Research on intelligent storage monitoring of power grid based on YOLOv5 converged attention
Xiaodong Tu, Zhan Gao, Wang Liujun, Yong Wang, Chai Lianxin · 2023
With the promulgation of Made in China 2025, as an important support for the realization of industrial informatization, smart warehousing has become the focus of current enterprises and scholars. In this paper, aiming at the real-time monitoring of large cargoes in the grid storage scenario, based on the target detection network YOLOv5, combined with the ECA channel attention mechanism and SIoU loss function, a real-time and efficient cargo detection method based on YOLOv5 fusion attention is proposed for the grid operation site. In order to alleviate the imbalance between positive and negative samples, make full use of the information of difficult samples and improve the detection effect of difficult samples, Focal loss function is introduced. At the same time, on the basis of the original SPP multi-scale fusion, the improved YOLOv7 multi-scale detection method was fused. In view of the lack of power cargoes data sets at home and abroad, we have built a small but high-quality power cargo datasets. The experiment shows that the method in this paper has significantly improved in accuracy, speed, accuracy and generalization ability, and can achieve effective detection in complex operation scenarios, which can provide technical reference for the safety of power grid operators and reduce the intensity of human monitoring.