Automated Pile Monitoring System with SAM2 Segmentation

This study presents an innovative automated system for real-time monitoring of pile verticality and height in construction sites, integrating advanced visual measurement techniques with the SAM2 segmentation model. The proposed framework leverages binocular stereo vision, machine learning-based segmentation, and geometric reconstruction to achieve high-precision, non-contact measurements. Key innovations include promptable pile segmentation, dynamic PTZ camera control, and robust verticality synthesis through plane intersection analysis. It can enhance safety, efficiency, and accuracy in pile monitoring, and it can provide a scalable and cost-effective alternative to traditional manual and high-end technological methods.