SDK advances face recognition for large surveillance systems

Neurotechnology announced VeriLook Surveillance 2.0, a SDK for biometric face identification using live video streams from single or multiple high-resolution digital surveillance cameras.

It provides real-time identification of faces and can be used in a wide range of surveillance systems for retail and commercial areas, entrance monitoring and counting, automated time-attendance systems, law enforcement applications and transportation security.

The new, integrated face tracking algorithm in VeriLook Surveillance 2.0 includes a robust, dynamic face model which can adapt to visual appearance changes as subjects move across the scene. It continues tracking of subjects even when their faces briefly disappear from the frame or when they are partially blocked by other objects or even other faces (a common problem while tracking multiple faces)

Because it can now simultaneously process video streams from multiple surveillance cameras, VeriLook Surveillance 2.0 is suitable for use in large surveillance systems.

VeriLook Surveillance 2.0 incorporates the VeriLook 5.1 face recognition algorithm, which enables detection of faces with up to 45 degrees out-of-plane rotation in yaw angle. The new face tracking algorithm uses motion prediction models to re-localize faces that have undergone full occlusion, such as when a subject has been fully obstructed by a wall and emerged on the other side.

The dynamic face model allows the system to track faces in front of backgrounds and ensures that subjects can be localized in all video frames. Face images can then be matched against internal databases, such as criminal watch lists or authorized personnel, and VeriLook Surveillance will report recognized faces to the system.

VeriLook Surveillance 2.0 enables one computer to process images from multiple cameras in the same VeriLook Surveillance process. In large, extended surveillance systems data may be coming from multiple cameras with processing occurring across an array of computers.

The new SDK provides connections between VeriLook Surveillance units deployed on different computers, synchronizing the databases so that the system works as a whole within the logical network. When a new subject is enrolled in one of the surveillance instances, the data is sent to other surveillance processes, enrolling the subject in all running surveillances. The synchronization routines are provided as sample level code, leaving the customer free to modify the communication or logic behind any or all synchronization processes.




Share this