Design of Foreign Object Intrusion Detection Model for Electrified Railway Contact Network Based on YOLOv5s and PP-LC
In order to detect foreign object intrusion in the contact network in a timely manner and provide efficient and accurate monitoring results for staff,a contact neural network foreign object intrusion detection model was constructed by combining YOLO series version 5 algorithm and lightweight CPU convolutional neural network.This model introduces a lightweight CPU convolutional neural network,attention mechanism,and pyramid pooling structure based on the traditional YOLO series version 5 algorithm,and improves its backbone network.At the same time,a new convolution module is introduced to improve the neck structure of the algorithm.The results showed that the average detection accuracy of the research design model was 95.74%,and in comparative experiments with other popular models,its detection speed was 0.012 s/frame,significantly faster than other models.This design model can achieve high efficiency in foreign object detection while ensuring detection accuracy.
万方数据
维普
