The use of personal protective equipment （PPE） is crucial for ensuring worker safety in high risk environments such as industry and construction. However， traditional PPE compliance checks often rely on manual monitoring， which is inefficient and prone to errors. The article proposes an automated PPE detection system based on computer vision， aiming to improve the accuracy and efficiency of detection. The system uses the YOLOv5s deep learning model， combined with a large annotated image dataset， to perform real time detection of PPE such as safety helmets， reflective vests， protective goggles， gloves， etc. Through data augmentation techniques and transfer learning， the article significantly improves the detection performance of the model， providing an efficient and reliable solution for PPE detection with broad application prospects.