Abstract: To address the challenges of inefficient cross-modal interaction and insufficient fusion in current multimodal remote sensing object detection for unmanned aerial vehicles (UAVs), this paper proposes a novel object detection algorithm named PCMFNet. First, a Progressive Cross-modal Consistency Enhancement module (CMCE) is designed to achieve feature alignment and capture complementary information by computing channel-wise consistency maps between visible and infrared modalities. Simultaneously, residual connections are utilized to preserve intra-modal discriminative features, thereby efficiently facilitating crossmodal interaction. Second, a lightweight Cross-modal Self-Attention Fusion module (CSAF) guided by fusion outcomes is introduced. This module employs deep infrared channel features as value vectors to model global contextual information, enabling effective cross-modal feature fusion and enhancing representation capacity. Experimental results demonstrate that the proposed method achieves significant performance improvements on two visible-infrared remote sensing object detection datasets, DroneVehicle and VTUAV-det. Specifically, it elevates mAP0.5 by 3.2% and 0.7%, respectively, over the baseline model, significantly enhancing UAVbased multimodal object detection accuracy in complex scenarios.