To address the issue of the lack of targeted detection of infrared dim small targets by existing deep learning network structures, a method for recognizing infrared dim small targets based on an improved YOLOv8, named UT-YOLOv8 (YOLOv8 enhanced with UniRepLK Block and Triplet Attention) is proposed. In this method, a triplet attention mechanism is introduced in the detection head at the output end of the feature fusion network. Additionally, new small target detection layers and detection heads are added within the feature fusion network, and large kernel convolutions are incorporated within the spatial pyramid pooling layer of the feature extraction network. These enhancements are tailored to the imaging characteristics of infrared dim small targets. Validated on real infrared image data, the experimental results indicate that the UT-YOLOv8 algorithm, while maintaining high detection speed, has effectively enhanced the network's precision in recognizing infrared weak small targets, achieving a mean Average Precision mAP@0.5 of 95.9%.