Research On Semantic Segmentation Of High-resolution Remote Sensing Image Based On Deep Learning

With the rapid development of satellite remote sensing technology,massive remote sensing image data has also come.How to find useful information in a large amount of data is a serious challenge for researchers.Artificial intelligence technology is considered to be one of the three cutting-edge science and technology in the 21 st century.Its rapid development provides a new solution to the intelligent and fast processing of remote sensing image data.The high-resolution remote sensing image records detailed feature information,including the geometric shape,geometric features,texture and other characteristics of the target.Its application fields include map drawing,agricultural testing,disaster warning,urban planning and many other directions.Therefore,how to realize the efficient and intelligent information extraction of high-resolution remote sensing images is an important problem that needs to be solved urgently.At present,the traditional commonly used methods in dealing with high-resolution remote sensing images are mainly machine learning methods,or a combination of machine learning and artificial methods,which not only consume a lot of manpower and material resources,but also difficult to achieve the desired effect.In this paper,firstly,the full convolutional neural network method is used to semantically segment the high-resolution remote sensing image,and the matrix expansion technology is combined to improve the training speed and realize the effective semantic segmentation function.In order to solve the problem of misjudgment and limited precision improvement caused by the uneven classification of sample data,the polar coordinate transformation method is used to enhance the data set to increase the amount of data and improve data diversity.Finally,aiming at the limited problem of precision improvement of the full convolutional neural network,the method of training and learning using DeepLabV3+ model on the enhanced dataset is proposed.The good segmentation effect is obtained,and the effectiveness of the proposed method is improved.The segmentation accuracy of remote sensing images has certain reference significance.