods is described and implemented in a mapping application. This system could be largely automated with suitably powerful computers. DA - 1995 DB - OpenUCT DP - University of Cape Town LK - https://open.uct.ac.za PB - University of Cape Town PY - 1995 T1 - Image understanding and feature extraction for applications in industry and mapping TI - Image understanding and feature extraction for applications in industry and mapping UR - http://hdl.handle.net/11427/15942 ER - en_ZA
dc.identifier.urihttp://hdl.handle.net/11427/15942
dc.identifier.vancouvercitationCalitz MF. Image understanding and feature extraction for applications in industry and mapping. [Thesis]. University of Cape Town ,Faculty of Engineering & the Built Environment ,Division of Geomatics, 1995 [cited yyyy month dd]. Available from: http://hdl.handle.net/11427/15942en_ZA
dc.language.isoengen_ZA
dc.publisher.departmentDivision of Geomaticsen_ZA
dc.publisher.facultyFaculty of Engineering and the Built Environment
dc.publisher.institutionUniversity of Cape Town
dc.subject.otherDigital photogrammetryen_ZA
dc.titleImage understanding and feature extraction for applications in industry and mappingen_ZA
dc.typeDoctoral Thesis
dc.type.qualificationlevelDoctoral
dc.type.qualificationnamePhDen_ZA
uct.type.filetypeText
uct.type.filetypeImage
uct.type.publicationResearchen_ZA
uct.type.resourceThesisen_ZA
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
thesis_ebe_1995_calitz_michaelangelo_franco.pdf
Size:
8.77 MB
Format:
Adobe Portable Document Format
Description:
Download
Collections
PhD / Doctoral