Law enforcement agencies have long used any means of verifiable markings to link a suspect to eyewitness accounts of crimes. As a fairly permanent and distinct marker in identifying an individual, tattoos have been an invaluable tool in pursuing “persons of interest” over the years.
Now, science has brought us a tool to help the process of searching through databases of thousands of tattoos to find that special someone, that much easier and quicker.
Anil Jain and Jung-Eun Lee of Michigan State University’s Department of Computer Science and Engineering have developed a new methodology to categorizing and identifying scars, marks, and tattoos (SMTs). They have labeled the new process Tattoo-ID and believe that it will help law enforcement agencies more accurately and quickly link an SMT with the individual they are interested in.
Currently, law enforcement agencies use the standards for SMT classification as stated by ANSI/NIST-ITL 1-2007. Which, according to Mr. Jain and Mr. Lee, is subjective, time-consuming, and is not scalable to meet the rapid growth in tattoo design.
With Tattoo-ID, the researchers believe their method can meet the needs of SMT identification as the needs of law enforcement grows.
Our approach is one of content-based image retrieval using features (e.g., color, shape, and texture), instead of labels or keywords, to compute the similarity between two images.
Currently the program is seeing 835 out of 1000 images correctly identified with the first attempt out of a database of 64,000.
Although blurred images and low quality image sources create lower success rates, Mr. Jain and Mr. Lee feel that by tweaking their current process and with the addition of new algorithms in their software, the tool will be able to resolve a larger number of SMTs quicker and more accurately, with even larger image databases.