In order to solve the problems of infrared target detection (i.e., the large models and numerous parameters), a lightweight detection network, MSIA-Net, is proposed. Firstly, a feature extraction module named MSIA, which is based on asymmetric convolution, is proposed, and it can greatly reduce the number of parameters and improve the detection performance by reusing information. In addition, we propose a down-sampling module named DPP to reduce the information loss caused by pooling down-sampling. Finally, we propose a feature fusion structure named LIR-FPN that can shorten the information transmission path and effectively reduce the noise in the process of feature fusion. In order to improve the ability of the network to focus on the target, we introduce coordinate attention (CA) into the LIR-FPN; this integrates the location information of the target into the channel so as to obtain more expressive feature information. Finally, a comparative experiment with other SOTA methods was completed on the FLIR on-board infrared image dataset, which proved the powerful detection performance of MSIA-Net.

Thermal infrared detection technology can enable night vision and is robust in complex environments, making it highly advantageous for various fields. However, infrared images have low resolution and high noise, resulting in limited detailed information being available about the target object. This difficulty is further amplified when detecting small targets, which are prone to occlusion. In response to these challenges, we propose a model for infrared target detection designed to achieve efficient feature representation. Firstly, an interval sampling weighted (ISW) module is proposed, which strengthens the fusion network’s spatial relationship modeling, thereby elevating the model’s generalization capability across diverse target-density regions. Next, a detection head founded on 3D attention (TAHNet) is introduced, which helps the network more comprehensively understand the feature details of the target. This enhances the accuracy of the model in identifying the target object’s location, reduces false positives and false negatives, and optimizes the network’s performance. Furthermore, to our model, we introduce the C2f module to transfer gradient information across multiple branches. The features learned using diverse branches interact and fuse in subsequent stages, further enhancing the model’s representation ability and understanding of the target. Experimental outcomes validate the efficacy of the proposed model, showcasing state-of-the-art detection performance on FLIR and KAIST thermal infrared datasets and showing strong antiocclusion and robustness in complex scenes.