Abstract:In worker casualties, head injuries account for the vast majority of casualties.Aiming at the problems of worker safety helmets with low correlation between traditional methods and detection scenarios and poor practicability, this paper proposes an algorithm that combines facial features and neural networks.The algorithm uses MTCNN to extract facial features and VGG convolutional neural networks to detect helmets.The detecting model is of small memory consumption high detection accuracy, and strong algorithm practicability.It can effectively supervise the wearing of helmets and provide workers with construction security guarantee.