We propose two applications for two different situations : individual behavior detection (presented here) and crowd detection (there).

In a public area, when somebody has a suspicious behaviour, several indices can be used to raise an alarm depending on the situation. Here, the person seems to wander slowly among cars. EVITECH's Video analytics  software can automatically identify and raise these situations to the interest of guards  :

• frequent stops,
• wrong direction walk,
• low or high speed,
• crawling or squatting silhouette,
• person staying or wandering to long in the area (using signature of the object),
• too long stop close to a sensitive area,
• same person seen successively under different cameras,
• groups, group size,
• abandoned object, tagging, laying object,
• conditions combinations, other observations ...

Possibilities are without limit, especially with the target signature function that enables target tracking even in case of other persons crossing, and from a camera to another if the signature is shared on the camera network. Our integration teams are at your disposal for any case of situation where you want to study an opportunity.

The smart video sensor connects any kind of image data (BNC, IP, WIFI, DSL, Firewire, ...), uncompresses all formats (PAL/SECAM, H.264, MPEG4, MJPEG, ...), and offers an intelligent analysis of images, resistant to any kind of outdoor conditions, for sensistive sites protection applications : intrusion detection, abnormal behaviour, abandoned object, ... It sends alarm messages through IP messages, dry contactors, e-mails, SMS or MMS via GSM/GPRS/UMTS ...

Here is the example of abandoning an object :

The EAGLE & JAGUAR software allow to detect abandoned, even small sized, objects, even very similar to other objects close from the laying location. On the images below (SONY tri-CCD camera), you can see a test under snow, a person laying a small white bag on the wall, just similar to a snow falling flake. In the first image, we see a person coming (P-1, for Person and first threat), with the bag. Biggests falling snowflakes (see on the upper left corner) are filtered by the software, they are not classified as a threat : they cannot raise an alarm.

In the second image, the person leaves while having laid the white bag on the wall. It is then detected as threat number two (I-2), where I stands for unclassified Ignored-type threat.

Falling snow is still unidentified as threat. At last, in the following image, the person has left the scene, and the bag is still there, as the last threat :

Conclusion : The software tells the operator that only two threats have been observed in the scene, despite the falling snowflakes : the person (P-1) and the bag (I-2). An alarm triggered on an immobile object will finally raise the expected alarm.