Abstract: To address the missed-trigger problem of traditional Automatic Emergency Braking (AEB) systems at unsignalized intersections caused by the uncertainty of the target vehicle’s maneuver intention, an intention-recognition-based AEB control strategy, namely Intention Recognition Automatic Emergency Braking (IR-AEB), is proposed. At the perception and decision-making level, a target-vehicle intention recognition model is developed based on naturalistic driving data. Combined with intersection topology, multi-branch candidate trajectories are generated, and an intention-probability-weighted Expected Time-to-Collision (ETTC) metric is further established for collision risk quantification. At the control execution level, a dynamic primary-risk-target selection mechanism is introduced to handle multi-target interaction scenarios, and a braking trigger method combining fixed time thresholds with hysteresis logic is designed. The proposed strategy is validated on a MATLAB/Simulink closed-loop simulation platform under the Car-to-Car Front Turn-Across-Path (CCFT) scenario specified by C-NCAP, as well as typical two-vehicle conflict scenarios and extended three-vehicle interaction scenarios. The results show that the intention recognition model achieves a classification accuracy of 95.84% and a macro-average F₁ score of 0.955. Even when the maximum deceleration of IR-AEB is limited to 5 m/s² while the baseline TTC-AEB is allowed to use a higher braking limit, the proposed strategy still achieves earlier trigger timing and a higher collision avoidance success rate. In addition, in complex three-vehicle scenarios, it can stably realize accurate locking and dynamic switching of the primary risk target, thereby effectively improving scenario adaptability and safety.