Section: New Results

Brick & Mortar Cookies

Participant : Anaïs Ducoffe.

The objective of the BMC project is to create a software that aims to present attendance and attractiveness of the customer in stores, based on automatic video analysis. This final system is designed to be used without changing current camera network of the customer store, dedicated to security purpose. Analysis should be given at different time and space resolutions. For instance, day attendance can be as interesting as year attendance. Moreover, shop owners want to be able to compare two given years or months, etc... As space resolution is concerned, the software should be able to give information about the global attractiveness of the store but should also analyze some specific zones.

IVA embedded on Bosch cameras

Intelligence Video Analysis (IVA) is embedded in some models of Bosch cameras. The algorithms are composed of human detection and tracking. They can be configured directly on the camera interface via tasks.

Following Bosch tasks were selected and studied:

• Loitering and idle object in a field tasks enable to detect stop actions in a zone, when they happened and the stop positions.

• Entering and leaving field tasks enable to know when a person enters or leaves a zone.

• Detect people in a field task enables counting people in a zone.

• Crossing lines tasks for counting people entering or leaving shop. We are able to know when the line was crossed and in what sense.

It is not possible to get people trajectory when metadata from Bosch cameras are acquired in offline mode. Then we studied live connection to get metadata directly from the camera stream using a RTSP connection. Metadata information is saved in XML format.

The previously enumerated tasks use algorithms to detect people and get their trajectories. STARS team has developed similar algorithms and has adapted their parameters values to the specific needs of this software. Moreover these algorithms can be run on any type of video cameras (live and offline modes) whereas Bosch IVA can only be run in live mode on compatible Bosch cameras. Stars algorithms can also combine several cameras at the same time in order to track people across the camera network. We need those algorithms to sell a system that doesn't need a new camera network but can be used with existing ones. They will be integrated in the final product.

Tests in real conditions

A system for testing cameras and our software was installed in partner store (Super U). Cameras were installed and configured to process all our use cases and test our mechanism to extract the metadata. We used only Bosch camera with embedded IVA. We successfully acquire 2 hours of the desired metadata. The results of embedded algorithms are reliable on realistic data : we get good results in counting people and trajectories are accurate.

Metadata storage in database

Metadata have to be stored in a hierarchical way as request of the metadata by the application should be easy and quick. We choose to store metadata in a database. This database design was constraint by data storage speed and a quick access for live computation. Different parts of database (store information, devices description etc...) were designed to be as much independent as possible.

Web interface (GUI)

The graphic interface design is in progress. The interface will be a web based one to narrow compatibility problems: the application should be used as well with a computer as a tablet.