Urban Growth Prediction using Sentinel Satellite Images by Neural Network Method

This research aimed to study urban growth modeling of Rasht city using remote sensing and neural network techniques. To this purpose, land use changes have been detected by analyses of Landsat and sentinel images. Due to improvement of spectral and spatial resolutions of sentinel images compared with Landsat ones, it seems to observe improvements in the accuracy of image processing and monitoring of temporal changes. Map production from images was carried out by combining several classification methods using a decision tree approach and achieving best results from the Sentinel image with the Kappa coefficient of 0.92. For growth urban modeling, the images captured in years 2000 and 2011 were used in a neural network. In order to validate the model, the 2017 map was predicted using the generated model. The matching of the predicted map with the 2017 reference map based on the overall accuracy and Kappa coefficients was 0.9113 and 0.8422, respectively. Finally, according to efficiency of the model, the proposed method was used to predict the 2025 map.