Detecting　whether　workers　in　surveillance　images　and　videos　are　following　regulations　to　wear　safety　helmets　plays　a　vital　role　in　reducing　safety　accidents　on　a　construction　site.　Although　some　recent　algorithms　can　fulfill　this　task　with　the　help　of　emerging　deep　learning　techniques.　The　detection　accuracy　is　still　hard　to　meet　the　needs　of　real　applications.　So,　a　new　accurate　and　real-time　method　for　safety　helmet　wearing　detection　is　proposed.　Firstly,　the　YOLO　V5　network　is　improved　by　combining　implicit　and　explicit　information　to　enhance　the　context-aware　ability.　Then,　based　on　the　improved　network,　an　accurate,　fast,　and　stable　safety　helmet　wearing　detection　algorithm　is　proposed.　Finally,　a　series　of　experiments　are　performed　to　validate　the　proposed　algorithm.　It　shows　that　the　proposed　algorithm　outperforms　the　state-of-the-art　methods.　©　2022　IEEE.