Overview

Demolition productivity and material recovery depend on how heavy equipment is coordinated, but reliable operational data from real sites remain difficult to collect. Manual tracking is labor intensive and error prone, while continuous video monitoring raises concerns about personally identifiable information, accidents, and other sensitive events. This interdisciplinary project develops an automated, privacy-preserving data acquisition system that keeps raw video within the demolition site, protects sensitive content, and preserves the operational information needed for equipment productivity analysis.

System design

  1. Thin sensing nodes: Vehicle-grade industrial cameras and long-range Wi-Fi client bridges provide reliable video capture without placing expensive computing hardware on heavy equipment.
  2. Resilient site connectivity: A multi-access-point Wi-Fi network, overlapping coverage, segmented video, and local buffering support mobile equipment in obstructed and rapidly changing demolition environments.
  3. On-site edge processing: A local gateway aggregates camera streams, adapts recording quality to network conditions and operational value, and keeps raw data inside the site perimeter.
  4. Privacy protection: Computer vision models detect sensitive regions and events, then apply targeted blurring, pixelation, masking, or inpainting before data are used beyond the protected gateway.
  5. Equipment analytics: Detection, tracking, activity recognition, and transfer learning identify demolition activities such as crushing, scooping, dumping, and loading, enabling equipment cycle-time estimation from privacy-preserved video.

Research goal

The central research goal is to test whether sensitive instances can be removed from demolition-site video while retaining enough operational information to accurately estimate equipment cycle times. The project combines privacy-by-design data handling with practical edge deployment, aiming to make large-scale field data collection acceptable to contractors without sacrificing the measurements required for productivity analysis.

Expected impact

The resulting system can provide a foundation for data-driven demolition planning that balances productivity, cost, project duration, and material recovery. It can also support future work in equipment safety, sustainable demolition, and intelligent decision-making in other field environments where access to operational video is limited by privacy concerns.