Research on the Application of Data Augmentation with Large Model in Visual Monitoring of Transmission Lines in Complex Scenarios
Aiming at the low object recognition accuracy and high rates of false detections and false alarms are common in the visual monitoring of transmission lines under complex scenarios,an combined method that leverages the data augmentation by large model with optimized YOLOv10 is proposed.Firstly,data augmentation is performed based on the Stable-diffusion large model to address the shortage of available sample data,which can enlarge the number of samples.Secondly,based on the limited number of training samples,the YOLOv10 algorithm is optimized by further enhancing the image feature extraction algorithm and optimizing the object recognition algorithm to improve its accuracy and performance.The final experimental results show that,compared to existing method,the proposed approach significantly improves the target recognition accuracy for visual monitoring of transmission lines in complex scenarios,from an original 52.6%to 54.3%.
NETL
NSTL
万方数据
