Incheon National University Researchers Develop Multi-Camera AI System for Excavator Tracking

Researchers from Incheon National University have developed an automated multi-camera system to improve excavator tracking accuracy in dynamic construction sites, announced in a press release. The system integrates deep learning-based instance segmentation with a reliability evaluation framework to address frequent occlusions that occur when multiple machines operate simultaneously.

The research, published in Automation in Construction, introduces a quantitative method for assessing camera reliability at the frame level. The technique automatically selects the most reliable camera view during occlusion events, reducing errors in equipment tracking. The study also defines critical occlusion thresholds—0.7 for the excavator arm and 0.5 for the body—beyond which tracking accuracy significantly declines.

By continuously evaluating reliability and using existing camera infrastructure, the approach improves tracking continuity without requiring new installations. This method enhances the accuracy of equipment operation logs used for carbon emission estimation and regulatory reporting while lowering operational and administrative costs.