Monitoring Progress and Safety Risks with Computer Vision

The Problem

Managing safety and quality across large construction sites presents a significant operational challenge. Project managers must monitor worker safety practices, track construction progress, and identify potential issues before they cause delays or hazards. Traditionally, these responsibilities rely heavily on manual site inspections and periodic reporting from field teams.

A commercial construction company faced these challenges across several active projects. Multiple crews were working simultaneously across large sites, often in areas that were difficult to monitor continuously. While site supervisors and safety officers conducted regular inspections, it was difficult to maintain consistent visibility across the entire construction environment.

The organization installed fixed cameras around construction sites and occasionally used drones to capture aerial imagery. These tools generated valuable visual data, but the images were largely reviewed manually (or often ignored). With thousands of images captured each week, identifying safety risks or progress issues required significant time and effort.

In addition, field engineers frequently captured photos using mobile phones during routine inspections. These images documented site conditions and potential problems but were often stored in separate systems or shared informally through messaging platforms. As a result, the organization lacked a unified way to analyze visual information across its projects.

The company needed a system that could continuously monitor construction activity, identify potential safety risks, and help project managers understand progress across multiple sites without relying entirely on manual inspection.

QSC’s Modeling Approach

QSC developed a computer vision system designed to analyze images captured throughout the construction workflow.

The system integrated imagery from several sources, including fixed site cameras, drone surveys, and photographs collected by field engineers using mobile devices. These images were routed through a centralized processing pipeline where they could be analyzed consistently and stored in a unified data platform.

Using these image datasets, QSC developed computer vision models capable of detecting key objects and activities relevant to construction operations. Deep learning techniques were used to train models that could identify workers, equipment, safety gear, and construction materials within images.

Object detection models allowed the system to recognize conditions such as workers entering restricted areas, missing safety equipment, or unusual equipment behavior. At the same time, anomaly detection techniques were used to identify unexpected visual patterns that might indicate emerging issues on the construction site.

By analyzing images from fixed cameras and drone flights, the system could also monitor visible changes in construction areas over time. This helped project managers track whether construction activities were progressing as expected and identify areas that required attention.

In addition to the modeling components, QSC developed software tools to make the system usable for project teams.

A desktop application provided project managers with a centralized dashboard where they could monitor site activity, review flagged images, and explore historical visual data from multiple projects. This allowed teams to investigate potential issues and compare conditions across different areas of a site.

A mobile application supported engineers and safety officers working directly in the field. Through the mobile interface, users could capture new images of site conditions and upload them directly into the monitoring system. These images became part of the training dataset and helped improve the computer vision models as new scenarios were encountered.

Together, the modeling framework and software tools created a unified visual monitoring system that combined automated analysis with field observations from site personnel.

Decision Support in Practice

Once deployed, the system provided construction teams with a continuous view of site activity and safety conditions.

Images captured by fixed cameras and drones were analyzed automatically as they entered the system. When potential safety risks or unusual conditions were detected, the system flagged the relevant images, highlighted them in the monitoring dashboard, and sent alerts to management.

The system also helped teams track construction progress by identifying changes in materials, equipment placement, and structural development across different areas of the site. By reviewing visual data collected over time, managers could verify whether work was progressing according to schedule and improve timeline planning for future projects.

Field engineers and safety officers used the mobile application to capture additional images during inspections. These images provided context for flagged events and helped expand the dataset used to train the computer vision models.

By combining automated image analysis with on-site observations, the system allowed construction teams to monitor safety conditions and project progress more consistently across large and complex worksites.

Outcome

The most significant improvement was the organization’s ability to maintain consistent visibility across its construction sites.

Instead of relying entirely on periodic inspections, the company gained a system that continuously analyzed visual data from cameras, drones, and field observations. Potential safety risks could be identified earlier, and project managers could investigate issues before they escalated into larger problems.

The integrated software platform also improved coordination between project managers, safety officers, and field engineers. Images captured across the site were available in a single system, making it easier to review conditions, investigate incidents, and track changes over time.

Equally important, the mobile application allowed field teams to contribute directly to the monitoring system by capturing images during inspections. These observations strengthened the system’s ability to recognize new conditions and provided valuable context for construction managers.

For the organization, the result was not simply a new inspection tool. It was a unified visual monitoring system that helped teams maintain safer construction sites while improving visibility into project progress.