Human Pattern Recognizer


HPR is a method for recognizing and tracking human walking in planar range data.

The core idea of the method is to stack consecutive 2D scans into a 3D space-temporal representation, where X,Y is the planar data and Z is the time dimension. This 3D representation combines information about the shape and size of objects with information about their motion in time. This provides richer information wherein clustering and pattern recognition can be applied to recognize human walking patterns and to track the different individuals in the scene.


The prototype is developed on Github,
The current stable release is v2.0.0.

History and Contributors

Version 2.0.0, 1 Mar 2017:
Katerina Zamani improved the method and updated the prototype accordingly.

Version 1.1.0, 8 Apr 2016:
Andreas Lydakis and Georgios Stavrinos largely re-implemented HPR and dramatically improved integration as a ROS node.

Version 1.0, 13 Mar 13 2015:
HPR was originally developed in Matlab by Theodoros Varvadoukas and Ioannis Giotis. This version was ported to Python by Athanasia Sapountzi. The method that this prototype implements is described in the paper "Detecting Human Patterns in Laser Range Data" [ECAI 2012].