Open construction sites are highly complex environments for onsite tracking. The large amount of items present along with the amounts of occlusions/obstructions, make efficient onsite personnel tracking very difficult. Current tracking methods rely mostly on Radio Frequency technologies, such as Radio Frequency Identification (RFID), Global Positioning Systems (GPS), Bluetooth and Wireless Fidelity (Wi-Fi, Ultra-Wideband, etc). These technologies require manual deployment of tags and record keeping of the people they are placed on. In open construction sites with numerous people working simultaneously, sensor installations and maintenance increases the cost and time needed to implement these tracking methods. This paper presents a new, less obtrusive method for open site tracking of personnel using video cameras. Video feeds are collected from on site video cameras and presented to the user. The user can then select the person that is to be tracked. The person is subsequently tracked in each video using 2D vision tracking. In each frame, epipolar geometry is used to calculate the depth (3D position) of the person. This method addresses the limitations of radio frequency methods since it uses existing construction site equipment (security cameras) to perform tracking. The method has been implemented in a C++ prototype and preliminary results show effective 3D positioning of personnel in construction sites.
3D location, personnel tracking, image alignment