Gangs use graffiti, not only to deface public property, but to mark territory and communicate. Identifying these markings can aid police in crime control, since it is estimated that some 80 percent of US crime is gang related. A computer scientist has developed a system to do just that.
Graffiti can be monochrome and simple or colorful and complex
Anil Jain at Michigan State University (MSU) has a program to compare an image to a data base to help identify which gang is responsible for the piece. Sometimes the program can even point out which person did the drawing. Graffiti is drawn by hand, so each piece, like anything handmade, is slightly different. However, Jain?s system looks for key points. It uses scale invariant feature transform (SIFT) to extract distinctive and salient feature points.
Hand drawn text often requires human intervention to decipher
So far the system is by no means foolproof. Graffiti often involves text, including the gang?s name. That text is put before people working in Amazon?s Mechanical Turk service which provides an on demand, scalable workforce to perform human intelligence tasks. Humans decipher what Jain?s program is not yet sophisticated enough to interpret.
Jain, whose research includes pattern recognition, computer vision and biometric recognition, has had previous success in other fields of identification. His work on altered fingerprints, Beyond Fingerprinting (PDF download), was given passing recognition in the 2008 CSI:Miami episode, Wolfe in Sheep?s Clothing. He held six patents related to fingerprints which were transferred to IBM in 1999.
Working together, MSU and Tsinghua University?s Department of Automation in Beijing, China developed an algorithm that analyses fingerprint images and marks them as normal or unusual. The unusual ones are then examined manually to look for alterations. Jain explains: "Government agencies worldwide encounter individuals who have gone to extreme measures to alter their fingerprints to avoid being identified by Automated Fingerprint Identification Systems (AFIS)." Fingerprints can be disfigured by burning or cutting to scar the fingertip.
Tattoos, on the other hand, are embedded so deep in the skin that even severe skin burns rarely destroy them. Tattoos are used not only to identify the bad guys, but also to identify victims, such as those killed in the 9/11 attacks in New York. Skin art is said to be more distinctive in identifying a person than the traditional demographic markers of age, height, race, and gender.
The National Institute of Standards and Technology?s Data Format for the Interchange of Fingerprint, Facial & Other Biometric Information (ANSI/NIST-ITL1-2011) defines eight major classes, such as human, plant, animal, or symbol, and has 70 subclasses therein such as skull, rose, dragon, for matching a query image. However, its success is limited due to the subjective nature of labeling. Tattoo-ID, Jain?s precursor to his graffiti identification software, is instead a content-based image retrieval system that extracts features from a query image and retrieves near-duplicate tattoo images from a database.
The Biometrics Research Lab at MSU presented a paper in 2010, Graffiti-ID: Matching and Retrieval of Graffiti Images that illustrated just how invasive gangs are, noting that one affiliation has 50,000 members worldwide with 10,000 of them in the US. At the time of the presentation, they had a Web DB which contained 1,265 graffiti images downloaded from the web, and 5,000 images from a single Southern California city in the Cal DB. Gangs, unfortunately, are prevalent in the west coast state. San Diego, for example, is known to have at least 90 gangs. The graffiti software?s average time for feature extraction was 1.08 seconds with an additional 27.33 seconds for matching, using an Intel Core 2, 2.66 GHz, 3 GB RAM processor.
Jain?s team will present at ACM Multimedia 2011 in Scottsdale, Arizona November 28th through December 1st. Other topics will cover multimedia systems and middleware, security in multimedia related to source identity, content integrity, privacy, trust and anonymity. Arts and Contemporary Digital Culture covers the innovative use of digital technology in the creation, analysis or critique of cultural artifacts, tools for content preservation and curation; environments, dynamic and interactive multimedia artworks and processes. Graffiti itself, in general, has evolved into a world wide art form.
Not all graffiti carries an offensive message
Gang graffiti specifically can be likened to hobo codes or Underground Railroad markers; it acts as a bulletin board, understood only by the initiated. Studies are underway by both scholars and law enforcement.
Graffiti, like rap, has hit the artistic level. However, Graffiti is Dead, a DVD of graffiti artists from around the world, isn?t mainstream and you won?t find it on Amazon. One writer commented that whether the work is by a gang member or an acknowledged master such as Salvador Dali, the observer is challenged to decipher the hidden meaning. Simple tagging is analogous to signing your autograph; bombing takes the tag a step further, incorporating more information into the design. Entire pieces of graffiti have become the inner city gang?s artistic expression. Graffiti adorns subways, neighborhood fences, alleyways, sidewalks, street signs. Although still considered to be vandalism, once in while it conveys a seasonal message, rather than a nefarious one.
Where do these hoodlum street artists come from? The community that produces gang members typically is an economically depressed urban center where schools are underfunded, jobs are scare, and prospects for advancement are rare. We may be producing artists with unique talents, but we are also producing kids who are alienated from society, angry, and dangerous. Creating software to identity them after the fact isn?t the answer to eliminating the problem.