Detection Method of Helmet Wearing Based on UAV Images and Yolov7

In power line construction working environment, the safety helmet plays an important role in protecting the safety of the workers from injury or reduce the injury, so safety helmet detection is important for power industry safety inspection. Detection of helmet based on UAV images is an efficient detection way. The traditional safety helmet detection algorithms have low accuracy and poor robustness. An novel detection method for construction helmet wearing in Yolov7 model is proposed in this paper. Yolov7 adopts extended efficient long-range attention network (E-ELAN), model scaling based on cascade model, convolution weight parameterization and other strategies. Based on the images captured by UAV, the experimental results show that the improved Yolov7 algorithm improves the average accuracy over the previous Yolo algorithms,which enhances the generalization ability in the scene of small targets, meets the performance requirements of helmet wearing detection in the power line construction site, and has a certain role in promoting the construction safety of the power industry.