Research On Shadow Detection And Removal Of High Resolution Remote Sensing Image

With the advancement of aerospace and photography technology,remote sensing has gained more and more attention.Remote sensing images has become the most convenient and convenient way to obtain ground information,and it is widely used in target detection and recognition,image segmentation and change detection.However,due to the light conditions and topography fluctuation,as well as the dense and tall buildings in the high resolution city image,the shadow is caused by the occulting relation.Although the shadow can help to establish the three-dimensional scene of the senses,but it might cause interference in the extraction of surface information and image interpretation,reduce the image quality,affect the accuracy of quantitative inversion,as well as details information extraction has negative impact.In order to eliminate the problems caused by the shadow in remote sensing image,it is necessary to restore the texture and color information of the shadow,so as to improve the efficiency of remote sensing image.We studied the theory and method of shadow detection and removal for single image remote sensing image,and divides remote sensing image into mountain region and urban area by terrain.In this paper,we respectively proposed the methods of shadow detection and topographic radiation correction of Landsat remote sensing images with DEM,and the detection and removal methods of high resolution urban remote sensing images without DEM support.At last,based on the theoretical foundation of deep learning,the application of shadow detection is carried out.This paper focused on the difficult problems in the detection and removal of high resolution remote sensing image.The main contents are as follows:(1)Review the current research of topographic radiation correction,shadow detection and removal,and summarize the existing methods and ideas.(2)Describe the causes and classification of the shadow in the mountainous area and study he mountain shadow detection,topographic radiation correction and shadow removal of Landsat satellite images is carried out with DEM(3)Research the shadow detection and coarse error elimination,analyze the deep learning methods in shadow detection,and study shadow removal methods combining with image segmentation,moment matching and local optimization(4)Through the experiment,the shadow detection and removal algorithm proposed in this paper is verified,and the applicability and limitation of the method are expounded.The experimental results show that the methods can improve the effect of the traditional terrain radiant correction.In the mountain area,DEM assisted methods can get more ideal shadow removal efficiency,in the urban area,the methods can get more accurate shadow detection results while combining with deep learning methods.The experimental results show that the methods proposed in this paper can achieve the shadow removal based on the four-band high resolution remote sensing image well.