Presentation on theme: "FACE RECOGNITION HISTORY, TECHNOLOGY, IMPACT AND RESEARCH."— Presentation transcript:
FACE RECOGNITION HISTORY, TECHNOLOGY, IMPACT AND RESEARCH
OUTLINE What is Face recognition? Brief History of the approaches and techniques Where and how is it used today? Current state-of-the-art and ongoing research Social and moral implications
A HARD PROBLEM Humans (and non-human primates) are really good at face recognition and many other forms of pattern recognition – handwriting, speech, abstract illustrations. These skills probably helped us survive in the face of predators, identify close family members, and analyze facial expressions Human infants possess this sophisticated skill at very early stages of development But because of the complexity of the problem, computer algorithms that can perform face recognition have only been developed fairly recently
FACE DETECTION VS. FACE RECOGNITION Face detection is the first step – using a technology to locate a human face in a complex image that contains noise and distracting objects, and a face that might have a variety of orientations, illuminations, partial obstructions, etc. Retrieved 2/1/11 from How many faces are there in this picture? (at least 13, maybe more….)
FACE DETECTION The technology is now very common – many consumer cameras have face detection built in to facilitate focus on a person; Google, Flickr, Facebook and many other web sites perform face detection (and even the next step, recognition) on images. The algorithms employ a variety of methods, including pattern matching, identifying particular facial characteristics, and even looking for areas of skin tones Downloaded 2/1/11 from
I’VE DETECTED A FACE, HOW DO I RECOGNIZE IT? Start by storing one or more images of the face that will need to be subsequently recognized. These images are processed in a way that will allow the recognition algorithm to compare the features of the target image to the stored image(s) Starting in the 1960s, some procedures were developed that were semi-automated, it required an operator to manipulate the image so that a computer algorithm could process it successfully.
ALGORITHMS Geometry-based methods – compare distances and other geometric features on the detected face with those of the stored image, an try to identify the closest match Direct Correlation Methods – Compare the detected face with the stored images, and find one with the closest match by statistically comparing color and other dimensions pixel by pixel Many other complex mathematical approaches are also used Retrieved 2/1/11 from 1/face_recognition.jpg
USES: SURVEILLANCE Retrieved 2/1/11 from cartoon.jpg Large populations and public areas can be surveyed and persons of interest (suspected terrorists, missing children, etc.) could be automatically identified.
USES: SECURE LOGIN AND ACCESS Many computer systems, including Windows 7, now enable secure login through face recognition; this supplements traditional methods such as passwords and fingerprint swipes and is potentially more secure Retrieved 2/1/11 from blink-face-recognition/screenshot-yjsynopf.html
USES: FACE RECOGNITION IN CONSUMER APPLICATIONS Retrieved 2/1/11 from photo-gallery-facial-recognition/3966/ Face recognition is now used in many consumer photo applications – Flickr, Picasa, Facebook, and desktop photo organization software.
ADVANTAGES OF FACE RECOGNITION Retrieved 2/1/11 from security.gif/ /biometrics-security.gif As a security technology it is less intrusive than other biometric techniques – it requires no physical contact or even cooperation from the subject
CHALLENGES TO FACE RECOGNITION Retrieved 2/1/11 from s.aspx Face recognition suffers from several deficiencies. It can be easily circumvented in security situations by a disguise. It also depends heavily on the quality of the captured image, and therefore is very subject to illumination in uncontrolled environments
HOW ABOUT PEOPLE THAT LOOK VERY SIMILAR, E.G. TWINS, OR CORPORATE EXECUTIVES? Retrieved 2/1/11 from orange.fr/biometrics/types/face.htm
FACE RECOGNITION AND BIOMETRICS Retrieved 2/1/11 from Attribution: By Welleman (Own work) [CC-BY-2.5 (www.creativecommons.org/licenses/by/2.5), via Wikimedia Commons Because facial recognition on its own might not be sufficiently reliable for some applications (e.g. it can easily be circumvented by disguises), it can be supplemented by other biometric techniques such as fingerprints, retinal scans, voice analysis, etc.
FACE RECOGNITION: FUTURE OPPORTUNITIES Retrieved 2/1/11 from detection.htm Retrieved 2/1/11 fromhttp://www.sheknows.com/love- and-sex/articles/808147/signs-hes-lying Face recognition algorithms are being expanded to identify features that correlate with various psychological states and emotions. Is there any validity to TV’s “Lie to Me” ?
SOCIAL, MORAL AND LEGAL IMPLICATIONS Retrieved 2/1/11 from communities-for-afghanistan/ Staff Sgt. Matthew Valek uses a biometric camera to store data for an Iraqi worker at a military base in northern Baghdad in Sept.http://www.aolnews.com/2009/11/16/coalition-plans-gated- communities-for-afghanistan/ Biometrics are used by the U.S. military as a key tool for identifying insurgents in Iraq and Afghanistan, but this program expands its application to the civilian population in increasingly restive Helmand.identifying insurgents Because facial recognition can be performed without any awareness by the subject, does this constitute an invasion of privacy or personal freedom? What are the consequences if we biometrically identify a complete population? Perhaps this is already being done, since in many populations government agencies already have photographs of all individuals (via passports, drivers licenses, student IDs, etc.)
REFERENCES Facial Recognition Technology: A Survey of Policy and Implementation Issues Face Recognition Facial Recognition recognition-computers-can-see-through-disguise.html