Abstract: To address visual inspection issues in prefabricated laminated panels, such as missing embedded parts, non-parallel internal reinforcement frames, and non-compliant outward bending hooks, a target detection algorithm based on an improved YOLOv5s model is proposed. To mitigate interference during target detection, the CBAM(Convolutional Block Attention Module) Attention Mechanism and Transformer module are integrated into the backbone network layer of the YOLOv5s model. Additionally, a prediction module is added to the output network to enhance the model's ability to extract image features. The CBAM Attention Module optimizes the feature channels, allowing the network to focus more on the feature representation of the target object. The Transformer Module captures global long-range dependencies in the image, enabling the model to better focus on the target layer, thereby further optimizing the model's output results. The results indicate that the improved YOLOv5s model increases the Mean Average Precision (mAP) by an average of 12.24% compared to the original YOLOv5s model across three types of defect detection experiments. The enhanced model demonstrates superior feature extraction and fusion capabilities in the quality inspection of prefabricated laminated panels.