Automatic extraction of shadow and non-shadow landslide area from ADS-40 image by stratified classification

Abstract

The objective of this study is fast and accurate to detect the landslides automatically from shadow areas and non-shadow areas that use ADS-40 airborne multispectral image by stratified classification method. First, the shadow area was detected by the brightness method. The shadow and non shadow images were calculated Normalized Difference Vegetation Index (NDVI), and we used iterative self-organizing data analysis technique (ISODATA) unsupervised classification to classify the area of vegetation and non-vegetation. The highest overall classification accuracy of shaded and non-shaded Landslides was 85.75% and 92.75%, respectively. The classification of shaded area by 12-bit image radiation information has a certain capacity. This automated process can be effectively and quickly obtain information of Landslide.