Forest fires are characterized by large-scale,dynamic,and unstructured spread,whereas high-rise building fires occur in vertical,multi-story environments with severe occlusion.Conventional fixed monitoring systems and ground-based rescue operations often struggle to simultaneously achieve rapid situational awareness,continuous tracking,and safe operation.Based on review,classification,comparison,and synthesis of the literature,this paper summarizes recent advances in key unmanned aerial vehicle(UAV)technologies for complex fire scenarios,including UAV platform and payload matching,fire detection,fire-source localization,fire-spread prediction,path planning,multi-UAV coordination,and emergency supply delivery.The review indicates that fire perception is evolving from traditional image-feature-based methods toward lightweight deep learning and visible-infrared multi-modal fusion,fire-source localization is progressing from discrete coordinate estimation to continuous fire-state estimation,risk prediction increasingly incorporates real-time observational feedback,and path planning increasingly adopts online replanning.Nevertheless,existing studies remain constrained by limited real-fire data,insufficient cross-scenario generalization,fragmentation among perception,prediction,and planning,and inadequate field validation under high temperatures,dense smoke,strong winds,and unreliable communications.Future research should prioritize integrated perception-prediction-planning frameworks,autonomous coordination of heterogeneous UAVs,and explainable decision-making under uncertainty,thereby improving the autonomy,reliability,and real-world deployability of UAV systems in complex fire environments.