Evaluating the efficiency of voluntary geographic information in Sentinel-2 images’ classification for urban land cover mapping

In recent decades, urban areas have undergone many changes due to increasing urbanization. Today, with the help of multi-temporal remote sensing images, it is possible to monitor land use changes over decades. High resolution satellite images provide great opportunities to produce urban LU/LC maps. However, such images are expensive and their access are limited. Hence, medium-resolution satellite images such as Sentinel-2 has been widely used in urban applications. Supervised image classifications techniques need accurate training data to detect urban features. Training data collection is difficult and time-consuming and is not easily possible for historical images. Alternatively, voluntary geographic information (VGI) has become widely available from online sources such as OpenStreetMap (OSM), and it may provide a useful source of training data in image classification. This study aims evaluate the efficiency of VGI in classification sentinel-2 time series images (for the years 1394 and 1397) to identification LU changes have been done. For this purpose, the accuracy of classification of Sentinel 2 images with training samples obtained from voluntary geographical information with the accuracy of classification of the mentioned images with training samples obtained from Google Earth images has been compared by T-test at 95% significance level. The results of T-test for 1394 and 1397 show that there is no significant difference between the data set of Google Earth images and VGI. Therefore, the results confirmed that the use of VGI training samples provides good results in monitoring the land use changes.