Dual-model hybrid pattern recognition method based on a fiber optic line-based sensor with a large amount of data

Abstract

A dual-model hybrid pattern recognition based on a fiber optic line-based sensor with a large amount of data is proposed. The vibration signals are converted to gray-level images to reduce the memory requirement. The ResNet18 model for classification is used. To reduce the false positive rate, the over-zero rate and short-time energy are extracted from the intrusion signal, and a support vector machine (SVM) is used. Finally, a discriminator is constructed to determine the types of events by combining the two models trained on the validation dataset. The results demonstrate the excellent average recognition accuracy of this method, which achieves the 97.1% for six events.

Intrusion signal recognition method based on φ-OTDR fiber distributed sensing research progress

“…In 2022, Cheng 63 et al proposed a dual-mode hybrid pattern recognition method based on fiber optic wire sensors with a large amount of data.The authors converted vibration signals into grayscale images for the classification of the ResNet18 model.To reduce the false alarm rate, the excess zero rate and short-time energy were extracted from the intrusion signal, and support vector machine (SVM) classification was used.By combining the two models trained on the validation dataset The two models trained on the validation data set to construct discriminators to determine the event types, and it is shown through experiments that the method can achieve 97.1% recognition accuracy for six types of events: blowing, shaking, shearing, knocking, walking, and tree tapping.…”

Feature fusion-based fiber-optic distributed acoustic sensing signal identification method

“…In contrast, the neural network structure through deep learning can automatically learn more feature information, and its powerful learning capability makes it possible to express complex functions through multiple neural network layers and extend new computational spaces to solve a series of complex problems. Therefore, deep learning using convolutional neural networks (CNNs) can be effective for intrusion signal classification, and several studies have focused on event recognition in fiber-optic distributed sensors using techniques from neural networks [17][18][19][20][21]. However, most scholars have used two-dimensional (2D) convolution kernels to construct convolutional layers to obtain correlations between the duration and frequency of the signal to improve recognition accuracy.…”

“…However, most scholars have used two-dimensional (2D) convolution kernels to construct convolutional layers to obtain correlations between the duration and frequency of the signal to improve recognition accuracy. Therefore, to deal with one-dimensional (1D) sensing signals, scholars always convert the original signals into 2D images or data matrices by time-frequency analysis; however, this leads to an increase in computational complexity and a decrease in recognition accuracy [17][18][19]. By contrast, a 1D convolution can describe the temporal evolution of the signal more intuitively, and it is easier to determine the characteristics of the signal in a specific time period and facilitate the comparison of the differences between different signals.…”