Biometrics: Facial Recognition

INTRODUCTION: WHAT IS BIOMETRICS As the first few years of the 21st century unfold, technology is making great steps toward making life easier and safer. Daily life involves being part of an automated society and interacting with anonymous parties through the World Wide Web. As a result, identity has become an important issue. Simply knowing login identification to access sensitive information is costing the economy far too much in fraud. Transportation terminals have difficulty authenticating passports to ensure the safety of other travelers. Accordingly, Biometric Authentication is gaining use in several areas.
The definition of biometrics translates to life measurement, but more commonly the term is associated with unique physiological characteristics that are used to identify an individual. These personal features that are used in identification are the face, iris, and fingerprint. Typical settings are in security and corporate systems, point of sale and consumer electronics. Weaknesses in most sophisticated security systems have recently been discovered due to events like terrorist attacks. The most promising biometrics has emerged as the human face. This is due to the social acceptance and nonintrusiveness of its measurement through imaging. Minimal or no cooperation is needed from the subject which makes it ideal for surveillance and theft control. Typical applications for facial recognition software are; mug shot identification for law enforcement, gateways to limited access areas, personal identification such as driver’s licenses and credit cards, and surveillance or crowd behavior. Face recognition is thought of as the signifier of biometric identification because of significant advantages over other methods of identification. However accuracy is not the best known advantage to this type of technology. Although facial recognition software cannot compete with the error rates of other types such as, fingerprint or iris systems, it does have its own features. It is applied at its own convenience and does not need cooperation from its subjects. It is benign in its assault to the unaware bystander.
FACIAL RECOGNITION: HOW DOES IT WORK

