The rapid iteration and proliferation of micro-UAVs (<250 g) are reshaping security dynamics. In 2024, the global market for such drones exceeded $18 billion, with infrared obstacle avoidance and AI path-planning technologies challenging traditional defense systems. To address detection limitations in existing dual-network and target-tracking models—particularly underutilization of video sequence data and object features—this study proposes a Siamese Multi-Frame Tracking (SiamMFT) for infrared small-target tracking. SiamMFT enhances search-area features by aggregating multi-frame historical data and performs multi-template matching to refine twin feature maps. Experimental results demonstrate superior performance over state-of-the-art models in key metrics.