Participants: Francois Bremond, Alberto Avanzi, Frederic Cupillard, Magali Maziere, Monique Thonnat, Christophe Tornieri and Marcos Zuniga.
In December 2002, the Orion VSIP platform adapted for the CASSIOPEE project has been installed on a PC located in a bank agency. The platform processes color images coming from a surveillance video camera looking at the agency, at a framerate of 10 images per second. The goal of the platform is to detect a predefined scenario called bank attack. This scenario involves 2 actors : an employee and a robber. The robber enters the bank agency, goes towards the employee, standing behind the counter and threatens the employee to open the safe door.
A performance evaluation has been done by playing several times with actors the same bank attack scenario. Other ``normal'' scenarios (scenarios which are not supposed to arise an alarm) have been played too, always by actors, to evaluate the robustness of the platform against false alarms. The results showed 80 % of true positive and 0 % of false positive.
Figure 1. People Counting based on 3D parallelepiped Classification into 3 people classes : 1Person, 2Persons, 3Persons, Unknown.
In 2005, the VSIP system was installed and evaluated in four bank agencies near Paris with different configurations using both recorded videos and live acquisition. The system has been designed to recognize bank attack scenarios. However it has been mainly devoted to count people in the secured technical room of the bank (ETS) since only a maximal number of persons are allowed to be together in this room. Concerning the counting people evaluation, the recorded videos corresponds to twenty days and 4 days were associated with Ground Truth. The situation with one person in the ETS was correctly detected 75 times out of 75 with no false detection. The situation with more than two persons in the ETS was correctly detected 6 times out of 11 with no false detection. The evaluation performed on live videos gave similar results. The main errors were due to two problems: the room was dark (the light was not turned on) and the position of the camera could not permit to observe the people evolving in the room. No false alarm has been observed during the whole evaluation process which was a main requirement from both end-users, the bank and the video security operators. The system has also been evaluated in live conditions for a few weeks. This evaluation has been performed directly by the company in charge of the remote surveillance for a few weeks and by ourselves for several days. Thus we have been able to verify the effectiveness of our approach: the alarms are sent in real time to the surveillance headquarter. The ORION startup Keeneo is currently in charge of commercialization of the system.