Scan face of video feed

1. A system, comprising:

a processor; and

a memory to store instructions that, when executed by the processor, cause the processor to:

determine a plurality of velocities corresponding to movements of a plurality of human objects identified in a video feed, wherein the plurality of velocities comprise a given velocity corresponding to the movement of a given human object of the plurality of human objects;

determine a probability distribution based on the plurality of velocities;

trigger an abnormal event in response to the given velocity corresponding to a threshold number of standard deviations from a mean of the probability distribution; and

control a camera in response to the triggering of the abnormal event to scan a face of the given human object.

2. The system of claim 1 , wherein the instructions, when executed by the processor, further cause the processor to:

track a position of the given human object over a time interval to determine a trajectory; and

track the given human object to determine trajectory coordinates.

3. The system of claim 1 , wherein the instructions, when executed by the processor, further cause the processor to:

aggregate the plurality of velocities to determine a pattern for a plurality of pixel positions across a video frame of the video feed during a normal state; and

trigger the abnormal event in response to a crowd behavior observed in the video feed being significantly different from the pattern at more than a pattern threshold number of the plurality of pixel positions.

4. The system of claim 1 , wherein the instructions, when executed by the processor, further cause the processor to:

detect the given human object by analyzing a plurality of pixels of a video frame using at least one of a Histogram of Oriented Gradient (HOG) and a cascading classifier.

5. The system of claim 1 , wherein the instructions, when executed by the processor, further cause the processor to:

direct the camera to scan a plurality of faces of the plurality of human objects in response to the abnormal event being triggered, wherein the camera uses at least one of pan, tilt and zoom to scan the plurality of faces;

direct the camera to scan the plurality of faces sequentially; and

scan a face of the plurality of faces corresponding to the given human object at a higher resolution in response to the camera zooming.

6. The system of claim 5 , wherein the instructions, when executed by the processor, further cause the processor to:

direct the camera to assign higher priority to scanning the faces of the plurality of faces closer to a boundary of a video frame;

direct the camera to assign higher priority to scanning the faces of the plurality of faces less than a threshold number of degrees between a line from the camera to a face of the given human object and a horizon;

direct the camera to assign higher priority to scanning the faces of the plurality of faces further away from the camera; or

direct the camera to assign a higher priority to scanning a frontal-view of at least the given human object.

7. The system of claim 1 , wherein the instructions, when executed by the processor, further cause the processor to:

create an index of scanned faces, the index to include at least one of associated time and location data; and

use the index to match the scanned faces with faces of a database via a face recognition system.

8. A method, comprising:

determining a plurality of velocities corresponding to movements of a plurality of human objects identified in a video feed, wherein the plurality of velocities comprise a given velocity corresponding to the movement of a given human object of the plurality of human objects;

determining a probability distribution based on the plurality of velocities;

triggering an abnormal event in response to the given velocity corresponding to a threshold number of standard deviations from a mean of the probability distribution; and

controlling a camera in response to the triggering of the abnormal event to scan a face of the given human object.

9. The method of claim 8 , wherein the triggering further includes:

focusing at a center of the given human object;

zooming to at least one of a height of the given human object and a maximum zoom of the camera;

tracking the given human object while a face of the given human object is scanned; and

repeating the focusing, zooming and tracking for a second human object of the plurality of human objects, after the face of the given human object is scanned.

10. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor of a device, cause the processor to:

determine a plurality of velocities corresponding to movements of a plurality of human objects identified in a video feed, wherein the plurality of velocities comprise a given velocity corresponding to the movement of a given human object of the plurality of human objects;

determine a probability distribution based on the plurality of velocities;

trigger an abnormal event in response to the given velocity corresponding to a threshold number of standard deviations from a mean of the probability distribution; and

control a camera in response to the triggering of the abnormal event to scan a face of the given human object.

11. The non-transitory computer-readable storage medium of claim 10 , wherein:

a plurality of faces of the plurality of human objects are scanned via a plurality of pan, tilt and zoom (PTZ) cameras in real time, and

the instructions, when executed by the processor, cause the processor to determine a priority for scanning at least one of the human objects for one PTZ camera of the plurality of PTZ cameras based on whether a face of the at least one human object of the plurality of human objects has been scanned by another PTZ camera of the plurality of PTZ cameras.