Ali Akbar Matkan; Mohammad Sohrabinia; Saeid Sadeghian; Dadfar Manavi
Volume 6, Issue 4 , July 2009
Abstract
In this study, we have analyzed how to update large scale maps with the help of IKONOS images. To do this, a complete frame IKONOS image from aerial photos (1:5000 scale) and 1:2000 scale digital maps of the city of Urumia have been used as test data. Here, our objective is to exploit the spatial precision ...
Read More
In this study, we have analyzed how to update large scale maps with the help of IKONOS images. To do this, a complete frame IKONOS image from aerial photos (1:5000 scale) and 1:2000 scale digital maps of the city of Urumia have been used as test data. Here, our objective is to exploit the spatial precision of a pan-chromatic band, the spectral richness of a multispectral image and the spatial and spectral capabilities of a pan-sharpened product at the same time. Meanwhile, geometric correction, image fusion, image information extraction, change detection and incorporation of the changes into the old maps are the main subjects of discussion in the current research. At first, geometric correction of IKONOS image has been analyzed with the help of polynomial, rational and RPC functions, with tables and charts used to compare statistical results of these three models. Overall, the rational model with a third coefficient gave the best result. Geometric correction with the RPC model without any control points gave an RMS error of 15 meters, which decreased to 70 cm. when only one control point was applied. The model with the best results was used to produce ortho-images of IKONOS pan-chromatic and multispectral images. To extract different object classes from the IKONOS image, visual interpretation, pixel-based and fuzzy object extraction methods have been used. Aerial photographs and old maps were used for editing and accuracy assessment of the results. Image analysis methods for visual interpretation, training samples for supervised and fuzzy classification and interpretation of the output classes of unsupervised classification all proved to be very helpful. Further, to detect occurrences of changes occurred relative to the old maps, comparison of the old maps with new extracted maps and comparison of the old maps with the IKONOS image were carried out. Finally, the information content of the IKONOS image was compared with object classes of 1:5000 and 1:2000 scale maps. For 1:5000 scale maps, most of object classes were detectable and recognizable, however, only a limited number of classes in 1:2000 scale maps were detectable and recognizable. In sum, it was found that IKONOS images are capable for the revision of 1:5000 scale maps but has some deficiencies in 1:2000 scale maps.