SureMatch 3DT
Technest Inc. (formerly Genex Technologies) – Technest Inc. is a 3D imaging company which specializes in intelligent surveillance, imaging, high-speed sensor processing and image display technologies. We’ll limit our view of the company to Facial Recognition where they focus their efforts on Advanced Recognition Methods and Aging. Due to limitations in 2D facial recognition technology, companies like Technest are investing large amounts of R & D dollars into discovering new ways to quickly image and recognize facial features. They are also working hard to fully understand the way the human face ages, and key elements which they can look for or abandon in the effort to correctly identify a subject. Facial lines, resolution of the image and perceived wrinkles, weight changes, and bone breaks are only a few of the age related items which can make proper detection difficult. Technest believes that 3D systems offer huge advantages over legacy 2D systems and is resolute to develop methods which allow the use of vast quantities of 2D images in Face Recognition Systems (FRS) databases. According to Technest “When we look at a photograph, we see many distinctive characteristics and subtle nuances in a face. When a computer looks at a photograph, it sees pixels. One limiting factor is the flat, lifeless aspect of a 2D photo.” ("Genex technologies," ) This company is using key elements of facial features in 2D images, and years of research to understand aging in the face to create virtual 3D models of the face of a subject. The intention is that their 3D technology and tools will allow organization interested in migrating to new and developing technology to continue to use historical data or systems while moving into the future. For some organizations like law enforcement, that is a real return on investment to leverage the FRS which they currently employ.
A suite of tools for one purpose – The SureMatch 3DT (SM3DT) suite of tools by Technest, is unique compared to other facial recognitions solutions. It is a suite of five different discrete tools comprised to mitigate the issues normally experienced with facial recognition and can generally be consigned to the image enhancement or specialized matcher sectors of the market. Three major issues with 2D facial recognition solutions have historically prevented the technology from being adapted for mainstream use; or caused issues with accurate subject matching. SM3DT addresses the problems related to pose, lighting, and expression by using different applications within the suite. One of the newest pieces of the suite is the Capture 3D FaceCam which can capture a 3 dimensional geometry and texture detailed images of a subject in less than 400 milliseconds. For situations where non-compliant subjects are involved, this is an important capability. Additionally, unlike other historical 3D capture devices the lighting used by the camera is not harmful to human eyes. The ability to analyze small facial features which are unique is critical in accurate matching. Lines, points, and irregularities all add to the identification of subjects and can be captured in greater detail with 3D imaging devices. Compare 3D is another application in the suite which uses the results of captured geometry and texture to do exact matching of 3D images. This is a far more accurate comparison than historic systems could perform. But the real value in a suite like SM3DT is the ability to tie together the thousands of images in databases which are all in a 2D format and have been compiled from poor imaging solutions dating back many years. “…when surveillance photos are compared with a database of frontal enrollment pictures, matching performance drops dramatically and positive identification fails.” ("Genex technologies,"). Recent research has allowed progression in this area. Technest “Enroll allows seamless integration with existing 2D recognition databases.” ("Genex technologies," ) Using 3D Enroll a 3D image taken from a subject can be converted into many varying 2D images accounting for different lighting situations, expressions, and poses. A single resulting 2D image generated with Enroll 3D can then be compared to an existing database to find a suitable match. Replication of the off-pose angle and lighting environment can be simulated in a preprocessing environment and then compared to available images. This greatly increases the accuracy of the comparison and offers more precise matches than have been experienced previously. Especially in cases where off-pose or low quality surveillance video is used. Next, the 3D Sketch Artist component of the suite enables sketches of subjects to be manipulated in a virtual environment. “Using Genex's patented 3D morphing technology, law enforcement professionals can now sketch a truly accurate composite with amazingly life-like 3D features. 3D SketchArtist transforms typical sketches into rapidly evolving 3D mock-ups that can change poses, expressions, and even lighting in mere seconds.” ("Genex technologies," ) Technest acknowledges the difficulty and the effort required by law enforcement agencies to accurately sketch the face of a subject. Traditionally sketching subjects for law enforcement agencies meant hiring specialized professional sketching artists, and “Accurately adding features or adjusting existing elements such as hair or age will usually require a complete re-draw or significant time and effort.” ("Genex technologies," ) With the 3D Sketch Artist application investigators or professional sketching artists can manipulate any feature on the sketch. This feature enables more accurate depictions of suspects and saves many hours while working with witnesses. Since so much time is saved, the memory of the witnesses is more current and accurate. This improves the matching capability even further by helping to eliminate the human error factor. The computerized sketch can even be shared between the applications in the suite to allow for comparison and matching when the sketch is complete. 3D FaceMatch for SureMatch 3D Suite is another tool which allows for interoperability of 3D face maps with other biometric company’s solutions. It “provides both an enhancement for existing biometrics users and market-leading performance for new deployments.” ("Genex technologies," ) Despite the flexibility which the application suite offers in legacy solutions or in subject comparison, the company pushes two major integrated solutions. The systems are designed for circumstances where biometrics is used as a means of positive identification for entry, or where real time scanning of many individuals is required. The SureMatch 3D Suite for Watchlist is aimed at tackling the issue of high volume traffic areas. The system is built to compare large volumes of subjects against a ‘Watchlist’ of known terrorists. The strategy of the system is to convert the volume of 2D watchlist images into an enhanced database with 3D geometry and texture detailed images. Once the 3D data has been created using specialized algorithms, the 3D data is converted back to a collection of 2D within the database accounting for changing pose, lighting, and expression conditions. On the fly comparison can then take place as subject images are taken from strategically placed image capture devices, and “The result is a SureMatch 3D "enhanced" database that enables 2D recognition systems to achieve superior performance.” ("Genex technologies," ) Finally the Access Control integrated system is a combination of hardware and software which is intended for organizations which need to positively identify trusted individuals for entry. This is a more focused solution where care can be taken during the subject image acquisition process. Unlike populous situations where subjects are not necessarily compliant, this systems would be more accurate and offer much greater recognition capabilities. Technest also offers a software developer kit for installations where multiple models are required for positive identification. All of the products offered by Technest are plainly designed with security, law enforcement or government requirements in mind. For simple access to your Small Office or Home Office (SOHO) computing resources, Technest is likely not the solution.

