It is the contractor's usual practice to collect information from construction sites to measure and verify the progress. This is carried out through manual observations. However the level of accuracy is highly dependent on human errors and limitations in data. As an alternative to the current system, this study introduces a conceptual frame work of a fully automated real time tracking system using image processing techniques for data extraction from video files of construction activities. The worker tracking system provides an opportunity to track the construction worker location and their movements in a specified work area. The research study proposes the "construction hardhat" as an object which represents the presence of a site person (worker, supervisor, engineer, etc.) at any given time. Hence, it is used as the key tracking object for worker recognition in the study. Worker tracking algorithm is purely based on characteristics of the hardhat such as unique shape and colour. Feature based template matching is used as the pattern recognition technique. Proposed system uses a combination of four cooperating detection and tracking modules to track workers in camera images. Modules used are motion detector, colour recognizer, shape detector and 3D localizer. A stereo camera based system is used to track 3D coordinates of the target object.
展开▼
