Technology Research On Shadow Detection And Compensation In Optical Remote Sensing Images

Sunlight is the only nature source which irradiates surface features from top to bottom,so the surface features with certain height will shelter the sunlight and form a shadow region.The imaging process of optical remote sensing image relates to the interaction between natural light and ground objects,therefore shadow become an unavoidable objective phenomenon in optical remote sensing image.On one side,the existence of shadow can provide information for the height measurement of ground objects and the reconstruction of ground objects.On the other side,the illumination condition in shaded region has changed,so the color of surface features in shaded region is different to the surface features in non-shaded region.This phenomenon will influence the post process,such as change detection,classification,dense matching and target recognition.To better utilize the information contained in the shadow areas of the optical remote sensing image,at the same time,as far as possible to reduce the negative impact on the image quality,this paper takes the visible light image obtained by the UAV platform as the research object,mainly studies the shadow detection and compensation technology of the urban area building,which can improve the detection success rate of the shadow detection algorithm to the easily confused area.In addition,the whole quality of shadow compensation is improved by analyzing and improving the existing shadow compensation algorithm.This paper's main work and innovations are as follows:1?Aiming at the problem of excessive noise and apparent void phenomena in pixel-based shadow detection results,the object-oriented shadow detection is realized by SLIC(Simple Linear Iterative Clustering)super-pixel segmentation algorithm.Because there is no proper judgment basis for the segmentation numbers in SLIC super-pixel segmentation,a method of estimating the number of hyper-pixel segmentation combining edge information is proposed.Based on the idea of initial segmentation optimization,the algorithm of edge information distribution in the initial cluster point grid is the basis for judging the number of segmentation,and it is simple and effective to estimate the number of segmentation required for the image to be fully segmented.2?In this paper,common detection algorithms based on shadow feature are researched.Firstly,the appropriate detection indices are screened to construct the initial test conditions,and then the dark area and bright shadow area of the initial detection result are extracted separately by using the NDUI detection index to construct the easy-confused region extraction condition.Next,the detection criterion based on spatial relation is used to distinguish dark region and bright shadow area,finally,the results of the preliminary detection and the discriminant results based on the spatial relation detection condition are finally obtained.The experimental results show that the algorithm can improve the detection success rate of dark areas and bright shadow region more effectively on the basis of guaranteeing the overall detection accuracy.3?Based on the research of common shadow compensation algorithms,an improving algorithm is proposed for the region compensation model.Aiming at the problem of insufficient compensation in the case of regional compensation algorithm without artificial optimization,compensating effect of the compound shadow areas with multiple objects,error sources are analyzed.The compensation formulas and the solution of compensating parameters are optimized.The experimental results show that the optimized regional compensation method can obtain a relatively good compensation effect without artificial optimization,and reduce the subsequent workload of manual adjustment.