Mapping Changes in Urban Areas Using Satellite Imagery with an Emphasis on Maximum Use of Old Maps

Due to the rapid transformation of the societies، and the consequent growth of the cities، it is necessary to study these changes in order to achieve better control and management of urban areas and assist the decision-makers. Accordingly it is essential to design a proper solution for image classification. For this purpose، there are numerous methods that can be divided into two general categories; pixel-based and object-based methods. Object-based classification and information extraction methods are usually known as being more effective for classification and interpreting purposes especially in urban areas. However supervised object-based classification، like other supervised methods، requires accurate and sufficient training data. Due to these facts it is highly motivated to design a efficient method to provide reliable training data from existing data sets. In this regard، the training samples was extracted from map and purified by using K nearest neighbor and k-means methods. The results of training data editition improved over 15 percents in the classification accuracy. And change map extraction accuracy in the best case was 98. 08 percent.