Safety Detection Method for Work at Height Based on Improved YOLOv8

The electric power industry faces numerous risks and frequent accidents. The traditional safety management methods at the construction sites cannot satisfy the current demands. In order to deal with the safety issues faced by workers in the electric power industry while working at height, this study developed a safety detection method based on YOLOv8. Our team collected a large number of photos from electric power industry construction sites to build a comprehensive and effective dataset. These photos are intended to provide visual information about actual scenarios during the construction process. We took these photos at different locations and times to ensure the diversity and authenticity of the dataset. Our method combines the use of shuffle attention and Focal-EIoU loss function to enhance the performance of YOLOv8. The implementation of this method enables the effective and reliable monitoring of workers' safety hazard behavior, thereby enhancing safety measures. This innovative solution will significantly improve the safety of workers in high risk environments while reduce the occurrence of accidents.