| Remote sensing image has played an important role in many fields,such as precision environmental management,agricultural land area detection and military investigation.Affected by environmental and technical factors,remote sensing images often lack of information.How to classify the remote sensing images which contain a lot of spatial information is a problem worthy of study.Aiming at the problem of insufficient spatial information of remote sensing image,this paper proposes a dual-mode remote sensing image information filling algorithm.This paper introduces the text file of mobile phone user visit information corresponding to each single-mode remote sensing image,analyzes and counts the time information in the text first,then clusters the time information,and saves the extracted time information into three dictionaries,then expands the boundary of single-mode remote sensing image,and finally fills the eigenvalues in the dictionary into the designated area to get the time information Bimodal remote sensing image.The dual-mode remote sensing image information batch processing filling system is designed.The filling system first extracts all the text time information of mobile phone users’ visit information text training set and test set,and saves them in the dictionary set with a unified naming method.Secondly,it completes the boundary zero filling for the single-mode remote sensing image training set and test set,and then fills the corresponding eigenvalues in the dictionary set into the single-mode remote sensing image Finally,the interactive interface of the batch processing system is designed to check the filling progress and verify the correctness of filling information.In this paper,we propose a criterion to determine the spatial information deficiency of remote sensing image.On the one hand,the information of single-mode and dual-mode remote sensing image is analyzed.When the mean and variance of dual-mode remote sensing image are greater than the corresponding single-mode remote sensing image,the image belongs to the remote sensing image with insufficient spatial information.On the other hand,the criterion is applied to the original single-mode remote sensing image data set to analyze the proportion of remote sensing images with insufficient spatial information.In this paper,RESNET,CNN,BP neural network,random forest and1d-cnn are used to classify the single-mode remote sensing image data set and the dual-mode remote sensing image data set.By comparing the classification accuracy of the five classification algorithms,the results show that the classification accuracy of the dual-mode remote sensing image in CNN network is the highest,increased by 4.6%.Thus,it is proved that the bimodal remote sensing image added with mobile phone users’ visit information text data is more effective,and the recognition degree and difference of remote sensing image with insufficient spatial information are improved. |
