Research On Fusion Of Optical Remote Sensing Data And Geochemical Data In Dacaotan Area,Eastern Tianshan,Xinjiang

As an important data,geochemical data are widely used in different fields,such as mineral exploration,environmental research,and resource potential analysis.However,Geochemical data is limited by accuracy in the analysis.In cases of harsh geographical conditions,geochemical surveys are difficult to carry out.To obtain large-scale geochemical data requires a lot of material resources.With the development of earth observation technology,the acquisition of remote sensing data is more convenient.Remote sensing data of high spatial resolution and rich spectral information can quickly and massively be acquired over a large area.Based on these advantages,remote sensing data have been widely used in the field of geoscience.Remote sensing data can reflect the absorption characteristics of electromagnetic waves by surface materials.This physicochemical relationship makes the remote sensing data and geochemical data fusion possible.Remote sensing and geochemical data fusion methods have low accuracy and few available fusion methods.For the above reasons,these methods are not suitable for practical production and application.In order to improve the accuracy of geochemical data,the fusion of remote sensing data and geochemical data is studied.In this paper,linear regression and logistic regression are used to realize remote sensing data and geochemical data fusion.This paper applies the deep learning theory,and puts forward a fusion optimization framework.The main research results are as follows:(1)A local linear regression fusion method is implemented.The linear regression model describes the physico-chemical relationship between geochemical data and remote sensing data in a local area.Remote sensing data and geochemical data fusion is realized in a subregion.The fusion results obtained by this method are better than the result obtained by global linear regression.(2)The fusion methods of local logistic regression and global logistic regression are implemented.They are based on Logical Regression Binary Classifier.Multiple logistic regression binary classifiers can be used to map remote sensing data and geochemical data.The results show that the global logistic regression fusion method is better than the local logistic regression fusion method.(3)A fusion optimization framework based on deep learning methods is established.This framework uses multilayer perceptron and convolutional neural network to realize remote sensing data and geochemical data fusion.The framework uses the fusion results of local linear regression and global logistic regression as the data source,the measured 1:50,000 geochemical data as the label.The framework by training multilayer perceptron and convolutional neural network,obtain the remote sensing and geochemical fusion results.The results matched with the actual geochemical distribution.which show that the fusion optimization framework improves the accuracy of the results.