Unsignalized intersections are accident-prone locations due to numerous conflict points and unclear right-of-way. This issue is exacerbated at night when road markings become less visible, leading to increased accident rates. Self-luminous road markings, a new type of proactive traffic safety control facility, have garnered increasing attention and are being gradually promoted due to their intelligent, stable brightness and variable characteristics. To explore the warning effect of self-luminous road markings at unsignalized intersections at night, this study designed three types of warning schemes: continuous-illuminating pedestrian crosswalk advance warning marking (CPWM), continuous-illuminating yield text advance warning marking (CYWM), and transition-illuminating pedestrian crosswalk advance warning marking (TPWM). Based on previous research, nine indicators were selected for comprehensive evaluation from the perspectives of driver's operating, visual characteristics, and psychology. Finally, an entropy-based matter-element model was constructed to comprehensively evaluate the warning effect. The results show that the CYWM and CPWM schemes can help drivers effectively reduce the mean speed and potential lateral conflict risk at intersections, enhancing drivers' awareness of road environment information. However, the TPWM scheme had adverse effects due to greater visual stimulation. The effectiveness of the three schemes ranks from high to low as CYWM, CPWM, and TPWM. This study confirms that self-luminous road markings are effective at intersections, providing new insights for infrastructure upgrades and offering a general framework for evaluating the effectiveness of traffic safety facilities.