Recognition And Extraction Of Specific Targets In Remote Sensing Images Based On Deep Learning

In recent years,with the rapid development of information technology and space technology,the number of high-resolution satellites has gradually increased,the resolution of remote sensing images has become higher and higher,and the amount of information has also increased.The importance of high-resolution remote sensing image target recognition and extraction has become increasingly prominent,and it has become an important subject of high-resolution remote sensing image application research.Traditional extraction methods rely on manual assistance and have poor adaptation to targets.Therefore,it is of great significance to study automatic identification and extraction methods for target identification and extraction of high-resolution remote sensing images.In this thesis,based on the method of deep learning,the identification and extraction of blue-roofed houses and open-pit mines in remote sensing images are studied.The main research contents are as follows:1.The dataset of blue roof room and open pit of high-resolution remote sensing images are constructed.In view of the fact that there are few public datasets for blue-roofed houses and open-pit mines,this thesis annotates remote sensing images with a resolution of 0.5m in this city and makes sample labels.And the data set was enhanced,and a high-resolution remote sensing image blue-roofed house and open-pit mine data set was established.2.In order to improve the accuracy of the model,the attention mechanism is introduced into u-net to make the network model more about the useful feature information and weaken the useless feature information,so as to strengthen the ability of the model to extract features and make the extraction effect of the model better,Compared with the related semantic segmentation methods,the feasibility and effectiveness of this method are proved.3.The identification and extraction of blue-roofed houses and open-pit mines of remote sensing images are realized by instance segmentation,and on the basis of the Mask R-CNN network model,the feature pyramid network is optimized by multi-path feature fusion,thereby improving the feature extraction ability of the algorithm and improving the segmentation accuracy.Comparative experiments verify the effectiveness of the optimization algorithm for remote sensing image recognition and extraction with complex background and dense target.4.In order to better meet the needs of users,a comprehensive use of Python programming language,Qt Designer and Py Qt5 is used to complete the development of a visualization platform for specific target recognition and extraction of remote sensing images,and the usability of the visualization platform is verified through testing.