Facial recognition using a sphericity metric
AssigneeRaytheon Company
InventorsSteven J. Manson, Tara L. Trumbull
Life of this patent
Estimating the term…
Abstract
A facial recognition system and a method for performing facial recognition are provided. The facial recognition system includes a memory configured to store a target data set identifying a plurality of predefined points on a face of a target and a processor. The processor may be configured to receive an arbitrary number of photographs including a face of a subject, each of the photographs being at an arbitrary angle and at an arbitrary distance from the subject, create a subject data set identifying the plurality of predefined points on the subject's face based upon the received photographs, and perform facial recognition on the subject data set by comparing the subject data set to the target data set.
Claims25
A facial recognition system, comprising:
- a memory configured to store a target data set identifying a plurality of predefined points on a face of a target;
- and
- a processor configured to:
- receive an arbitrary number of photographs including a face of a subject, each of the photographs taken from an arbitrary angle and at an arbitrary distance from the subject,
- create a subject data set identifying the plurality of predefined points on the subject's face based upon the received photographs,
- determine a bounding object based upon at least three points within the subject data set,
- determine a circular object within the bounding object which touches each side of the bounding object,
- remap the bounding object and the circular object therein such that the bounding object, corresponding to predefined points on the subject's face, matches an object corresponding to the same points in the target data set, and
- determine a sphericity of the remapped circular object.
A method for performing facial recognition, comprising:
- receiving, by a processor, photographic data from an arbitrary number of photographs including a face of a subject, each of the photographs being at an arbitrary angle and at an arbitrary distance from the subject;
- creating, by the processor, a subject data set identifying the plurality of predefined points on the subject's face based upon the received photographic data;
- determining, by the processor, a bounding object based upon at least three points within the subject data corresponding to predefined points on the subject's face;
- determining, by the processor, a circular object within the bounding object which touches each side of the bounding object;
- remapping, by the processor, the bounding object and the circular object therein such that the bounding object, corresponding to predefined points on the subject's face, matches an object corresponding to the same points in a target data set stored in memory;
- and
- determining, by the processor, a sphericity of the remapped circular object.
An apparatus, comprising:
- a camera configured to take at least one photograph of a target, the photograph being taken at any arbitrary angle relative to the subject and at any arbitrary distance from the subject;
- a memory configured to store the at least one photograph and further configured to store a database including facial recognition data for at least one target identifying a predetermined number of points on the target's face;
- and
- a processor configured to:
- create a subject data set by analyzing the at least one photograph to determine the location of the predefined points on the subject's face,
- determine a bounding object based upon at least three points within the subject data set corresponding to predefined points on the subject's face,
- determine a circular object within the bounding object which touches each side of the bounding object,
- remap the bounding object and the circular object therein such that the bounding object, corresponding to predefined points on the subject's face, matches an object corresponding to the same points in the target data set, and
- determine a sphericity of the remapped circular object.
Dependencies are inferred from the claim text.