Research On Thick Cloud And Shadow Removal In The Remote Sensing Image

Remote sensing image is vulnerable to the interference of cloud in the process of imaging, especially by thick clouds and thick cloud of remote sensing information shadow area is completely lost, reduces the efficiency of the use of remote sensing data, affect the accuracy of extracting thematic information of late. This is important for an accurate method to remove the thick cloud and its shadow. A thick cloud of remote sensing to the purpose of remote sensing information is effectively reduced by occlusion, and to ensure that the remote sensing data of other regions without distortion.A thick cloud of conventional detection method have the following methods:support vector machine, based on the histogram matching method, homomorphic filtering method, based on multi information fusion algorithm of remote sensing image, the algorithm of multi spectral information. In addition to cloud algorithm of wavelet transform, multiple images superposition method, the spatial structure of repair, based on machine learning methods. With the features of the algorithm and the time difference of different. Multiphase to method of thick cloud, is the use of the same area and the ground state and the little change of Landsat ETM in high resolution remote sensing images are fused using wavelet and support vector machine method to the cloud. In the cloud, or by using the linear regression method or by histogram matching method based on remote sensing images and spectral characteristics of phase contrast adjustment. The method is also based on the relative change of the spectral characteristics of images of different bands, enhanced model to design a thick cloud and its shadow region classification, combined with the model and the conventional classification model, detect the thick cloud and cloud shadow region.Guangzhou and Zhuhai area high resolution satellite remote sensing test data used in this paper are from the Landsat ETM shooting. In the experiment, large number of thick clouds exit in remote sensing image in Guangzhou, take on different phase image to a thick cloud of thick clouds and control way of shadow region detection using wavelet transform. One-temporal remote sensing image of Zhuhai also has thick clouds and shadows, by contrst with thick clouds area detection algorithm, thick control watershed method,shadow enhancement pattern method and improved method for shadow detection and shadow enhancement mode. And three remove shadows methods for effect comparison. The experiment proved that, the wavelet transform of the image in different phases, have very good effect to the cloud.In accordance with the thick cloud region shadow range search, canroughly detect thick cloud shadow region of the original remote sensing image, it can solve the confusion of water area and cloud shadow region problem. Improved shadow enhancement model shadow detection more effective than before improvement effect.It can solve the mixed of water region and shadow region problem and more effective than the before model.