Light pollution map identification of urbanization using npp viirs dnb band and artificial neural networks

In this study, It has been aimed to produce ten citities (İstanbul, Adana, Ankara, Mersin, Antalya, İzmir, Bursa, Gaziantep, Şanlıurfa ve Konya) development maps that selected according to their population status by using the day/night (DNB) band of Visible Infrared Imaging Radiometer Suite satellite(VIIRS), MODIS multispectral image Normalized Difference Vegetation Index (NDVI) and Artificial Neural Networks (ANN). For this purpose, COPERNICUS URBAN ATLAS 2012 of year data and images of June 2012 for VIIRS DNB and MODIS NDVI has been used in order to be able to compare with the maps obtained as a result of ANN. Vegetation Adjusted NTL Urban Index (VANUI) has been calculated using DNB and NDVI image values and VANUI map has been produced. While creating the ANN model, VIIRS DNB, MODIS NDVI and VANUI as input data; reference data as output data(DNB>40 and NDVI<0.4);and sigmoid function as activation function has been used. When the ANN model was run, the success rate has been calculated to be 99.8%. The success of the development maps created for 10 city has been measured as 84% for 2012 and 90% for 2019. When the 2012 forecast images and COPERNICUS data are compared, it has been calculated that there is an average difference of 9% for 10 city. It has been seen that this difference was due to the classification differences in the COOPERNICUS data. As a result of ANN, city maps increased by 13% proportionally to the average population of 10 city between 2012 and 2019. It has been seen that the relationship between population and city area (R2 = 0.95) was directly proportional. As a result of the study, it has been detected that light pollution maps and artificial neural networks algorithm can be used so as to create city area maps. Key Words: Remote Sensing, GIS, Image Processing, Artificial Intelligence, Artificial Neural Networks, City Area Determination, Satellite Data