Remote Sensing Image Dense Target Detection Based on Multimodal Interactive-guided Network
Object detection in remote sensing images is the technical basis for risk assessment and rescue of large-scale crowded urban scenes.It faces such challenges as light interference,occlusion,density crowdedness and others.The existing work is mainly based on visible light monomode remote sensing data with a limited feature representation of dense targets and object separability.A multimodal interaction-guided network for remote sensing vehicle detection is proposed.By building a dual-flow network and a multimodal interaction-guided learning mechanism,the separability of dense vehicle targets can be significantly improved and the issue of poor detection performance of dense vehicles caused by small inter-class distance and large intra-class distance are addressed.The experiments on the Potsdam public datasets and DLR music festival data demonstrate the robustness and effectiveness of the proposed algorithm.
NETL
NSTL
万方数据
