الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 179 |  |  |
| | | from | 179 | | |
Abstract
Feature extraction for the purpose of mapping from space images became important in recent years to acquire accurate information on urban land use/land-covers and their change over time. Land cover classification of very high resolution (VHR) imagery over urban areas is an extremely challenging task. Impervious land covers such as buildings, roads, and parking lots are spectrally too similar to be separated using only the spectral information of VHR imagery. Additional information, therefore, is required for separating such land covers by the classifier. Thematic maps representing the characteristics of the Earth’s surface have been widely used as a primary input in many land related studies. Classification of remotely sensed images is an effective way to produce these maps. The value of the map is clearly a function of the accuracy of the classification. selecting proper size of samples and classification methods are essential issues to produce accurate thematic maps. Updating maps is a serious and costly task that requires extensive and time-consuming field revision. Thus, it becomes crucial to develop unconventional strategies for map updating. In the present study, three stages of investigations have been implemented. The first stage, equal training data sets at various sizes and a proportional sample size used to investigate the effect of the training set size on the classification accuracy to set the appropriate sample size for each classifier. Six supervised classification methods with different characteristics were applied to produce land use/land cover thematic map of the study area. The used classifiers include: Parallelepiped, Minimum Distance, Mahalanobis Distance, Maximum Likelihood, Neural Network and Support Vector Machine (SVM). Also a proportional training sample size has been applied and a comparison was held. In the second stage, we study the effect of using auxiliary data in the classification results using the outputs of first stage (best classifier and training size), and generated Co-occurrence matrix attributes. Three investigations have been implemented: 1) using single attribute of each band. 2) using the group attributes of each band. 3) Using the all attributes of bands. In the third stage, the object oriented technique was implemented which depends on the segmentation of the classified image. The effect of using both spectral and spatial attributes is available too. • The results showed that optimum sample size differs from classifier to another. In the case of limited number of training pixels, SVM and Maximum Likelihood classifiers produced higher classification accuracies than the rest of classifiers. The group of attributes of Blue band performed the best of the three bands. All classifiers are performing better using all attributes except for Neural Network.