Enhanced YOLOv8 Model for Drug Detection in Compound Robot System

The sudden emergence of the novel coronavirus (COVID-19) has had a catastrophic impact on global populations, with healthcare workers (HCWs) facing unprecedented challenges due to the high infection rate among this group. The deployment of composite robots, which combine the functions of fast movement and flexible grasping, can potentially address the shortage of HCWs and facilitate the delivery of essential medical supplies and other tasks in isolation environments. This paper presents a key technology for composite robots to perform the function of picking up and delivering medicines. The identification and detection of medicines is a crucial aspect of this technology. To enhance the accuracy and real-time performance of composite robots in detecting medicines, a target detection algorithm based on the improvement of YOLOv8 for medicines boxes is proposed. Firstly, a global attention mechanism (GAM) is incorporated into the backbone network to enhance the network's focus on crucial information through cross-latitude information interaction and to retain channel and spatial information to a greater extent, thereby improving the ability of the backbone network to extract features. Secondly, the weighted bidirectional feature pyramid network(BiFPN) is integrated into the neck network, which facilitates multi-scale feature fusion by assigning appropriate weights to each input feature and refining the feature fusion process. The experimental results demonstrate that the proposed model achieves an average detection accuracy (mAP) of 99.20%, a recall of 98.90%, and an F1 score of 99.05%, while maintaining a detection speed of 158.73 FPS. In comparison to YOLOv6, YOLOv8n, and YOLOv8s, the proposed model exhibits the highest accuracy index and satisfactory overall performance, aligning with the requirements of composite robots for rapid and precise identification of medicines. The model enables the composite robot to accurately detect and identify medicines in real time within an isolated ward environment.