Research On High Precision Remote Sensing Image Road Recognition And Extraction Based On Deep Learning

Road information,as one of the important geographic information,has been widely used in many fields such as military,maps,transportation,navigation,and so on.With the maturity of related technologies such as carrier rockets and satellites,the recognition and extraction of road information in high-precision remote sensing images have received widespread attention and research.Remote sensing images are prone to interference such as clouds and fog,image distortion,and loss of ground features,which greatly affects the accuracy of road extraction.At the same time,the road information obtained by existing road extraction algorithms has problems such as detail loss,edge blur,and poor road network continuity.Therefore,this article focuses on the in-depth research of road extraction and recognition algorithms for high-precision remote sensing images in complex environments,with the following main tasks:(1)To address the problem of cloud and fog occlusion in remote sensing images,we proposed a remote sensing cloud removal algorithm based on generative adversarial network called CBAM-GAN(Generative Adversarial Networks Based on Convolutional Block Attention Module).The model mimics the human visual attention mechanism to identify and focus on the important features in the cloudy remote sensing image information,thereby enhancing important feature information and restoring cloud information to generate better quality cloud-free images.Compared with three existing cloud removal algorithms on the open source RICE dataset,the CBAM-GAN achieved the best performance in in terms of Peak Signal to Noise Ratio and Structural Similarity Index Measure.It is demonstrated that CBAM-GAN is effective in remote sensing cloud removal tasks,with good background restoration ability,color restoration ability,and detail preservation and restoration ability.(2)To address the problems of low road integrity,poor edge smoothness,and poor road network connectivity in road information extracted by existing road extraction algorithms in complex scenes,we propose a road extraction algorithm for remote sensing images,called AGD-Link Net(Link Net Based on Attention Mechanism,Gated Decoding Block,and Dilated Convolution),which integrates attention mechanisms,gated decoding block,and dilated convolution.This model mainly consists of three parts.Firstly,the Stem Block is used as the initial convolution layer of the model to reduce the information loss;Secondly,the series-parallel combined dilated convolution and coordinate attention block into the center of the network,which enlarges the receptive field of the network as well as improves the feature extraction ability of spatial domain and channel domain information;Finally,in the decoder part,gated convolution is introduced to improve the extraction the of the road edge.Compared with the three different road extration algorithms on the open source Deep Globe dataset,the proposed AGD-Link Net has improved the pixel accuracy,mean intersection over union and F1-Score index of road recognition by 1.41%-11.52%,0.0077-0.1473,0.0057-0.1292,and has certain effectiveness and feasibility in many scenarios in rural areas,urban,and suburbs.And can be apply to the tasks of road recognition and extraction in resolution remote sensing.