DHANet: Dual-Stream Hierarchical Interaction Networks for Multimodal Drone Object Detection

Drone-based remote sensing has become pivotal for high-resolution dynamic monitoring. However, the differences between day and night modes will trigger a mismatch in multiscale object features under extreme lighting conditions. In this article, we propose a dual-stream hierarchical interaction network (DHANet) for multimodal drone object detection, called DHANet, which enhances the distinguishability between multiscale objects and background for each modality. Specifically, the DHANet is designed with a modality-adapted asymmetric attention module (M-AAM) that enhances object-level semantic representations through global and local attention mechanisms. The M-AAM employs global context attention and local positional attention to replace conventional multiscale context extraction, thereby effectively integrating spatial-channel information of objects. Furthermore, the network is equipped with a multimodal scale-attentive convolution (M-SC) module that dynamically generates modality-specific feature aggregation weights. This design enables global cross-modality information fusion while reducing computational complexity. Experimental results on two multimodal remote sensing benchmark datasets [DroneVehicle and vehicle detection in aerial imagery (VEDAI)] and two natural datasets (low-light vision visible–infrared paired (LLVIP) and FLIR) demonstrate the robustness and generalizability of the DHANet. The codes will be openly and freely available at https://github.com/Victoria-xin1009/-IEEE_TGRS_DHANet for the sake of reproducibility.