The face search module
Face search module from, Axxon Next features a face capture and search algorithm, that automatically detects a face in the field of view or in a provided video footage of one or several cameras, and then searches for that same face in other video provided by networked cameras.
This face search module makes it possible to accumulate a database of faces which have come into the field of view of cameras and then search this database for similar faces. The web report system provides an interface for this module and makes it possible to aggregate data from all video servers on which the face search module is installed.
The face search module makes it possible to find similar faces in the archive quickly and collect statistics on their appearance in the scene for various cameras.
How the face search module works
For each face which comes into the field of view of a video camera, vector biometrics are compiled and recorded in the archive in parallel with the captured face image. Thus a database of all the faces which have appeared in the cameras’ field of view is created.
In order to conduct a search for similar faces in this database, you must either indicate a frame in the video archive containing a face, indicate a link (URL) to an image of the face, or upload a graphic file with the face image to the system. The face search module conducts a search for all similar faces in the accumulated database and classifies them by degree of likeness. The results are displayed as a list of photographs sorted by degree of likeness. You can set a minimum similarity threshold for displaying results; photographs with a degree of likeness lower than this threshold will not be displayed.
In the case of a large archive, the first results are displayed immediately, and you can work with them while the search is still in progress. For each face found you can view the relevant video footage. The printed form of the report will contain the photograph of the face presented for the search and photographs of all similar faces.
The results of each query are saved in the system for later viewing without repeating the search. After displaying the results you can select one of the faces found and conduct a search based on it, and you can also change the minimum similarity threshold for quicker operation.
The Cognitec 5.0, Cognitec 8.0 and VeriLook libraries are used for creating vector biometrics.
The face search server provides a web service which can be accessed using HTTP protocol; this enables simple integration with any software systems regardless of platform.