YOLOv8: Detection of Concealed Objects in Terahertz Images

Terahertz imaging technology has attracted considerable attention as an emerging inspection technique in the field of security checks. This paper proposed a terahertz security object detection based on deep learning. In this paper, we focus on the BIFPN structure that exhibits characteristics of different resolutions on different scales, allowing objects of different sizes to be represented with appropriate features at their corresponding scales. By fusion multi-scale features at different scales, the model's effectiveness is improved. Additionally, we incorporate a 4 × 4 tiny object detection head to detect tiny objects in terahertz images. Through the collaborative efforts of the detection head, accurate detection of hidden items of various sizes in terahertz images is achieved. Results from the experiments demonstrate an enhancement in detection accuracy with the proposed method for terahertz security images under the YOLOv8n model.