Research On High Resolution Remote Sensing Image Artificial Target Recognition Algorithms Based On Sparse Representation

With the rapid development of aerospace technology and sensor technology,satellite remote sensing images have been widely used,especially the high spatial resolution remote sensing image has become an indispensable source of data for civil and military applications.At present,high-resolution remote sensing images have been widely used in many fields and have obtained some achievements on applications,such as urban planning and environmental monitoring,natural disaster monitoring and emergency treatment,ground observation system,military monitoring and investigation,land use and Adaptability evaluation.The spatial information of the feature in the high resolution remote sensing image is relatively complete and clear,and can provide the massive data information for the above application.However,its massive data information,feature shape,spatial structure and complex structural features bring new challenges to the automation and intelligent processing of remote sensing data.The traditional remote sensing image processing technology far cannot satisfy the requirements of the current high score remote sensing image application,therefore,it is particularly important to explore a new method.In order to solve the above problem,this paper introduces the sparse representation theory of signal,studies and discusses the sparse representation theory and low rank theory deeply.The main contributions of this paper are as follows:(1)The sparse representation theory is deeply studied and discussed.According to the study of biological vision,the sparsity of high-resolution image is studied and discussed,we apply the sparse representation theory to high spatial resolution remote sensing image,an artificial target recognition method of high-resolution remote sensing image based on sparse representation is proposed,and a series of experiments are conducted on Uc Merced and Wh U datasets to verify the effectiveness of the proposed method.(2)The interference of clouds in remote sensing images is considered,which leads to the degradation of image quality and effects the subsequent image processing.According to the characteristics of the cloudy cloud in the remote sensing image with excessive slowness,uniform distribution and strong self-correlation of the spatial texture structure,we assume that the thin cloud information is low rank.Based on this,a thin cloud elimination method of remote sensing image based on low rank matrix decomposition is proposed,which can effectively eliminate the thin cloud of the remote sensing image for subsequent target recognition.Experiments show that this method is efficient.(3)Considering the sparsity and the low rank of artificial target,combined with the theory of low rank matrix decomposition,a new target recognition method is proposed.Firstly,the method performs low rank matrix decomposition on remote sensing image,the two parts of the sparse information are obtained,then by the K-SVD algorithm,the dictionaries of the two parts are studied,which are combined to build he complete dictionary of sparse representation,finally through the sparse representation algorithm,we obtain the sparse coefficient of the classification target in the complete dictionary,and classify the target according to the maximum criterion of the sparse coefficient.Similarly,the corresponding experiments are performed on Uc Merced and Wh U datasets.The experimental results show that the proposed method is efficient.