Research On Remote Sensing Image Retrieval Method Of Urban Area Based On Multi-feature Fusion

In recent years,with the continuous development of remote sensing technology and the rapid growth of massive image information,people can now get the information they want through remote sensing data.Remote sensing image retrieval plays a more and more important role in obtaining a large number of remote sensing data information,which is one of the hot spots of current research.However,at present,the research of remote sensing application still needs to be improved.It is unable to extract useful information accurately and effectively,resulting in a great waste of resources.Therefore,how to retrieve effective information efficiently and accurately from a large number of remote sensing images is the main problem to be solved.Urban area is an important part of land management,and it is closely related to people's life.People's life scope is basically in urban area,so it reflects the relationship between the current situation of land management and people's daily life.Now urban planning is inseparable from the acquisition of remote sensing image information.How to accurately extract the regional distribution and layout of each city is the focus of relevant departments.At present,urban planning methods can be divided into different categories according to the different functions,but the research on each urban area is still insufficient.Starting from the practical application of urban regional sense image data retrieval,this paper proposes a multi feature fusion method for remote sensing image retrieval.Based on the existing research on remote sensing image retrieval,this paper makes a detailed study on the feature extraction method,retrieval algorithm and feature fusion method of remote sensing image.The main work of this paper is as follows:(1)Analyzed the characteristics of remote sensing images.Considering the complexity of remote sensing images,a single feature alone could not represent all the characteristics of remote sensing images.Therefore,a variety of feature fusion methods were proposed to attempt remote image retrieval.Global features and local features can complement each other's advantages,greatly improving the description ability of remote sensing images,and effectively enhancing the retrieval effect of remote sensing images.(2)Choose a suitable feature fusion method.When there are appropriate multiple feature descriptions,a suitable method is needed to fuse the extracted feature descriptions together.Therefore,the concept of multi-core learning is introduced in this paper.Multi-core learning is used to fuse multiple features.Retrieve remote sensing images.Multi-core learning is an extension of single core learning,which can obtain the optimal results through continuous training learning according to different needs.In this paper,multi-core learning method is used for feature fusion,which combines global features and local features,and improves the accuracy of retrieval through the complementarity between features.(3)The global Gist feature and the local SIFT feature are specifically selected as the remote sensing image features for fusion,and the Gist features of the remote sensing images of all urban areas in the sample database are extracted and saved as the Gist feature database.Subsequently,Scale Invariant Feature(SIFT)is further introduced to construct a Bag-of-Features(Bo F)model and a Bo F feature vector library is obtained.The generated two features are combined into a synthetic kernel according to a linear addition method,and a classifier is trained by a multi-core learning method to obtain a final retrieval result.