Laser-triggered and vision-validated collaborative method for node recognition of rebar tying robots
Year: 2026
Type: article
Abstract: The accuracy of node perception in rebar tying robots determines the safety of the tying. The current laser-triggering-only solution cannot judge the state of the node, thus resulting in incorrect tying. Based on the multi-channel binary laser signal; this paper puts forward a collaborative recognition method that can be used for image collection and YOLOv5 classification: When the laser array is activated, a local image is obtained, and You Only Look Once version 5(YOLOv5) is then used to determine the category of the node; only nodes that are tieable will be tied. Based on the 60-image held-out test set, the model has an overall mean Average Precision at IoU=0.5 ([email protected]) of 92.0%; and the tieable node (normal_node) is 96.8%. A comparison of 91 physical nodes has shown that the false-binding rate of the above-mentioned method is only 8.2%; it is better than the laser-only scheme (45.1%) and close to that of the baseline using vision only. In addition, offline inference tests on the Intel Core i7 processor showed that the laser-triggered mode requires that only about 49.5% of the frames be processed; and thus the work required for YOLO inference has been reduced to about 50.5%. Without changing the existing hardware, the proposed method has increased the reliability of the decision to tie and provided a very cost-effective perception solution for engineering deployment.
Source: Smart Construction
DOI: 10.55092/sc20260021
Authors: Jingwen Liang(Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing), Yingdi He(Validation), Cheng Zhou(Writing – review & editing), Yan Zhou(Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Writing – review & editing), Shifeng Wen(Investigation), Xi Zeng(Validation)
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7234160600
