Research On Spectral Classification Method Based On Few-shot Learning ——Taking The Xingcheng Area Of Liaoning As An Example

Rock spectrum is the comprehensive embodiment of rock physical and chemical properties,composition and structure.Now it has been widely used in the research of rock classification.Due to the difficulty of collecting the data of rock spectrum,it often needs to be collected manually,which not only causes great labor cost,but also leads to the limited data of rock spectrum collected.When the rock spectral classification model is trained with a limited number of training samples,the Curse of Dimensionality will occur,which means the accuracy of classification will decline with the increase of feature dimensionality,and the data of the rock spectrum just conforms to this characteristic which has a highdimensional feature number.Therefore,in order to achieve great classification results,a large number of training samples need to be used in the training model of traditional rock spectral classification model,which is usually multiple times than the feature dimension.If the number of samples is small,we need to reduce the dimension of the feature to get the ideal classification accuracy.Therefore,when the number of samples is small,how to obtain more accurate classification effect on rock spectral data has become a hot research topic nowadays.This paper collects the spectral data of typical rocks in Xingcheng,Liaoning Province.Based on python programming language,four traditional machine learning methods which include Decision Tree,Random Forest,Support Vector Machine and K-Nearest Neighbor,are used to establish the classification model in the case of little training samples.By drawing the learning curve,it is verified that these four traditional machine learning methods do not have good classification function in the case of small samples.Then the Few-shot Learning method and Siamese Network are introduced in detail,and the Siamese Network classification model is established.Taking Triple Loss as the loss function,the 3-way1-shot classification model is realized,and the prediction accuracy of 97.8% is achieved in the verification set.Finally,we use the trained classifier to build a rock spectrum classification system with extended function.