Thursday, January 20, 2005

Bonding Tsunami Victims with Facial Recognition

Ushering in 2005 was an unforgettable experience to Edwin Yang and his colleagues. After Phuket of Thailand was ravaged by the tsunami, they went to local hospitals and temples to take images of the bodies or surviving victims.

For those four days, the director of Regal Cyber in Hong Kong and his team managed to capture some three hundreds images. And combined with the contribution of database from the International Victim Coordination Center at Phuket, the company established a facial recognition web site to assist families in searching for their missing members in the catastrophe.

The web site,, was set up on December 31st, 2004. It allows relatives to upload a photograph of their missing friends or family. The images will then be checked against a growing database of the faces of surviving or dead victims.

The PeopleMatch system has the ability to place up to 1,700 unique characteristics for an individual face in its database. It will then match these points with similar points from an uploaded facial photo. Upon verification, the system will call up a set of photos that matches the victim.

Due to the factors such as facial injuries, as long as reaching to a recognition rate of 65 percent, the web site will notify the respective families immediately, Yang told A&S.

He informed that four matches were successfully done in the first week of January. And the database has collected up to 1,300 facial images just within the first few days after the Internet set-up.

In addition to setting up the PeopleMatch’s platform, Regal Cyber partnered with local distributors of Viisage and Cognitec in Hong Kong. The two companies supplied facial recognition solutions for free for back-end systems.

“In this case, facial recognition is the most convenient tool to help the sufferers. It is a much faster alternative than DNA or tooth identification,” he stated. And indeed, the biometric technology has offered hope in the natural disaster.

Tapping onto Facial Recognition

Facial recognition industry looks set to achieve continuing growth this year. However, vendors said that there are still some issues to be ironed out before it can embark on full-scale growth.

The worldwide market of facial recognition applications this year is expected to experience continuing growth.

The most significant development in 2004 was the progress made by the ISO SC37 working groups, which worked on the standardization of biometrics in travel documents, an official with Cognitec, a German facial recognition company, informed A&S.

“This development, initiated and supported by the International Civil Aviation Organization (ICAO), is opening up one of the most important business areas for biometrics in the near future,” the official claimed.

And the positive outlook is expected to go on in 2005.

“In the area of facial recognition, we have seen several technological advancements during the last 12 months. This year, we will have to bring these improvements to the market. Another fast-developing area is happening in the mobile biometric devices, where we have seen a lot of progress,” said the official.

The government sector will apply this technology widely in 2005 as there are currently a number of initiatives within homeland security and international travel security taking place.

Facial recognition technology is needed for issuance of travel documents and verification of biometric solutions containing facial biometric data. Therefore, electronically-readable travel documents will boost the implementation of facial recognition applications.

“Once biometrics of travel documents is perceived as a commodity, we expect growth to take off in other areas as well,” the official added.

Anton Kuip, a biometric expert with Nedap’s security management team, agreed. He said that the technology will be used in passports and the first ISO standard is expected to take off this year. Nedap is a Netherlands outfit delivering solutions for security management, access control and biometric identification.

“This will most likely promote the use of facial recognition in general,” he stated.

Joining Forces with CCTVs

Indeed, the revolution of the industry is underway with the participation of CCTV vendors.

“Facial recognition is an interesting technology. It was introduced along with the development of CCTVs,” commented Jison Hsu, biometric solutions division manager of PenPower Technology Ltd. It is a Taiwan-based company offering recognition technologies including handwriting, voice, facial recognition, as well as wireless communication.

“Nowadays, surveillance cameras can be everywhere. People are starting to change their mindset that although cameras do invade their privacy to a certain level, they have become a part of everyday lives for better security,” he claimed.

As surveillance cameras at the streets, buildings and public places become more prevalent, people will get used to being watched at all times. This will spur the growth of facial recognition solutions as people are more at ease with them.

“The combination of facial recognition with CCTVs will definitely be the future trend,” Hsu stated.

Cognitec seconded, adding that facial recognition can be integrated in existing video surveillance application and supports security officers with hint about suspect persons that are previously enrolled. “This is the most challenging face recognition technology application and still at the very beginning,” an official said.

