Research On Nondestructive Testing Based On Machine Learning

Near and middle infrared detection technology is to obtain the spectral information of the measured object through transmission or diffuse reflection,from which the internal quality of the measured object can be reflected.Preprocessing is the premise and basis of spectral analysis based on machine learning.Its function is to eliminate irrelevant or redundant features and highlight absorption peaks related to the physical and chemical properties of the object under test.By combining SNV and PCA to process spectral features,the complexity of subsequent spectral modeling can be effectively reduced and the error of predicting its physical and chemical properties can be reduced.Due to the influence of environmental factors when actually collecting spectra,the adoption of the linear model has a large error in spectral feature learning.Therefore,the introduction of the Xgboost algorithm into the nir nondestructive testing technology can effectively improve the prediction results of the linear mode.Finally,the prediction effect of the proposed method and PLSR method on fruit sugar degree under different pretreatment methods was compared.Experiments on 196 apple samples demonstrate the effectiveness and efficiency of the proposed method.By comparing root mean square error,it can be found that SNV-PCA combined with Xgboost algorithm is more effective than the traditional MSC combined with PLSR method.