Detection And Recognition Of New Buildings In Airport Clearance Area Based On High Resolution Images

Airports are one of the important public transportation infrastructures.In recent years,the continuous increase in the total number of airports has put forward higher requirements for the construction and operation management of airports.Clearance environmental monitoring is an extremely important part of airport security work.The detection and identification of buildings in the area is the key task of the research work on the protection of the airport clearance area.Due to the large clearance area of the airport,it is difficult for the traditional manual monitoring methods to meet the actual needs.Exploring the supporting role of high-scoring remote sensing technology in airport life-cycle management is of positive significance for improving the level of airport management.Traditional remote sensing image interpretation methods based on manual interpretation are difficult to efficiently cope with the huge amount of remote sensing data.The new wave of artificial intelligence since 2012 has provided sufficient theoretical foundation and technical support for the intelligent extraction of spatial information.Efficient intelligent interpretation methods of remote sensing images are becoming more and more important.Focusing on the practical application task of extracting new buildings in the airport clearance area,this paper proposes a new building extraction process based on highresolution remote sensing images to enhance the local semantic description capabilities of deep learning to improve the accuracy and efficiency of visual tasks.The remote sensing image change detection,sample expansion and scene classification methods based on deep learning can be summarized as follows:(1)Remote sensing scene change detection based on attention threedimensional convolutional neural network.This paper studies accurate and efficient change detection methods for complex airport environments,and proposes a semantic attention three-dimensional convolutional network for scene-level change detection in airport clearance areas.The network uses the classic Alex Net model as the main basic framework in an end-to-end form through multi-stage feature extraction to directly predict the change results from the input two-phase image blocks.The three-dimensional convolution structure reduces network parameters while improving timing The effect of image feature extraction and the addition of the semantic attention model is that the network can focus on target features,which can effectively highlight target scenes and objects in specific visual tasks,and enhance the expression of local change information.The self-made airport scene change detection data set was used to train and test the network,and it was applied to other remote sensing images around the airport for change detection,which verified the effect of the network model.(2)Remote sensing scene sample expansion based on deep convolutional generative adversal network.The scene classification algorithm based on deep learning requires a large number of images of different scenes as sample data.In order to solve the problem of insufficient sample images for deep learning scene recognition in this paper,how to use a limited amount of airport scene image data to expand to scale For the issue of a larger dataset with a unified style,this article uses a high-performance deep convolutional generation confrontation network as the basic framework,on which reasonable parameter adjustments and structural optimizations are carried out,so that it can better extract the local semantics of the image Features,strengthen the network’s learning of airport scene features in the original data set.Through the study of the feature distribution of the original airport sample data,the experiment generates a large number of images similar in style to the source data,completing the effective expansion of the small data set,and providing data support for other subsequent studies.(3)Remote sensing scene classification based on dense attention convolutional neural network.The complex scenes of the airport clearance area lead to the inundation of building information and the challenge of the accuracy of building extraction.In order to break through this limitation,this paper aims at efficiently and accurately extracting buildings from the airport environment.Deep learning scene classification method.This paper also proposes a Dense Attention Convolutional Network(DA-CNN).The network is organized by an intensive attention learning strategy,which provides stronger feature representation and transmission capabilities with fewer parameters and shallow depth through a dense connection structure;the intensive attention module is used to fuse salient features and volume features to enhance local scenes Theme expression;the channel attention module is introduced,which effectively avoids feature redundancy while compressing and optimizing convolution features.The performance of the network is tested through experiments on public data sets,and the method is used to further realize the detection task of newly added buildings on high-resolution remote sensing images of airports,fully verifying the effectiveness of the method.