OpenPTrack: People Tracking for Heterogeneous Networks of Color-Depth Cameras

Abstract

Abstract. This paper introduces OpenPTrack, an open source software for multi-camera people tracking in RGB-D camera networks. OpenPTrack provides real-time people detection and tracking algorithms from 3D data coming from Microsoft Kinect and Mesa SwissRanger. The software is able to track people at 30 Hz with minimum latency. A user-friendly cal-ibration procedure is also provided, so that the camera network can be calibrated in few seconds by moving a checkerboard in front of the cam-eras and seeing the calibration results in real time. The algorithms for people detection are executed in a distributed fashion for every sensor, while tracking is done by a single node which takes into account detec-tions from all over the network. Algorithms based on RGB or depth are automatically enabled while the system is running, depending on the luminance properties of the image. OpenPTrack is based on the Robot Operating System and the Point Cloud Library and has been tested on networks composed of up to six sensors.