Fiatech

Presented by: Man-Woo Park, Graduate Research Assistant, Georgia Institute of Technology

Reserve for this live webinar and earn 1.0 PDH

In this presentation, a novel method of tracking construction entities will be introduced. 3D object tracking on construction sites has a wide variety of applications. It allows identification and tracking of personnel, equipment, and materials to support effective progress monitoring, activity sequence analysis, productivity measurements, and safety management. In addition, tracking instantly enables the identification of critical activities and problems, which allows for on-site project control and decision making capabilities. Available tracking solutions are mainly based on Radio Frequency technologies including GPS and RFID, etc. They all work under the same principle of having a sensor attached on each entity to be tracked. These technologies have been applied, and proved to work excellent for most scenarios involved in construction management such as proactive work zone safety and material registration and installation. However, when it comes to large-scale and congested construction sites, the installation of the sensor system can be costly and time-consuming due to the large amount of items involved. Also, privacy issues can arise out of tagging workers. For these specific scenarios, vision-based tracking may have potential for use as an efficient alternative. This session presents a framework that promises to determine the spatial location of project related entities in a large-scale congested construction site, such as construction equipment, personnel, and materials without installation of sensors. Under this framework, video streams are initially collected from a set of fixed video cameras that are placed at a project site. The views of the cameras should have large overlapping areas so that the entities appear at least in two views. In each view, project related entities are automatically identified and localized when they first appear. The identified entities are tracked in the subsequent frames. This process provides 2D pixel coordinates across frames for each view. In order to obtain 3D coordinates of the entities with two 2D pixel coordinates at each frame, the geometric relation between the camera views is discovered. Once the distance between camera positions and the angle between the camera views are revealed, the 3D coordinates are calculated by triangulating two cameras and the entity. This technology is tested on videos collected at real construction sites. The accuracy is measured by comparing with total station data, which demonstrates its competitive and promising performance for on-site tracking purposes.