Remote Sensing Images Understanding With Deep Learning

Deep learning(DL)neural network methods became a hotspot subject of research in the remote sensing filed.The deep learning neural network is a recent development that has become the subject of research in the computer vision and remote sensing disciplines.Super-resolution reconstruction refers to the technique of reconstructing a high-resolution image from a single or a series of low-resolution images by digital image processing.This technology can not only increase the high-frequency information of the image,but also eliminate the low-resolution.Different satellite data are used to predict the performance of each deep learning model.Deep Learning has made breakthroughs in modern digital image processing.Compared to traditional algorithms Bicubic and maximum likelihood(ML),a series of challenging image processing problems such as image classification and target detection needs to find a fast and reliable solutions so,our thesis will focus in the most important stages in remote sensing field after applying deep learning methods on it as follow:1.Super-resolution reconstruction refers to the technique of reconstructing a high-resolution image from a single or a series of low-resolution images by digital image processing.This technology can not only increase the high-frequency information of the image,but also eliminate the low-resolution.Deep Learning has made breakthroughs in modern digital image processing.Deep convolutional neural networks learn through a large number of training samples,obtain relevant information within the image,and then use the information to achieve specific functions.Super resolution(SR)images can be obtained using deep neural network methods that achieve a higher performance than all previous traditional methods.Here,in this study,we proposed an enhancement deep convolutional neural networks called(EDCNN)applied on SR satellite images which achieve superior performance overcome existing traditional and deep learning methods for SR satellite images such as bicubic,CNN and DNC.2.Classification of aerial satellite images depending decently on spectral content,which is a challenging topic in remote sensing.Traditional supervised classification such as maximum likelihood(ML)has a lot of disadvantages.With the aim of accomplish a high performance and accuracy of Egyptsat-1 satellite image classification,in this work we proposed to apply convolutional neural network(CNN)as first deep learning method used to get the classification on Egyptsat-1 Satellite images.CNN considered one of the famous deep learning methods which appear in this work.CNN developed to classify aerial photographs into land cover classes such as urban,vegetation,desert,waterbody,soil land,road,etc.In our work,a comparison between maximum likelihood(ML)and CNN method conducted and CNN outperform ML by 9%.SegNet used to enhance the performance of our work and it get better classification result than traditional Convolutional neural network.In addition,the experiments showed that SegNet is the most satisfactory and effective classification method applied to classify Egyptsat-1 satellite images.3.Change detection in remote sensing images is used to detect changes during different time periods on the surface of the Earth.Because of the advantages of Synthetic Aperture Radar(SAR),which is not affected by time,weather or other conditions,change detection technology based on SAR images has important research value.At present,this technology has attracted the attention of increasingly more researchers,and has also been used extensively in diverse fields,such as urban planning,disaster assessment,and forest early warning systems.Our objective in this work is to combine both the change detection of SAR images with the deep neural networks and proposed a new method that can compare its efficacy with existing methods.Our experiments,conducted on real data sets and theoretical analysis,indicates the advantages of our proposed method.Our results suggest that deep learning algorithms can further improve the change detection process.