Infrared small target detection from complex backgrounds is increasingly vital for military and civilian fields. Nonetheless, most of the existing methods are too restrictive to portray infrared targets from multidimensional and omnidirectional. In this article, we propose a low-rank differential cascade network (LDCNet) to integrate the physical properties and deep cascade features of infrared images. First, the cascade feature extraction module is designed via a multilevel coplanar cascade encoder–decoder structure, which integrates the deep-level and low-level features of infrared targets and backgrounds. Then, to provide a better understanding of the context capture of the scene, the differential attention mechanism based on the change differential analysis and robust principal component analysis is introduced. Finally, the multilevel feature fusion module is designed to adaptively integrate the spatial and semantic information of different depth feature maps to predict the final detection result. During the research, a new maritime small targets detection dataset is also constructed. Experimental results compared with other related methods on three datasets have demonstrated the effectiveness of LDCNet.

Infrared small-target detection is now commonly used in maritime surveillance, flight guidance, and other fields. However, extracting small targets from complex backgrounds remains a challenging task due to the small-target scale and complex imaging environment. Many studies are based on designing model structures to enhance the precision of target detection, and the number of Params and FLOPs has been significantly augmented. In this work, a knowledge distillation-based detection method (KDD) is proposed to overcome this challenge. KDD employs the small-target labeling information provided by a large-scale teacher model to refine the training process of students, thereby improving the performance and becoming lightweight. Specifically, we added efficient local attention (ELA), which can accurately identify areas of interest while avoiding dimensionality reduction. In addition, we also added the group aggregation bridge (GAB) module to connect low-level and high-level features for the fusion of different feature scales. Furthermore, a feature fusion loss was introduced to enhance the precision of target detection. Extensive evaluations have demonstrated that KDD performs better compared to several methods, achieving extremely low Params and FLOPs, as well as higher FPS.