Συντάχθηκε 12-12-2011 09:25
από Eleni Stamataki
Email συντάκτη: estamataki<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: σύνταξη/αποχώρηση υπάλληλος.
Τμήμα Ηλεκτρονικών Μηχανικών & Μηχανικών Υπολογιστών
ΠΑΡΟΥΣΙΑΣΗ ΔΙΠΛΩΜΑΤΙΚΗΣ ΕΡΓΑΣΙΑΣ
ΚΑΝΕΛΛΙΔΗ ΕΛΕΝΗ
με θέμα
“Ακριβής αναγνώριση περιεχομένου φωτογραφιών φύσεως”
“Registration and Content Recognition of nature pictures”
Τετάρτη 14 Δεκεμβρίου 2011, 11πμ
Αίθουσα 2042, Εξωτερικά στο Κτίριο Επιστημών, Πολυτεχνειούπολη
Εξεταστική Επιτροπή
Καθ. Σταύρος Χριστοδουλάκης (επιβλέπων)
Καθ. Μιχάλης Ζερβάκης
Αν. Καθ. Ευριπίδης Πετράκης
Abstract
This thesis is based on the SPIM framework and introduces the SPIM+ framework. The SPIM framework is used for semantic spatial information processing. The framework exploits the modern digital camera’s potential for capturing contextual parameters through the use of sophisticated sensor devices, information found in specially annotated semantic maps and industrial standards. It processes images with embedded positional and directional information. The objective is to effectively manage and associate the information and semantic objects contained in both the semantic maps and the images. In addition, the framework employs the use of image processing and other algorithms to enable the automatic annotation of the images.
Although SPIM achieves its objectives with considerable success, there are still some aspects that can be improved. SPIM+ is based on the SPIM implementation and uses its principal components in order to add functionality to the initial system and improve it.
The SPIM framework registers images of nature with the model which is produced by the 3D land maps. A number of experiments under various conditions (time of day, season, weather condition, etc.) were performed and it was observed that on many occasions, the SPIM failed to registered the images correctly. The aim was to improve its performance. To do that not only were improvements made to the original algorithm but also alternative detection objectives (shore lines, island lines etc.) were employed, and the position of the sun was exploited by taking into account the time and season. The results were incorporated into the SPIM+ main algorithm. The SPIM+ main algorithm was then tested in a new set of images and the quality of its results was evaluated and compared with the original SPIM algorithm. SPIM+ significantly improves the registration performance of the original algorithm.
Finally, the initial system was extended in order to process images which have only positional and no directional information embedded in them.
