Lithological Classification Using Remote Sensing Image And Accuracy Assessment Combined Sample Purification

Using of remote sensing images for lithological classification and identification is one of the important applications of remote sensing in the geological field.The traditional research of remote sensing petrology is mainly based on the spectral characteristics of rocks to identify lithology.With the continuous deepening of the application of remote sensing technology in lithological identification,the accuracy of lithological information extraction based on spectral features has been difficult to achieve the application requirements.The texture information has been used as auxiliary information to improve image classification accuracy for lithological recognition.Window size is an important parameter for texture extraction,which will affect the extracted texture results.On the other hand,in the process of lithological classification of remote sensing images,classification accuracy evaluation is an indispensable link in remote sensing classification technology and has important significance for application in remote sensing classification.In assessing the classification accuracy,the test sample is usually regarded as true or accurate.However,the existence of mixed pixels in image usually causes errors in test samples.Such errors are often simply ignored and attributed to classification errors.Therefore,the result of the accuracy evaluation is unreliable.For the two problems above,this paper proposed a lithological classification method combining the optimal texture window size selection and test sample purification.Firstly,the optimal texture extraction window size was pre-estimated based on semivariogram.Secondly,optimal extraction window size and four features of the Gray Level Co-Occurrence Matrix including Angular Second Moment,Entropy,Correlation,and Contrast were used to express image texture.Thirdly,Optimal Index Factor(OIF)was used to select optimal feature combination.Fourth,a support vector machine(SVM)classifier was employed to classify different combinations of features.Fifthly,using the proposed sample purification method and texture features of image,sample purification rules were established based on attribute coherence to remove the test sample points that conflict with the rules,and the purified test samples were obtained.Sixthly,the effectiveness of the semivariogram-based optimal texture extraction window selection method was verified by lithologic classification based on the Angular Second Moment of different window sizes combined with spectralfeatures.Finally,the accuracies between different combinations of classifications were assessed by test samples with and without sample purification.Experimental results show that the pre-estimated texture window size can guarantee a classification result with high classification accuracy for lithological classification.The results also demonstrated that the accuracy of lithological classification based on spectral features and ASM textural features was the highest.The overall lithological classification accuracy and kappa value,without sample purification selected by stratified sampling,were respectively 87.4% and 0.84,however those with sample purification were respectively 88.01% and 0.85.The results show that the proposed method is capable of yielding more reliable lithostratigraphic identification.