Abstract: Addressing the issue of insufficient comprehension efficiency of safety warning signs in complex construction environments, the research constructs a multimodal assessment framework combining eye-tracking and subjective questionnaires, and employs integrated analysis of both to systematically quantify the patterns of attention allocation toward warning signs in real-world scenarios such as confined underground spaces, open surface areas, and light-dark transition zones. The research finds: warning signs placed centrally in narrow passages significantly enhance visual capture efficiency; a layout at heights of 5-7 meters in surface construction zones can balance the conflict between occlusion and visibility; visual capture intensity is significantly lower in dim environments compared to bright ones, while bright areas within light-dark transition zones attract initial fixation faster. Interference from LED signs leads to a significant attenuation in the cognitive depth of warning signs, with total fixation duration reduced by 71% and fixation count decreased by 73%, revealing their attentional resource competition effect and compensating for the inadequacy of existing research that predominantly focuses on static sign characteristics. Therefore, for complex water conservancy and hydropower construction environments, establishing context-adapted safety warning strategies characterized by "centrally in narrow spaces, lower in open areas, medium-high on surfaces, and proximity to light sources while preventing interference" can effectively enhance workers' safety awareness and reduce accident risks.