Adaptive Feature Fusion Network for Infrared Small Target Detection

Boyuan Li, Xiuhong Li, Songlin Li, Yuye Zhang, Kangwei Liu

Abstract: Infrared (IR) small target detection aims to separate small targets from complex backgrounds in infrared images, which has a wide range of applications in both military and civilian fields. However, IR small targets lack valid information such as shape and texture, which makes it difficult to effectively detect small targets in complex backgrounds. To address this issue, we introduce an innovative solution an Adaptive Feature Fusion Network (AFFNet). Initially, UNet serves as the backbone network for extracting features at different levels. Subsequently, the receptive field enhanced refinement (RFER) module is applied to emphasize the target and suppress background interference. Finally, a top-down adaptive feature fusion (AFF) module is used to generate the final prediction map. Experimental results validate the superior performance of AFFNet compared to other state-of-the-art methods in infrared small target detection.