Abstract: Infrared detection of small targets is a prerequisite for the realisation of automated infrared intelligent security, surveillance, and guidance systems. However, infrared detection of small targets suffers from high false positive and false negative rates, as well as severe information loss. To achieve real-time, high-precision infrared detection of small targets, we propose a lightweight infrared detection algorithm based on YOLO. First, targeting the feature information of infrared small targets, a dual-detection-head lightweight network structure, E-YOLO, is proposed to address the issue of target information disappearing during feature extraction; then, a feature-space-based filter pruning algorithm is proposed, which calculates the similarity between filters to prune redundant filters and improve inference speed; Finally, the cosPIoU-Varifocal loss function is proposed, which uses an adaptive penalty factor based on target size and combines angle balancing to balance sample and overlap region losses while performing positive sample weighting, thereby addressing the issue of redundant anchor box interference. Finally, validation was conducted on the SIRST and IDSAT datasets, achieving mAP values of 96.6% and 98.2%, respectively. The algorithm's computational load and network model size were compressed to 0.9 GFLOPs and 190 KB, respectively, with inference time below 3 ms. Compared to existing algorithms, the proposed algorithm achieves improvements in average detection accuracy, parameter count, and model size, meeting real-time detection requirements.