Analysis of Concrete Formwork Inspection and Distance Measurement based on YOLOv5

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The joint quality of prefabricated buildings directly determines structural performance. To realise the automatic positioning of concrete pouring equipment for small-size joints in prefabricated buildings, this paper develops an intelligent positioning system integrating YOLOv5 object detection and binocular vision ranging. For the screw-hole detection task in prefabricated components, a custom dataset was built using 2,000 on-site images collected under various viewing angles and illumination conditions. Five models, including YOLOv5n/s/m/l/x are compared in terms of recall, precision, mAP, and box regression loss. Considering detection accuracy, hardware cost, and inference speed collectively, YOLOv5l is selected as the object detection model. Zhang’s calibration method was adopted to calibrate the binocular camera and obtain its internal and external parameters. Using the SGBM semi-global stereo-matching algorithm, the depth and 3D positions of target screw holes were calculated based on the binocular triangulation principle. Data exchange between the host computer and PLC was achieved via the Modbus TCP protocol. The 3D coordinates of targets obtained through visual recognition are transmitted to the PLC to drive the end effector to complete positioning operations. A scaled-down indoor test platform was built for verification experiments. The effect of different shooting heights on screw-hole recognition was tested to determine the optimal working height. Experimental results showed that the average recognition accuracy of screw holes was 93.4% at a working distance of 600 mm. The measurement error of the target 3D coordinates was within the allowable engineering range, and the displacement values fed back by the encoder were generally consistent with the reference values measured using a laser rangefinder. The system achieves full-process automation of screw hole recognition, binocular ranging, positioning and operation. With moderate cost and practical applicability, the proposed scheme can provide technical support for intelligent pouring equipment for joints of prefabricated buildings.