Automated vision-based location monitoring for jack support

Accurate installation of jack supports is critical for structural stability in construction. Traditional inspection methods, such as visual inspection and tape measurements, are time-consuming, labor-intensive, and tend to involve human error, especially in large-scale projects with limited inspection resources. To address these issues, this study proposes an automated vision-based location monitoring model for jack supports, utilizing a markerless and geometry-based approach to achieve rapid and efficient inspection from a single image. The proposed system integrates two modules: a 2D localization module and a 3D localization module. The 2D localization module segments jack supports in the image and extracts their corresponding bottom reference points. The 3D localization module transforms the reference points into world coordinates using vanishing point-based calibration and a triangular similarity-based transformation. The proposed model showed a mean localization error of 0.15 m for supports located within 15 m of the camera. The proposed model required an average of 4 s per an image, demonstrating significantly faster performance than traditional methods and validating its potential for inspections with reduced manpower.