Image analysis – Applications – Personnel identification
Reexamination Certificate
2002-02-06
2004-10-05
Mehta, Bhavesh M. (Department: 2625)
Image analysis
Applications
Personnel identification
C382S190000, C382S285000, C345S420000, C345S679000
Reexamination Certificate
active
06801641
ABSTRACT:
FIELD OF THE INVENTION
This invention relates to the field of face recognition, and, in particular, to face-identification for law-enforcement and corrections applications.
BACKGROUND OF THE INVENTION
Law enforcement organizations rely heavily on witnesses to a crime for purposes of identifying criminals based on pictures of the criminal. Typically, a witness will review pages and pages of “mug shots” of criminals from books of mug shots maintained by the law enforcement organization. The witness hopefully recognizes the criminal from the mug shots.
Over the years, huge collections of mug shots have been developed by local, state and federal law enforcement agencies. Some of these collections of mug shots have been deposited in a national depository for mug shots maintained by the Federal Bureau of Investigation (FBI) of the U.S. Department of Justice. It is estimated that in the United States there are approximately 60 million mug shots in various mug shot collections across the country. However, the FBI has only about one-third of all U.S. mug shots in its own collection.
Mug shots are typically a frontal and profile views of an individual. A mug shot usually comprises two photographs of the front and side view of the face of an individual. To provide standardization of mug shots, law enforcement photographers follow specific guidelines on the positioning of the front view and side views of a person for photographing mug shots. Accordingly, each pair of mug shots, i.e., the front face and side face views, is uniform with respect to angle of view, distance and lighting of the person being photographed. These mug shots have been stored in photographic books, on film, on tape and now digitally on computer discs.
Mug shots show only two views of an individual—the front face view and a profile face view. To recognize an individual suspect from mug shots, a witness must recognize the front face or profile view of the individual. However, the witness may not have seen the suspect from the front or profile view. Rather, the witness may have seen an angled view of the suspect's face, such as from above, below or to one side. The view that the witness sees at a crime scene is often different than the front or profile view shown in mug shots. For a witness to recognize a suspect by reviewing the mug shots, the witness must convert in his own mind the angle from which he saw the suspect to either or both a front face or side view of the individual as shown in the mug shots. Not all witnesses can reliably convert in their mind the view actually seen of the suspect to the front face or side views shown in mug shots. Because all witnesses cannot make this conversion, witnesses from time to time have not been able to identify suspects. Moreover, witnesses at criminal trials are subject to cross-examination about how they recognized a suspect based on the frontal or profile view shown in a mug shot when the witness never saw the suspect from those particular views. Accordingly, criminals have not been identified or have been acquitted at trial because the witness did not reliably recognize the criminal suspect when reviewing mug shots, or because the witness was effectively cross-examined at trial with respect to the view that he had of the suspect.
There are many other facial features that are more difficult to describe and identify, such as the size of a person's nose, description of their face, etc. These other facial features are often recalled and described by witnesses to a crime, and can be used to recognize a suspect. However, these other facial features are not easily categorized and, thus, have prevented the creation of an effective system for categorizing or indexing mug shots. In the past, law enforcement investigators have attempted to characterize facial features. For example, law enforcement organizations have created standard measurement tools for identifying the size of person's eyes, nose, mouth, facial structure and ears. The criteria that are used are primarily used for identifying mug shots are limited 2-D criteria based on a full frontal view or profile view of a face. For example, the criteria as to nose width and eye shape may be limited to just front view figures, and not to profile views. Similarly, ear shape criteria may be applicable only to profile views and not to frontal view pictures of the individual. Thus, these criteria are of limited use at best if the witness observed an individual from a perspective other than a frontal or profile perspective.
Prior to the present invention, no known techniques were used for indexing facial features based on three-dimensional (3-D) framework. Prior approaches to cataloging and indexing facial features have assumed that the faces are shown only in 2-D full frontal and/or profiled views such as in mug shots. These indices of profile and frontal views of mug shots are helpful in identifying suspects when a front face or profile face view is available of the suspect. For example, if the suspect looks straight into a security camera such that there is a front face image, then that front face image can be indexed using standard face indices. Using the indices of the security camera picture of the suspect, the indicie values can be used to identify mug shot pictures having the same indicie values in searching for the criminal. Similarly, if a witness working with a law enforcement sketch artist can generate a frontal face and/or profile view of a suspect, that frontal face and profile sketch can be used to categorize the facial values of the criminal. These facial values can also be used to access the mug shots which are by the facial values indices used to obtain those mug shots having the same facial value indices.
The indices of facial features used to index mug shots are not useful when a frontal face or profile picture or sketch of the perpetrator is unavailable. Side views of a suspect cannot be readily categorized by facial indexing used with the current mug shots. A side view picture of a criminal is not susceptible to identifying the facial feature values used with current indices of mug shots. Unfortunately, the majority of photographs taken of suspects by security cameras are of angled views of the suspect's face other than frontal face or profile. Thus, when a security camera captures on film a suspect, often that picture cannot be directly used to identify the suspect by mug shots.
It has been long recognized to be an undesirable trait that the mug shot indices do not allow for correlation with pictures of criminals other than straight front or straight profile views. It is a long-felt need in law enforcement to have a computerized system for recognizing faces in which a photograph or other 2-D image of a criminal from any view can be quickly and reliably analyzed and compared to a mug shot collection to select those mug shots of possible perpetrators of a crime.
SUMMARY OF THE INVENTION
The present invention is a three-dimensional (3-D) face-identification system useful for human faces and other image identification applications. The invention also provides a fast search engine for searching a large collection of two-dimensional mug shot photographs for those that match a particular facial image. In addition, the invention provides a technique for indexing 2-D mug shots to a uniform set of 3-D facial feature parts, such that a 3-D face surface images are generated based on the 2-D mug shots. Moreover, the invention provides a method by which a large number of mug shots may be reproduced using 3-D face surfaces by reference only to the uniform set of 3-D facial surface features and a composite code that identifies the facial feature parts to be used to generate a particular individual's facial image. This method avoids the need to locally store a large amount of data that would be required if every mug shot were stored locally. In addition, the invention has the advantage of generating a 3-D facial image from any desired point of view, and of providing a searchable code to 2-D mug shot pho
Chawan Sheela C.
Nixon & Vanderhye P.C.
Wheeling Jesuit University
LandOfFree
Three dimensional face identification system does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Three dimensional face identification system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Three dimensional face identification system will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3296583