Abstract: To ensure the safety of personnel, safety helmets must be worn at the construction site of the power plant. In order to find and correct the behavior of not wearing safety helmet in time, this paper studies and designs a safety helmet wearing supervision system based on the new automatic RPA. First, the centernet framework is improved, and the pyramid structure is added to reuse the feature map to improve the detection accuracy of the helmet. Then face recognition is performed on the area where the helmet is not worn to confirm the identity of the person. Finally, the information of the person will be informed to the site administrator by email, and the person will be alerted by audible alarm on the site, using the mode of online and offline double alarm. The experimental results show that the map value of the improved centernet on the helmet wearing data set reaches 83.84%, and the FPS on the geforce GTX 1050 graphics card reaches 26.71, which meets the basic standard and can effectively detect whether the helmet is worn.