Aiming at the low accuracy and slow detection speed of rebar binding points in the target detection algorithm of rebar binding robots, we proposed a method for steel mesh binding point detection named FNBYOLOv5, which is based on YOLOv5. Firstly, by collecting the intersections of rebar grids to create a dataset. Secondly, FasterNet lightweight backbone network is introduced into the YOLOv5 network to enhance feature extraction capability and reduce network complexity. Then, the BiFormer attention mechanism improves the accuracy of feature extraction. And considering that small-sized targets dominate the detection task, the NWD loss function is used for normalization to optimize the accuracy of detecting binding points. Finally, a redesigned FGFPN feature fusion module is devised to enhance feature interaction and improve computational performance by incorporating skip connections and cross-scale connections. Comparative experiments with different models show that the precision, recall, and mean average precision (mAP) of the model are 99.82%, 99.10%, and 98.64%, respectively, representing an increase of 2.9%, 1.53%, and 1.63% over the original model. The frames per second (FPS) reach 44.8 f/s, an increase of 4.1 f/s compared to the original model, while the size of the weight model memory is reduced by 2.93MB. Experimental results demonstrate that the improved FNB-YOLOv5 model achieves higher accuracy and real-time performance on the rebar binding point dataset, providing technical support for the development of rebar binding robots in the construction industry.