Study On Rejection Of Shadows From Urban High Resolution Remote Sensing Image

In recent years, because of the fast improvement of the technology about remotesensing, the spatial resolution satellite sensors are constantly improving, theapplication of high resolution remote sensing image is becoming more and morewidely, mainly including topographic map drawing, change detection, digital cityconstruction, etc. However, the existence of the shadow will process a lot of adverseeffects to the result of the high resolution remote sensing image, such as imagematching, object recognition and extraction and so on. Therefore, the accurateextraction of the building shadow from high resolution remote sensing image and toget rid of it, it is one of the most important work in terms of dealing with highresolution remote sensing image.Based on the systematic research about the detection and removal of shadow athome and abroad, we found that the existing algorithms have many limitations, andthe processing result is not very ideal, there existing high rate of false detection.Aiming at these problems, this paper improved the existing algorithms, and putforeword on the new algorithms. The main works are described as follows:(1) This paper studied the characteristics of the shadow and took preprocessingoperations to the image. Details of the shape, type and spectrum characteristics of theshadow, the remote sensing image carries on some pretreatment, such as contrasttransform and median filter, it takes much convenient to handle the subsequent of theshadow.(2) The algorithm based on morphology is proposed to detect the shadow. Bythe means of top-hat transformation and low cap transformation with the remotesensing image, as a result, it enhances the contrast of the image. Using amorphological method and choosing a high threshold to extract the shadow area ofthe objects.(3) The region growing segmentation method is proposed to detect the shadow.After the pretreatment of the remote sensing image, then converting it to gray image,and normalize the histogram, combining with the regional growth function, bymeans of selecting a appropriate threshold, to accomplish the extraction of theshadow of the objects. The experimental results show that this method can accuratelyextracted the shadow of the objects.(4) Putting forward a method of detection and removal the shadow of the tree shadow among the shadow of the objects. By constructing the Lab and HSI colormodel to test the trees, after that, we take some experiments to remove the shadow ofthe trees. The final results show that the Lab color model of trees can well detect andremove the trees in the objects ares from the high-resolution multispectral remotesensing image.(5) Based on the analyzing and summarizing of the existing algorithm forbuildings’ shadow removal. We improve the algorithm for Wallis filter, and putforward a kind of shadow removal algorithm which combining with the theory ofcolor constancy. Through the experiment and quantitative evaluation, these twoalgorithms are verified by experiments that they can get a great improvement inaspect of precision than the traditional algorithm for shadow removal.