Abstract: Detecting small objects by their infrared signatures is of considerable importance in military reconnaissance, aerospace monitoring, and security, especially in low-light and complex weather conditions. Leveraging infrared thermal imaging technology enables relatively precise detection and identification of distant, faint targets. An enhanced CTCA-YOLOv8 algorithm is proposed to address the challenges of weak characteristics and low detection accuracy for small targets in infrared scenes. First, a layer is added to detect small targets while removing detection layers tailored for large targets to enhance detection of small targets while significantly reducing the parameter count of the model. Subsequently, BoT module, DFF and CTCA are incorporated to acquire global feature information and fuse features across different hierarchies. CTCA-YOLOv8 exhibited superior performance in terms of detecting small infrared targets. Experimental results on the FLIRv2 dataset demonstrate a 5.3% improvement in mAP50 metric compared to YOLOv8n model, with an increase in computational complexity of 3.5GFLOPs and a 37% reduction in parameters compared to YOLOv8n. CTCA-YOLOv8 achieved enhanced object detection accuracy while maintaining a lightweight design.