In complex highway environments, accurate identification and priority handling of critical events are crucial for effective monitoring systems. However, traditional attention mechanisms lack dynamic adjustment capabilities, sometimes leading to oversight of important events. To address this challenge, an adaptive attention mechanism based on dynamic event prioritization is proposed to enhance event detection and response efficiency in highway monitoring systems. The proposed novel adaptive attention mechanism combines predictive and priority information, incorporating a priority assessment network to evaluate event importance in real-time and dynamically adjust attention weights. Experiments on real highway surveillance data demonstrate the mechanism’s effectiveness, significantly reducing false alarm rates and improving critical event detection speed.