Phase-OTDR event detection on Phase-OTDR
99.74AccuracyCheng et al.
Evaluation results
| Method | Links | |
|---|---|---|
| Cheng et al.Preprocessing Method=TF transformation, masking, pre-training, global-local feature fusion, DL Model=Φ-GLMAE (Global-Local Masked Autoencoder), Num. of Classes=6, Pre-training configuration=pre-train2025.12 | 99.74 | |
| Li et al. (2025)Preprocessing Method=Long sequence classification, 10,000×12 sequence format, DL Model=DASFormer (Transformer-based), Num. of Classes=62025.12 | 99.6 | |
| Cheng et al.Preprocessing Method=TF transformation, masking, global-local feature fusion, DL Model=Φ-GLMAE (Global-Local Masked Autoencoder), Num. of Classes=6, Pre-training configuration=without pre-training (w/o)2025.12 | 98.97 | |
| GADF-GASF-RP Image-Based TransformationPreprocessing Method=GADF, GASF, RP -> RGB Image Transformation, DL Model=8 Different Transfer Learning Models, Num. of Classes=6, Evaluation Protocol=Holdout2025.12 | 98.84 | |
| Duan et al.Preprocessing Method=1D CNN, Modified Transformer Encoder (ISAT), Dropout, Batch Norm, DL Model=ISAT (Inter-sequence attention Transformer), Num. of Classes=62025.12 | 98.8 | |
| Chen et al.Preprocessing Method=Normalization, differentiation, spatio-temporal feature extraction, DL Model=Dendrite Net (DD) + VTTCCG, Num. of Classes=62025.12 | 98.6 | |
| Wang et al.Preprocessing Method=Min-Max normalization, FEA-TFW-LDA, DL Model=SVM, RF, KNN, NB, Num. of Classes=62025.12 | 98 | |
| Luo et al.Preprocessing Method=Downsampling, normalization, KD (knowledge distillation), DL Model=4-layer CNN (student), ResNet-34 (teacher), with KD, Num. of Classes=62025.12 | 97.92 | |
| Hu et al. (2025)Preprocessing Method=Gaussian noise addition, data reshaping, spatiotemporal maps, DL Model=SDENet, Num. of Classes=62025.12 | 97.7 | |
| Li et al. (2024)Preprocessing Method=Spatio-temporal feature extraction, channel and spatial attention mechanism, DL Model=MT-ACNN-SA-BiLSTM (Semi-Supervised), Num. of Classes=62025.12 | 96.9 | |
| Sasi et al.Preprocessing Method=Feature selection, hyperparameter optimization, DL Model=LightGBM, Num. of Classes=62025.12 | 95 | |
| Cao et al.Preprocessing Method=Spatio-temporal matrix normalization, feature extraction (32 features), DL Model=SVM, CNN, Num. of Classes=62025.12 | 94 |