Study on Shadow Detection of World View-II Remote Sensing Image Based on Band Standardization

Abstract

This paper presents a valid shadow detection method subjected to WorldView-II remote images combining band standardization with principal component transformation algorithm. Eight bands among The WorldView-II image, each band indicates a specified range of values. Firstly, each band is standardized to the same scale, then utilizing three optimal bands forms a new single band. According to the characteristics of principal component, if the gray value of dark material is positive, just take it for granted, which will be used in next operation. Once the gray value is negative, it will be inverted comparing with the former. In this study, selecting the above adaption builds up an advanced shadow index. Experimental results show that the shadow index can distinguish the shadow from other objects effectively, such as water in high brightness, vegetation, roads, buildings. Meanwhile, there is no significant difference between the water in low brightness and high when using the shadow index. To cope with this problem, we propose an advanced water index based on the normalized differential water index, which is constructed to eliminate the interference of water in low brightness. Results indicate that the indexes can accurately distinguish shadow regions in the WorldView-II image data.