Shadow Detection Method For Large And High Objects Based On Worldview-2 Imagery

In recent years, high-resolution satellite imagery is developing rapidly and is more and more widely used in different fields, especially as an important data source widely used in the geographical conditions monitoring conducted in the last two years. However, a large number of shadows exist in the high-resolution image because the sun is blocked by buildings, tall trees and other surface features. On the one hand, shadows caused the image degradation that leads to the loss of part of the shaded area contains feature information, causes difficulties to target identification and classification; on the other hand, shadows is one of the characteristics of the image, which includes the feature level and height and other information. Therefore, the shadow detection has become a key issue that must be solved in application of high-resolution imagery. However, high-precision shadow detection is a technical problem in image preprocessing due to various reasons such as image uncertainties and other reasons.Shadow detection is a prerequisite for shadow removal and use, high-precision shadow detection algorithm is one of necessary aspects and technical assurance of extensive application of remote sensing images. Through the study of extensive literature, although the existing shadow detection algorithms can effectively detect the shadow area, but still have some flaws, such as depending on a priori knowledge, parameter selection uncertainties, and other issues. For this reason, aiming at the WorldView-2 image commonly used in geographical condition monitoring, this paper presented a shadow detection method based on feature of City of tall objects in the WorldView-2 imagery. The main idea of the proposed shadow detection method is as follows. Firstly, shaded areas and non-shaded areas were divided by histogram threshold method that deals with C3 component in C1C2C3 invariant color space, and the shaded areas contains green vegetation information; Secondly, the vegetation information was extracted using multi-band operation based on counting and analyzing shadow and spectral values of typical features; Thirdly, the target shaded area was obtained based on difference computation between the zones containing shadow and vegetation area; Finally, conduct post-processing procedure for shadow extraction results, such as removal of small areas and morphological filtering.In the study, using a plurality of experimental area in WorldView-2 images, through a lot of experiments verified and analyzed the effectiveness of the algorithm. The result shows that the method is suitable for the WorldView-2 images. It can effectively avoid the influence of water bodies and bluish surface features, and accurately and quickly detect the shaded area, and the method is simple and adaptable.Due to the complexity and diversity of the shadow formation, and difficulties to obtained physical parameters and imaging scene and other information when imaging imagery, detecting all of the shadows still existed many problems. The proposed method is developed based on surface features spectral reflectance properties and mechanism of shadows in remote sensing. The method is expected to be more accurate to extract the shadows if it is combined with the physical characteristics when imaging imagery.