High Resolution Orthoimage Shadow Detection And Anisotropic Scattering Compensation Method

With the appearance of the new remote sensing platform for aerial observation and the rapid development of the supporting software,the production efficiency and the application of the very-high-resolution remote sensing products have increased significantly.Compared with the traditional 4D products,low-altitude platform imaging products with richer features,and more flexible and rapid access way,has gradually become important data sources of the city planning,land and resources exploration,emergency mapping,scientific research,Disaster prevention and mitigation,precision agriculture and other areas.However,in the production process of the orthoimage image only carried out the process of uniforming light and color and a large number of shadows are left untreated.Taking the urban area as an example,the shadows of some areas are long-term.Due to the relatively fixed position of the high-rise buildings and the solar location,shadows of some areas are permanent blind spots of urban remote sensing monitoring.Due to the lack of direct light,shadowed areas suffer from the spectral information loss and the blurring boundary,which lead to the difficulties in the inversion of the spectral parameters,classification of objects,change detection,target extraction,etc.At the same time,the shadows carry a large amount of information related to the geometry and spatial semantics of the projectile,which can be used as the auxiliary information of the 3D reconstruction technology.Through the deep study of the shadow characteristics,we can detect and compensate the shadows more effectively and tap the information of the image to improve the accuracy of the existing image processing algorithms and use data with higher efficiency.Thus,it can be concluded that the detection and compensation of the shadows in high resolution orthophotos are of great value in research and application.Due to the combination of multiple factors such as solar position,surface,atmospheric environment,the boundary,texture and spectral features of the shadow area are complex,and the blue-green features such as waters,vegetation,dark asphalt pavement,high reflective window structure,which have similar characteristics with shadows,often resulting in the failure of traditional shadow processing algorithms.Therefore,it is difficult to accurately locate and compensate shadows only by spectral characteristics.In this paper,the existing 2D image plane is extended to the 3D geospatial space by adding the digital surface model,and the causal model of the shadow is constructed by the relationship between the terrain and the light source.The radiation energy lost of shadowed areas are determined,and the purpose of detecting and compensating the shadow is achieved efficiently and automatically without interference from image quality and ground reflectance.The main contents and conclusions are as follows.1.In-depth study of the ground scene taken by hand-held cameras and the radiation characteristics of the shaded areas in high-resolution remote sensing images.Due to the influence of atmospheric scattering and the surrounding metamorphism,the shaded area is not dark,the amount of information is zero,but has a darker contrast and a lower contrast than the non-shaded area.In the remote sensing image,the information given by the shaded area can be reasonably compensated by the spatial and spectral characteristics of the surrounding environment.Therefore,the search for the image of the low amount of information,the smaller the signal to noise ratio of the shadow area in the unique multi-dimensional attributes,and features similar to the non-shadow to distinguish,to be followed by detection and compensation processing.Understanding the multidimensional features of shadows is a prerequisite for shadow detection and compensation.2.A shadow detection method is proposed to distinguish shadow areas with non-shaded areas without any human intervention.With the aid of the digital surface model,time information and coordinates of the image,the geometric relationship of the sun and the scene is obtained.Ray tracing method is applied to detect the hard shadow of the image as an initial value.Then,the initial shadow value is regarded as the image foreground and the non-shadowed areas as the background.The shadow mask is optimized by the matting technique,and the soft shadow with intensity information of the shadow is taken into account.The shadow mask is the input for subsequent shadow compensation.The robust and reliable shadow detection results are the basic prerequisite for shadow compensation of remote sensing images.3.A spectral compensation method is proposed based on the characteristics of the shadow area and its visibility of the sky.From the perspective of radiative transfer theory,shadow compensation is a ill-posed problem with many parameters and few constrains.Based on the existing spectral features and the geometric model between the light source and the scene during the shadow formation,the necessary parameters and their contribution to the shadow compensation are discussed.The basic radiation compensation model is built and the possible error sources are analyzed in the experiments and discussions.Based on the model,an anisotropic scattering shadow compensation method is proposed.The proposedmethod considered the influence of the visible factor of the sky and the sun position information and provided theoretical and methodological support for realizing the information restoration of the shadowed area.4.This paper uses two sets of high-resolution remote sensing image data sets.The shadow detection method and the shadow compensation method proposed in this paper are compared with two state-of-the-art shadow detection and shadow compensation methods.The shadow detection and compensation results were carried out in a comprehensive quantitative analysis.Compared with the other two methods,the proposed method can effectively avoid the influence of dark artificial objects(such as black vehicles),distinguish the water and superimpose on the water body while accurately detecting the typical shadows of buildings and the shadow of trees.The result of proposed shadow compensation method is the best result of information entropy in the three methods.It is shown that the information of the original remote sensing image can be improved by using the shadow detection and compensation method based on the surface elevation information proposed by this paper.