National Technical University of Athens (NTUA)
Αυτόματη εξαγωγή χαρακτηριστηκών από εναέριες και δορυφορικές εικόνες με τεχνικές υπολογιστικής όρασης
Abstract
dc:descriptionIn Remote Sensing, during the last thirty years, techniques for the analysis and processing of satellite data have been developing, with the help of algorithms for the definition of models, patterns and spectral characteristics. Respectively, in Photogrammetry the last fifteen years a rapid growth of techniques and methodologies has been performed, with automations in all kind of algorithms and for the production of various products, as orthoimages, 3d reconstructed objects, etc.. All these years, these two scientific disciplines tend to research, implement, evaluate and extend ideas and techniques from the fields (in the past) of Digital Signal Processing, Image Analysis and Pattern Recognition and (nowadays) from the fields of Computer Vision and Artificial Intelligence. In this framework and aiming at an automatic feature extraction from aerial and satellite imagery, in the present doctoral thesis, low and medium level computer vision techniques were researched and developed. More specifically, algorithms for were developed for: (a) image enhancement and simplification in nonlinear scale spaces, (b) the detection of points, blobs and edges and for (c) image segmentation. Results are described in the following seven bullets, in relation with the methodologies a, b and c, on which are mainly based on. Algorithms were developed, implemented and evaluated for: (a).1 image scale space simplification, where the novel ‘Anisotropic Morphological Leveling’ was introduced. (b).2 automatic point detection from stereopairs and image sequences. (b).3 automatic olive trees detection and counting from satellite images. (b).4 automatic coastline detection from satellite images. (c).5 automatic oil spill detection from radar imagery. (c).6 automatic man-made object detection, like roads, building and airports, from aerial and satellite imagery. (c).7 automatic and simultaneous optical flow estimation and object detection from stereopairs and image sequences. Results were evaluated both qualitatively and quantitatively and are judged promising. Future research based on developed, here, methodologies - which, in the framework of the present thesis, have been studied on a basic research level - towards the construction of automatic feature extraction operational systems, requires a further systematic evaluation, through algorithms application to various, in number and type, aerial and satellite data.
Degree
thesis:*- Grantor dc:publisher
- National Technical University of Athens (NTUA)
- Year dc:date
- 2007
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Karantzalos, Konstantinos
- Καράντζαλος, Κωνσταντίνος
Subjects
dc:subject × 4Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/16981