Recognition and classification of FBG reflection spectrum under non-uniform field based on support vector machine

Abstract

Abstract The reflection spectrum characteristics of fiber Bragg grating are very important for its sensing applications. A method of “Feature Extraction-Support Vector Machine (FE-SVM)” to identify spectral types is developed and experimentally demonstrated. The reflection spectrum characteristics of fiber Bragg grating are analyzed and extracted based on theory and simulation calculation. The characteristic data were preprocessed, and the distorted spectrum type recognition model was optimized. Training the data through the network, the recognition accuracy of Support Vector Machine (SVM) network for 1000 groups of FBG mixed spectrum reached 99.9%. To verify the recognition effect of reflection spectrum features, a time-varying temperature field was established as the non-uniform field. The accuracy rate reached 96.875%. The proposed FE-SVM method is characterized by fast response, high reliability and easy optimization, which has a promising application in environmental parameter measurement and substance classification.