This project focused on developing a machine vision–based monitoring system for construction sites to automatically track trucks and heavy machinery, replacing manual observation and counting processes. Previously, on-site personnel were responsible for visually estimating truck loads, counting inbound and outbound vehicles, and recording working times — a process prone to human error and inconsistency.

The vision system detects and tracks trucks, excavators, and bulldozers within defined site zones using video input. Each vehicle is assigned a persistent identity, enabling accurate measurement of arrival time, departure time, active working duration, and idle periods. By analyzing object movement, positioning, and loading interactions, the system determines whether trucks are fully or partially loaded and calculates the volume of material transported from the excavation area.

Additional metadata such as license plate numbers, vehicle brand, and company affiliation is extracted or associated with tracked objects, allowing activity reports to be automatically grouped by contractor. The system generates structured tables summarizing inbound and outbound truck counts, total working time, rest time, and estimated material volumes per company.

This automated workflow provides construction managers with reliable operational insights, improves transparency between stakeholders, and eliminates the need for continuous human supervision on-site.