Aiming at the problem that the bolt target detection algorithm for steel frame is difficult to deploy due to the large amount of calculation, and the detection accuracy is not high due to the dense distribution of bolts in the construction scene, a lightweight bolt detection algorithm based on improved YOLOv8 is proposed. The ScConv module is used to fuse the C2f module in the feature extraction network, and the SRU and CUR in the module are used to reduce the space and channel redundancy of the network, so as to lighten the model; The P2 small target detection layer is introduced into the neck structure, and the BiFPN network structure is fused to increase the two-way connection path, which promotes the feature propagation up and down, and improves the accuracy of the network for bolt detection. The experimental results show that the proposed algorithm performs well in the self-collected data set, and the mAP accuracy is improved by 9.9% compared with the original network, while the number of model parameters and the model size are reduced by 0.973í106 and 1.7 MB respectively.