Cognitec Systems
Cognitec Systems founders have been working on algorithms for the FaceVACS (face recognition technology) since 1995. One year later government and industry customers started to rely on the FaceVACS technology for many of their applications. Then in 2002 Cognitec Systems was founded due to the rapid growing need for software solution in biometrics. The headquarters are based out of Dresden, Germany, but have offices in the United States and Hong Kong. Cognitec not only develops market-leading face recognition technologies and applications, but are also the leader in the technology available on the market. They only focus on facial recognition products. Their products include products for facial database search, video screening, border control, ICAO compliant photo capturing and facial image quality assessment.
Cognitec is using three dimensional facial recognition devices, which is a fairly new method to come onto the market. How this new method works is it uses 3-D sensors to capture the facial shape information. It then identifies distinctive facial features like the contour of the eye sockets, nose, and chin. It also is determining the structure, shape and proportions of the face for example the distance between the eyes, nose, mouth, jaw, upper outlines location of the nose and eyes, and the area surrounding the check bones. Cognitec sees many advantages to using three dimensional facial recognition because it is not affected by changes in lighting, it can identify a face from a variety of angles, and it can be done from a distance without the individual even knowing that they are being scanned. This is a huge benefit in comparison to other biometric methods like fingerprint scanning or iris scanning which have to be with the individual knowing that it is taking place. The reason why Cognitec continues to expand is because there is nothing else like facial recognition since it can be used for surveillance purposes. It aids in searching for wanted criminals, suspected terrorists, and missing children. In fact the Los Angles Police department just found a wanted criminal because of the facial recognition that was purchased through Cognitec. Even though Cognitec uses three-dimensional imaging it is the software that examines the individuals that separates Cognitec from it competitors. They use software algorithms that identify faces by taking features from an image of a subject’s face. The algorithms normalize a gallery of face images and then compress the face data, only keeping the data in the image that can be utilized for facial recognition.
Cognitec’s FaceVACS has four different versions available for use through Cognitec. However, I will only focus on only two out of the four products, the fourth product that they offer is just an enhancement product, and the other product is used for two dimensional photos like ids. So we will focus on the other two individual products, which can be used in conjunction with each other.
The FaceVACS-DBScan with examiner surveys databases of biometric identities. This product is best used for verification of someone is in a large group of people. This device can compare non-similar identities automatically. The comparison process saves a lot of time because when the software is finished it only leaves the matches that closely resemble the facial scan. An added benefit is the software is doing the comparison so there is no room for human error. The examiner component of the product allows for enhancement of the images. If images are retrieved in low quality the image-processing filter can improve the quality of the images. The correction filter which uses 3D modeling technology, can generate frontal views of faces taken under lateral angles. This is a huge benefit to the enforcement officers using surveillance videos to try and obtain a criminal. The examiner tool also provides a set of inspection tools that aids in identifying the person in question in a timely manner. It eliminates any delay in the investigation, which is very important because it is everything when you are trying to catch a criminal. This product incorporates the new B6 algorithm, which was developed by Cognitec with higher matching speed. This product can run on Windows2003 server/ XP/ Windows 2008 server/ Windows 7. It also can run on Linux SLES 10. It does require a structured query language (SQL) database and 10g/11g oracle is recommended for use.
FaceVACS-VideoScan can work in conjunction with FaceVACS-DBScan because it can automatically scan incoming video streams, detect multiple faces, and will do compare checks for possible watch list matches. The difference between the VideoScan and the DBScan is that the VideoScan is live streaming video with real time face tracking on multiple video streams. The latest version of this product combines state-of-the art cameras and computer equipment. It can be used to detect criminals, but it can also be used to protect individuals of high statue when in large crowds. It also has an integrator kit for video systems. It works with C++ application programming interface (API) and web services API high scalability in terms of watch list size and number of video streams. This product uses comparison algorithm B5T8, it supports most camera types, and has improved camera control. It supports firewire, USB cameras and frame grabber connected cameras through DirectShow interface. VideoScan runs on Windows 2003 Server/ XP/ Windows 2008 server. It also runs on Linux SLES 9/10. This product also requires an SQL database and a 10g/11g oracle is recommended.
Cognitec’s products are mainly designed for law enforcement, but the easy to use applications also allows for businesses to use these products to protect their businesses. Even though that is a much smaller market it is a market that is growing. Cognitec realizes this that is why their products are tailored toward these individuals as well. They are not going to miss out on the growth in this market. Just think if more businesses started using these products when something did go wrong it would really help law enforcement catch the individual that committed the crime.
HOW ARE THEY DIFFERENT
3D facial recognition software overcomes shortcomings previously associated with 2D facial recognition software. Increased accuracy is accomplished by adding another dimensional plane. Differences in lighting are compensated for, along with particular pose position. Enough invariant information is given to deal with changes in expression, and other variables affecting a person’s natural look..
Since several of the applications for facial recognition software involve stationary and moving subjects, it is important that the images obtained are still viable in making a match. Capability to compare multiple angles of the face to a variety of possible images is something that 3D Enroll by SureMatch can do. Similarly, FaceVACS by Cognitec systems is able to match scanned images with less than perfect conditions. A difference that SureMatch has is the ability to analyze texture in the image.
Particularly benefiting law enforcement officials is the 3D Sketch Artist in the SureMatch suite. It allows for changes to be made to a sketch artist’s rendering on a 3D scale and this allows a more vivid example for witness confirmation.
It appears to be the dissimilarities of these two products is the mode of application. SureMatch is aimed more for larger agencies such as government uses. The ability to utilize and customize the multi-faceted range of programs that SureMatch offers makes it the sure winner, for developmental use.
For this type of software to be effective, it must be accurate and fast.

RECOMMENDATIONS
Body

REFERENCES & SOURCES OF INFORMATION

Genex technologies - 3d facial recognition products. (n.d.). Retrieved from http://www.genextech.com/pages/608/3d_facial_recognition.html ATTACHMENT #1