According to Hsu, CCTV systems nowadays can manage to cope with the requirements of general surveillance. However, it doesn’t mean that CCTV vendors therefore have a lower entry barrier into the biometric field.

These makers encounter numerous obstacles when exploring the market. The issues to be resolved are the existing problems of CCTVs: hardware, data storage, compression technologies, transmission speed and picture-capturing under various lighting conditions. To cite an example, CCTV players have been working hard to capture images under extremely dim lighting environment in recent two years.

However, facial recognition will be a valuable add-on feature to CCTVs. It is an extra intelligent function that says the system will be able to capture human faces and even identify the faces, he added.

Furthermore, CCTV solutions will become more powerful with the integration of facial biometrics. For instance, it will stop recording when no human are present within the cameras’ viewing areas, and will record at a slower speed simultaneously, thus cutting down data storage in return.

To his understanding, not every CCTV vendor has jumped into the biometric rush. They might have sensed the trend, and started to treat it as a long-term investment and do not expect fast adoption.

“But facial recognition functionality will set them apart from others who do not have this intelligence,” Hsu cautioned, advising CCTV and biometric-solution providers to form synergies to complement each other for the research and development.

Time for 3D?
It is becoming well acknowledged that 3D (three dimensional) facial recognition technology yields more accurate performance than the commonly-used 2D approach because 3D measures the depth of a face.

Because of the additional depth measurement, 3D is robust to facial pose variations and illumination changes, both of which are key factors that degrade the accuracy of 2D.

According to Kuip, most leading facial recognition vendors, including Viisage, A4 Vision and Identix, are now working on 3D.

“This technology allows the user more comfort and flexibility as it does not require looking straight into the camera. It can also recognize the faces of people passing by, thus enabling a faster people flow,” he explained.

As in the case of Cognitec, it has also been investing in the development of 3D facial recognition algorithms.

According to the company, 3D promises more robustness of biometric comparison against lighting and pose variations compared to 2D. However, the introduction of 3D requires complete new data gathering and will therefore be more expensive.

Hsu added that though 3D offers a more precise recognition rate, its computational loading is huge. It is thus impossible, at least for now, to provide “real time” recognition as the process will take much longer time than 2D.

As such, 2D is still the mainstream of the industry with 3D in the emerging stage.

Moving Forward
Compared to other forms of biometric applications such as fingerprint, iris recognition, or palm recognition, facial recognition is the only one that requires no contact with sensors at all.

However, “biometric-based recognition technologies inevitably have certain limitations,” Hsu noted. Compared to fingerprint--its more established biometric counterpart, the facial technology still has some distance to catch up with.

“The first applications for tracing wanted faces were unfortunately not very successful to this date. But the input from an automatic facial recognition system is always a helpful support for any kind of video surveillance system as facial technology offers intelligent tracking,” added Kuip.

Currently there is a great interest from airports in the use of facial recognition for both access control and video surveillance. One example of a situation in which it would be useful is passenger recognition at the moment of exiting an airplane. These passengers would then be traced again when appearing at the airport customs a while later.

However, the technology is not reliable enough yet for such applications at the moment.

Another area in which demand is surfacing quite strongly is in sports, for example checking the faces of fans at the entrance of football stadiums to trace or block particular persons, he continued.

As early in November 2002, Sydney International Airport had taken the lead in the industry by deploying new automated border crossing system with facial technology.

With this SmartGate system supplied by Cognitec, some 3,000 Qantas aircrew members arriving at the primary line, instead of presenting their passport to a customs clearance officer, place their passport on a reader, and look at a camera.

The gate opens after customs-and-immigration checks have been automatically performed by the system with facial features compared to the stored templates. The whole process is said to take less than 10 seconds.

Therefore, this process greatly enhances efficiency and at the same time maintains a high quality of secure border control.

Hsu envisions that such application cases which offer convenience and security will continue to mushroom at a larger scale this year.

“And as they become more commonplace, more people will consider deploying CCTVs with facial recognition in the next two to three years,” he summed up.