Guozhu Yang, Wei Du, Huamin Sun, Shirui Sun, Pengchao Sun
Transmission line safety; Real-time detection of wildfires; Predictionof wildfire spread; Lightweight; YOLO-LNet State Grid Power Space Technology Co., Ltd, Beijing, 102211,China; e-mail: [email protected], [email protected],[email protected], [email protected], [email protected] author: Guozhu Yang
Traditional methods have a limited monitoring range and delayed response in detecting wildfires on transmission lines, and existing image recognition methods are mostly limited to static fire source identification and lack dynamic prediction capabilities. A real-time prediction model based on a lightweight YOLOv5 detection mod- ule and a Bi-LSTM prediction module is introduced in this study. In terms of object detection, coordinate convolution and the Ghost module are introduced to improve the backbone network and neck structure of YOLOv5. Focus loss function and complete intersection- to-union ratio loss function are used to improve classification loss and achieve model lightweighting. In terms of spreading prediction, Bi- LSTM is employed to model the bidirectional dependence of temporal data, such as fire point location and fire area. The results indicate that the improved lightweight YOLOv5 model [email protected] Reaches 74.3%, which is 4.1% higher than traditional YOLOv5s and better than mainstream models such as EfficientDet (68.5%) and MobileViT (70.6%); The computational complexity has been reduced by 48%, and the model size has been reduced to 6.5MB. The complete YOLO LNet prediction model not only has the lowest error in different sce- narios, but its prediction accuracy (MAE: 5.23) and speed (43.8ms) are also superior to comparative models such as Seq2Seq and Trans- former, demonstrating significant comprehensive advantages. This model provides a high-precision and light-weight solution for real- time detection and dynamic prediction of wildfires in transmission lines, and offers new technical support for disaster prevention and mitigation of transmission lines.
Important Links:
Go Back
