作者: Masoud Hajeb , Sadra Karimzadeh , Abdolhossein Fallahi
DOI: 10.1007/S11069-020-03991-0
关键词: Remote sensing 、 Speckle noise 、 Change detection 、 Linear discriminant analysis 、 Geology 、 Correlation coefficient 、 Naive Bayes classifier 、 Random forest 、 Support vector machine 、 Synthetic aperture radar
摘要: The synthetic aperture radar SAR system with the capability of imaging during night, day, and all-weather conditions has a high potential in change detection on ground surface. In this research, we used three images ALOS-2 satellite over Sarpole-Zahab town west Iran that had an earthquake magnitude 7.3 November 12, 2017. effects speckle noise accuracy results were assessed based reduction filters. Correlation coefficient, difference intensity (in five window sizes), coherence texture six sizes) pre- post-event calculated, output parameters extracted. Then, damage assessment was carried out four machine learning classifiers, containing random forest (RDF), support vector machine, naive Bayes classifier, K-nearest neighbor. RDF showed overall 86.3%. Seventy percent dataset for training, 30% it prediction purpose (~ 300 buildings). Based training dataset, total number structures study area predicted (approximately 9200 Finally, discriminant analysis among damaged undamaged buildings.