Shadow Detection Method For Large And High Objects Based On GF-Series Images

Because of the sun is blocked by buildings,tall trees and other surface features,there are so many shadows exist in the GF-series images.The existence of the shadow will cause distortion of the image information,lower image quality and make some information obtain difficultly,which will have an bad effect on feature recognition and statistical information.Therefore,how to detect the shadow of GF-series images efficiently and accurately and how to eliminate the shadow is becoming a hot and difficult issue of application and research.GF-Series Images refer to the highest resolution satellite remote sensing images of China,According to literatures and experimental research works,there are lot existing shadow detection methods for GF-1 or GF-2.Then,this paper creates 3 shadow detection methods for GF-1 and 1shadow detection method for GF-2.The main research results:(1)One method of shadow detection research based on the characteristics of GF-1 imagesFirstly,use multi-spectral image and panchromatic data to do the data correction and information fusion,in order to use spectral resolution and information maximumly;Then protrude shadow information and increase the differences of the shadow and other features.Secondly,Establish PC1/NIR and Image characteristics calculation Rg_nir,the shaded area and water binary images are detected by multimodal histogram threshold algorithm;Finally,do logic operations on initial shaded area and water body,in order to achieve the purpose of excluding water.Then Use morphological closing operation to make the shadow region more continuous.(2)A method of shadow detection for High resolution remote sensing image tall ground objects based on HSV and PCAFirstly,the spectral values of typical surface features such as water bodies and shadows are counted and analyzed in GF-1 images.The shaded areas and non-shaded areas are identified based on principal component transform,but the shadow areas contains water,dark feature information.Secondly,the shaded areas and non-shadedar areas divided based on histogram threshold method for V component in HSV color space.The separation results contain green vegetation,but does not contain water and dark feature information.Finally,using logical operations on the two compound shaded areas.Get the shaded area and non-shaded area binary images.(3)Shadow Multi-scale Segmentation Method Based on the Characteristic ComponentFirstly,use multi-spectral image and panchromatic data to do the data correction and information fusion,in order to use spectral resolution and information maximumly;Then protrude shadow information and increase the differences of the shadow and other features.Secondly,Establish characteristics calculation PC1,V,G and NDVI,then use characteristics calculation to restructure band;Finally,Make new image multi-scale segmentation,and use the features to build rule set,then realizes the shadow information extraction.(4)One method of shadow detection research of GF-2 imagesFirstly,the spectral values of typical surface features such as water bodies and shadows are counted and analyzed in GF-2 images.The shaded areas and non-shaded areas are identified based on principal component transform.Secondly,Calculate the vegetation information by NDVI.Finally,do logic operations on initial shaded area and vegetation information.The result indicate that this method not only had high detection accuracy and efficiency for GF Series images,but weakened the influences of water bodies and bluish of shadows.To effectively solve the GF Series images series images shadow detection technical problems.