Fiber optic cable event recognition based on dual-channel GAF and dendrite networks

In order to solve the problems of complex signal feature extraction and low recognition accuracy in traditional optical fiber optic cable event recognition, this paper proposes a method for optical fiber optic cable event recognition using Gramma Angle fields (GAF) and dendrite network (DD). First, one-dimensional time series fiber optic cable event signals are normalized and encoded by polar coordinates. Then, dual-channel GAF method is used to retain the signal timing features and map them into two-dimensional images to form a dual-channel image training set with obvious features. On this basis, DD network is used to extract features and construct an event recognition model. The experimental results show that the classification accuracy of this method is 98.67%, which is 5.46%, 7.54%, 6.33%, and 1.55% higher than other methods such as KNN, SVM, ELM, and CNN, respectively. Therefore, the recognition accuracy of the DD model is higher than other methods, which verifies the accuracy of this fiber optic cable event recognition method.