Research On Shadow Detection And Compensation Technology For High Resolution Optical Remote Sensing Images

Space optical remote sensing technology has been increasingly developed in recent years.High resolution remote sensing images are widespreadly applied in both civilian and military fields.Shadow is a basic property as well as a common problem in remote sensing images.On one hand,shadow may reflect geometric information of surface features.On the other hand,shadow leads to remote sensing image interpretations more difficult,such as object recognition,surface parameter inversion and terrian classification.With the improved resolution of remote sensing images,shadow affects more seriously in remote sensing image interpretations.Hence,it is of great theoretical significance and application value to carry out the research on shadow detection and compensation methods of high resolution spatial optical remote sensing images in the fields of urban planning,vegetation extraction and mapping.Although many researchers have devoted themselves to processing shadow in remote sensing images at home and abroad,there are still shortcomings in current shadow processing technology when facing shadow in high resolution remote sensing images.In this paper,we carried out our research in terms of both shadow detection and compensation methods for high resolution spatial optical remote sensing images on the basis of both the radiative transfer theory and the space optical remote sensing imaging mechanism.The main research work and innovative results are described as follows.1.In order to effectively alleviate the tyical non-shadow misclassification problem for existing shadow detection methods facing high resolution spatial optical remote sensing images mainly covering urban buildings,we developed a logarithmic shadow index?LSI?based shadow detection approach in the 3rd chapter of this paper combining with hue,intensity and near-infrared components.Compared with typical shadow detection methods,the LSI shadow detection method proposed in this paper not only improves the overall accuracy of shadow detection results by more than 2%,but also reduces the misclassification rate by more than 3%and further improves the overall processing speed.The LSI shadow detection method proposed in this paper effectively alleviates the tyical non-shadow misclassification problem for existing shadow detection methods processing high resolution multispectral spatial optical remote sensing images mainly covering urban buildings.2.Aiming at the small shadow omission problem and the tyical non-shadow misclassification problem?especially prominent for the greenish vegetation misclassification problem?for existing shadow detection methods facing high resolution spatial optical remote sensing images mainly covering vegetation,we proposed a mixed property based shadow index?MPSI?shadow detection approach in the 4th chapter of this paper employing hue-intensity difference component and red-near-infrared difference component.Compared with typical shadow detection methods,the overall accuracy of the MPSI shadow detection method proposed in this paper is95.5%,the misclassification rate is reduced to 7.9%,the omission rate is reduced to2.81%,and the time consumption is 47ms.The MPSI shadow detection method proposed in this paper effectively alleviates the small shadow omission problem and the typical non-shadow misclassification problem for existing shadow detection methods processing high resolution spatial optical remote sensing images mainly covering vegetation.3.Aiming at the shadow color distortion problem of compensated images and the easily interference to the information of features in non-shadow areas for existing shadow compensation methods processing high resolution multispectral spatial optical remote sensing images,we presented an irradiance restoration based?IRB?shadow compensation method in the 5th chapter of this paper on the basis of both the radiative transfer theory and the space optical remote sensing imaging mechanism.Compared with typical shadow compensation methods,the IRB shadow compensation method proposed in this paper not only reduces the relative root mean square error r RMSE_shw-nshw of shadow regions in red?R?,green?G?and blue?B?components of the compensated image,but also shows no interference to the information of features in non-shadow regions.The IRB shadow compensation method proposed in this paper effectively alleviates the color distortion and the easily interference to the information of features in non-shadow regions in the process of high resolution multispectral spatial optical remote sensing images by existing shadow compensation methods.