Abstract:The sparse characteristics of target features poses significant challenges when using deep learning methods for infrared dim small targets. To tackle this issue, this paper proposes a novel multi-level sparse feature fusion network (MSFFNet) for detecting infrared dim small targets. A feature-level sparse feature fusion network fuses target features of the same level and different depths to express small target features. A decision-level sparse feature fusion network fuses features from different decision spaces to improve decision confidence. To enrich the feature representation of the target, different levels of target global features are introduced into the decision-level sparse feature fusion network. During the network training process, a deep joint supervision training strategy is proposed to supervise and train the multi-level sparse feature fusion network, aiming to fully learn the feature representation of the target. According to the experimental results, the proposed infrared dim small targets detection method outperforms existing popular methods under sparse target features.

MSFFNet: A Multi-Level Sparse Feature Fusion Network for Infrared Dim Small Target Detection