Research on quality recognition of rebar binding based on YOLOv5

As a crucial process in rebar engineering, the function of rebar binding is to ensure the stable position of rebar in the construction process. When the on-site personnel walk on the steel bar, and when the concrete is poured, the pressure and impact caused by the impact of the concrete and the collision of the vibrating rod on the steel bar are easy to make the steel bar deviate from its original position. Therefore, the role of rebar binding is very important, which provides a guarantee for the mechanical performance and durability of the structure in the future. With the continuous emergence of super high-rise buildings, their importance has become more prominent. The traditional technology and method of rebar binding detection have lagged behind obviously, and the inspectors mainly rely on manual detection and calculation with low efficiency and high cost, and they are still facing certain dangers. In order to detect the quality of reinforcement binding more conveniently, quickly and accurately, this paper conducts image recognition research on reinforcement binding based on the YOLOv5s deep learning network model, constructs a diverse and authoritative rebar binding target detection dataset through field materials from many local construction sites, uses the constructed dataset to learn and train the network model, analyzes and evaluates the training result data chart after the training, and compares and analyzes the training data of other algorithms. The model performance and data adaptation of YOLOv5 were verified. In addition, model improvement experiments are carried out according to the characteristics of the dataset and network structure in this paper. This method provides a solution for the automatic detection of rebar binding quality.