Image analysis – Applications – Personnel identification
Reexamination Certificate
2002-07-15
2004-06-15
Mehta, Bhavesh M. (Department: 2621)
Image analysis
Applications
Personnel identification
C348S014090, C348S014130, C348S211120, C382S125000, C382S172000, C382S294000
Reexamination Certificate
active
06751340
ABSTRACT:
BACKGROUND OF THE INVENTION
There are common features between infrared (IR) and visual images of the human body. Using the face as an example, head shape and size, and the relative location, shape, and size of features such as the eyes, mouth, and nostrils are the same in both imaging modes. A database of images can be segmented into classes using metrics derived from those common features, and the same classification will be obtained from either visual or infrared images. Height can be also used as a classification measure when it can be inferred from the collected image or from separate sensor data. An infrared image of an unclothed area of the body, such as the face, presents much more detailed and person-specific information than does a visible image. However, visible images are more commonly collected and large historical databases of visual images exist. It is therefore desirable to automate a process for comparing imagery from both the visual and infrared modes.
Infrared images are unique to each person, even for identical twins. Visual images are not unique because many people look similar and can disguise themselves to look enough like one another that an automated identification system cannot distinguish them. Therefore, in a large database, it is not possible to automatically perform a one-to-one linkage between infrared and visual images because the visual images are not sufficiently unique. However, for each infrared image, an automated system can eliminate all visual images which cannot be a match due to insufficient correspondence between minutiae characteristics. In general, it is estimated that more than 95% of a visual database can be eliminated as a match to a given infrared image. This has application to the use of infrared surveillance cameras to identify wanted persons for whom only visual images are on file. The infrared-visual matching system compares each person it sees in infrared and classifies him as either a possible match to someone on the visual image watch list or not a match. Persons who are possible matches can then receive greater attention from immigration or security authorities. This allows the use of infrared surveillance imagery to proceed without waiting until a large database of infrared images is established.
The use of infrared imagery also provides for the detection of disguises, whether worn or surgical, which may not be detectable from visible imagery. For example, artificial facial hair such as a mustache is readily detectable in an infrared image although it appears natural in visible images. The fact that infrared surveillance imagery shows a man with a fake mustache provides a clue to consider in matching against a visible image database. Surgical disguises such as a face lift leave telltale short and longer term variations in the facial thermogram, while the visual image may appear to be a different person and show no sign of surgery. The ability to detect in IR images that surgical changes have been made to a particular area of the face permits an automated system to broaden the parameters for searching for possible matching visual images in an historical database.
High definition visual images of the face and body are routinely produced and stored for medical, diagnostic and forensic use. Common examples are the photographing of criminal suspects through booking stations producing “mug shots”, driver's license photographs produced by each state, and passport photos used by the State Department. Many such large facial image databases exist, in hardcopy and in electronic form, and there is increasing research ongoing into automated matching of newly taken images with those databases. For example, there are frequent attempts to match surveillance images of a person using a stolen credit card at an ATM with photographs of persons previously convicted of similar crimes.
Visual imagery, particularly from surveillance cameras, is often of poor quality due to dim illumination at the scene. Low light level or infrared cameras are expected to become more widely used for surveillance as their cost diminishes. There is therefore a need to correlate between newly acquired infrared images and existing databases of video images. Even in the future, when simultaneous collection of video and IR images will generate correlated databases, there will always be a need to match images taken in one spectral domain with images taken in another. This can include matching images taken in one IR band (such as 3-5 micron) with images taken in another IR band (such as 8-12 micron).
Since IR cameras are passive, emitting no radiation and therefore presenting no health hazards, they may be used in conjunction with other imaging medical devices such as x-ray, sonogram, CAT scan devices, etc. Minutiae derived from the IR image may then be superimposed or annotated onto the resulting medical image. This presents a standard technique for generating standardized reference points on all medical imagery. Subsequently, the method and apparatus of this invention can be used to search a database of annotated medical images to find a match with a current IR image or current medical image annotated with IR minutiae.
Regions of Interest (ROI) may be utilized instead of minutiae, where the ROI may be elemental or other shapes including fractal or wavelet-derived structures, segments of blood vessels, locations underneath or otherwise relative to tattoos, moles, freckles, or other distinguishable features, or wiremesh or finite elements used for thermodynamic or visible modeling of the body. Rules may relate the shapes and positions of such elements, their centroids and other features. Time sequences of minutiae and ROIs may be compared, with the decision as to a possible match made on the basis of cumulative thresholds and rule tolerances over the sequence.
Facial expression and speech modeling has application to synthetic videoconferencing and face animation. Substantial bandwidth and storage reduction can result. Use of IR minutiae offers more precise modeling than current use of visual images. The present invention provides a technique by which IR images can be tied to the visual image being displayed.
BRIEF DESCRIPTION OF THE PRIOR ART
The identification of persons from infrared images is known in the art as evidenced by the Prokoski et al U.S. Pat. No. 5,163,094 which discloses a method and apparatus for analyzing closed thermal contours, called “elemental shapes” which are created by the vascular system interacting with the anatomical structure. Fifty or more elemental shapes can be identified for example in a human face imaged with an IR camera which has an NETD (noise equivalent thermal difference) of 0.07° C. and a spatial resolution of 256×256 pixels. Characteristics of those shapes, such as the centroid location and ratio of area to perimeter, remain relatively constant regardless of the absolute temperature of the face, which varies with ambient and physiological conditions. Two infrared images are compared by comparing the characteristics of corresponding shapes. A distance metric is defined and calculated for each pair of images. If the value is within a threshold, the two images are considered to be from the same person.
In the Prokoski et al U.S. patent application Ser. No. 08/514,456, there is disclosed a method and apparatus for extracting and comparing thermal minutiae corresponding to specific vascular and other subsurface anatomical locations from two images. Minutiae may be derived from thermal contours, or may be absolutely associated with specific anatomical locations which can be seen in the thermal image, such as the branching of blood vessels. Each minutia is then associated with a relative position in the image and with characteristics such as apparent temperature, the type of branching or other anatomical feature, vector directions of the branching, and its relation to other minutiae.
The comparison of thermal minutiae from two facial images is analogous to the comparison of sets of fingerprint minutiae, in that two images are
Desire Gregory
Laubscher, Jr. Lawrence E.
Mehta Bhavesh M.
LandOfFree
Method and apparatus for aligning and comparing images of... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for aligning and comparing images of..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for aligning and comparing images of... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3303498