EVITECH has installed crowd and mass counting applications in various real contexts in cooperation with organisms (authorities, and transport company) in a process where a manual counting was also processed in order to counter-evaluate the software results.

These installations were done on after-demonstrations recordings over public streets, and real time images on passage areas in transport sites (crossings, escalators, one way, ...). In both cases a manual counting was done over several hours in order to validate the measures outing from the Lynx software.

With a precision of 80% to 98% depending on the situations (image quality, angle of view, density), these installations allowed to validate the accuracy of EVITECH's algorithms for its LYNX product, and to better identify and precise installation conditions with respect to measure precision.

The relationship between the observation angle of the camera with the horizon and the accuracy of the measure, already demonstrated theoretically in our paper Peaceful_Crowd_Monitoring_V6b.pdf presented at the major French congress of global Security  WISG'2013 in January 2013 (available here), could be verified by the experience along the testing of these installations. An installation guide was written in order to provide advices to integrators and end-users in order to select installation conditions of the cameras for obtaining the best results.

One of the installations was selected for long term tests, which will allow progressively to reinforce the precision of LYNX software functions on the long term, and over real scenes.