Feature

Back in the late 1980s, I remember a prominent pattern recognition researcher being asked about techniques in our field that were in extensive commercial use. His response consisted of just one item, bar codes. To be fair, pattern recognition systems were already in use – for narrow applications – before this. Optical scanning of multiple-choice forms (such as school tests) was already performed. Money changing machines could distinguish paper currency. The military had so-called smart bombs, which could distinguish targets by their silhouettes (featured in the January 2004 issue of the IAPR Newsletter). And rudimentary speech recognition of “yes”, “no”, and numbers was in use by phone companies. All these systems used pattern recognition techniques to perform classification. However, for most of these, the problem was very constrained to enable the technology of that time to work (and for any of us who tried to change a slightly rumpled bill in a money changing machine, the technology did not always work!).

But that prominent pattern recognition researcher certainly identified a landmark event in pattern recognition, because barcodes were printed on every commercially sold item, they became a worldwide standard, and were in plain view of all of us. Directly following and influenced by barcode research, was optical character recognition (OCR). Although not evident to most people, OCR touched all of us through its adoption for postal recognition (which was featured in the April 2006 issue of the IAPR Newsletter [html] [pdf]). There were other pattern recognition successes subsequent to OCR, but this article focuses on some of the most prominent in the past six years that have been, not coincidentally, highlighted in this Newsletter.

Back in April 2003, we published a feature article entitled “Pattern Recognition in Sports”. This discussed the work of Gopal Pingali, Yves Jean, and Agata Opalach, then at Lucent Bell Labs, who had developed recognition technology for calling the lines and tracking players on televised tennis matches. The story described their debuts on CBS and ESPN broadcasts. This was a great pattern recognition success story because few things could be more prominent to people than sports on TV. Since then, pattern recognition technology has become commonplace in many televised sports broadcasts from tennis to football to hockey.

In June 2003, the feature article dealt with security and entertainment – a strange combination until one realizes that this refers to watermarking. Ingemar Cox, then at NEC Research Labs, and others were working on embedding watermark signals in pictures and movies to try to prevent illegal downloading, or pirating. Besides describing the technical underpinnings, the article also covered some of the political machinations that involved the Motion Picture Association of America, Recording Industry Association of America, DVD manufacturers, etc. If we fast-forward to 2008, it doesn’t appear as if this technology is winning the battle, however there is no question that it is still a hot topic both in and outside of technical circles.

In September 2003, the feature article described Herb Freeman’s and Steve Morse’s pattern recognition work at MapText for locating text labels on maps. That technology was initially used for US Census Bureau electronic map creation. Now electronic maps are everywhere from the GPS in your car to Google Maps, all using some pattern recognition technologies for their creation.

In the April 2005 issue [html] [pdf], the feature article explored not a commercial application, but one related to the basic science of astronomy. Building upon her insight on the data analysis needs in pattern recognition, Tin Kam Ho of Bell Labs, Alcatel-Lucent, built a tool for interactive pattern discovery, which has been incorporated into the project known as the National Virtual Observatory (NVO). This visual tool makes it convenient for astronomy researchers to analyze astronomical data from ground- and space-based telescopes worldwide that are archived in the NVO. As it turns out, many others find it useful in various engineering, business, humanities, medical, and military applications. Notably, Tin's larger body of work in statistical pattern recognition has earned her the 2007 Pierre Devijver Award, and she will present the Pierre Devijver Lecture at this year's S+SSPR workshop immediately before ICPR in December (see advertisement in this issue).

The tremendous leap in the number of surveillance cameras over the past few years elicits strong opinions both pro and con on their security application. The July 2005 issue of the Newsletter [html] [pdf] included a feature article on cameras that employed pattern recognition for improving traffic efficiency. Panos Michalopoulos, a professor at the University of Minnesota described the beginnings of the use of cameras in the field of traffic engineering for controlling traffic lights, and how later systems began to monitor all aspects of traffic flow. Now, OCR of license plates at 100 km/h is being accomplished for cars speeding through camera stations where toll booths once required full stops.

Despite the wide and public impact all these technologies have made, I’d say the one whose benefits impressed me the most was in the field of digital libraries. In the July 2006 issue [html] [pdf], the feature article covered the Greenstone Digital Library project, led by Ian Witten, a professor at the University of Waikato in New Zealand. Greenstone is a suite of software for building and distributing digital library collections, and is developed and distributed in cooperation with UNESCO and the Human Info NGO. One of the premises upon which it was built is that access to appropriate information is one of the overarching problems facing developing countries. Directly addressing this, Greenstone software has produced libraries on such topics as food and nutrition, humanity development, agricultural development, medicines, and a pictorial library for the world’s 20% illiterate (40% of those in sub-Sahara Africa, the middle-East, and South Asia) called First Aid in Pictures. The widespread, positive impact of Greenstone, in part employing OCR and document recognition technologies, makes one proud to participate in pattern recognition research and development.

Finally, although I can’t say that it has yet reached widespread use, the feature article that caught my sense of whimsy most was on pattern recognition for origami, the ancient Japanese art of paper folding. In the January 2005 issue [html] [pdf], I followed up on a technical paper I had read by Hiroshi Shimanuki, Jien Kato, and Toyohide Watanabe. This work used pattern recognition and geometric techniques (for which a 2003 MacArthur Award was given to Erik Demaine for his work in describing the polygonal shapes that can be folded from a single rectangular piece of paper) to guide the design from hand-drawn figure to a chosen 3-D shape. I liked two aspects of this story in particular: the use of pattern recognition for an ancient art and the researchers’ choice of a novel application.

There are many more fields where the use of pattern recognition has exploded in recent years. Biometrics is one that comes to mind (see January 2006 article which included NIST’s work on biometrics [html] [pdf]). Another is pattern recognition for cultural heritage (see IAPR Newsletter October 2004 ICPR Special Issue Invited Talk on “The Great Buddha Project” and IAPR Newsletter October 2006 Robert Sablatnig’s and Paul Kammerer’s report on the 1st EVA Vienna “Digital Cultural Heritage—Essential for Tourism and the EVA Vienna 2008 report in this issue). And there are many technologies that will appear from the labs in coming years.

On to more novel, beneficial, and impactful pattern recognition technologies and applications in the future!