[Lee Jong-gil's Autumn Return] Outdated Fingerprint Investigation Transforms into a 'Pinpoint Expert' with AI Added

Hunan Satellite TV's "I Am the Future" Production Team on "The Present and Future of Artificial Intelligence"

From March 29 to June 7, 2000, Park Hyun-gi's posthumous work was exhibited at the "3rd Gwangju Biennale." A video continuously shows each person's resident registration number appearing and disappearing over fingerprints. This signifies that countless humans are born and die, each possessing their own "code."

Fingerprints are geometric line patterns that appear on the surface of human fingertip skin. They form at 14 weeks of embryonic development and become lifelong unchanging patterns by 24 weeks. The epidermal fingerprint and the dermal fingerprint match exactly. Even if the epidermis is damaged, the dermis regenerates the same epidermal fingerprint as before. If lost, the original pattern can be confirmed from the dermal fingerprint. Francis Galton (1822?1911), a British eugenicist and the founder of modern fingerprint science, stated that a complete fingerprint contains 100 to 120 detailed features. These combine to create a unique fingerprint in the world, so he argued that fingerprints are unique characteristics that differ from person to person.

Ultimately, fingerprints exist from birth, remain unchanged throughout life, and differ for each individual. These characteristics make them a key tool for identifying suspects in public security and criminal investigations. They have been used as physical evidence in criminal litigation since the Song Dynasty in China. By the late 19th century, fingerprints were widely used in investigations and trials. In 1904, the U.S. government began collecting fingerprint data and adopted it as a strong form of evidence in 1911. Until the 1960s, fingerprints were classified and compared manually by people. However, this labor-intensive system became difficult to maintain as hundreds of thousands of fingerprint cards accumulated.

The book "The Present and Future of Artificial Intelligence," written by the production team of Hunan Satellite TV's "I Am the Future," introduces twenty-one core technologies leading the future of artificial intelligence (AI). "I Am the Future" is a science and technology program aired by Hunan Satellite TV in the summer of 2017. "The Present and Future of Artificial Intelligence" provides guidelines required in the AI era by combining future philosophy and technological directions with its content.

Xu Kai, chief researcher at the Third Research Institute of the Ministry of Public Security of China, is confident that the development of modern science and technology greatly contributes to national security and social stability. Chief Xu has achieved many research results in the fields of social public safety and public security scientific informatization. The public security science and technology team he belongs to has designed many core systems and equipment for the public security industry and promoted research, development, and practical application.

A representative achievement of this team is the enhanced computer-aided fingerprint comparison technology. Previously, fingerprints collected from crime scenes were initially analyzed and classified by experts. Opinions related to fingerprint features were entered into a computer database, and the fingerprints were compared to see if they existed in the database. Manual work compared all elements of the fingerprint. However, the computer-aided fingerprint method compares only the fingerprint features. These are broadly divided into three levels. Level 1 features are the types of fingerprint ridges: Arch, Loop, and Whorl. Level 2 features are the macroscopic shapes of fingerprints such as hook-shaped, eye-shaped, rod-shaped, and fork-shaped. Level 3 features are microscopic detailed characteristics of fingerprints, such as ridge endings, sweat pores, and wrinkles. The combination of these features and the overall distribution pattern of Level 3 features create a unique fingerprint in the world.

In the past, computer-aided methods could not properly utilize Level 3 features due to low resolution of fingerprint images. Therefore, they only partially assisted experienced experts. This method had clear limitations. The features entered into the computer system with expert opinions were only a part of the entire fingerprint features. There was a high possibility of errors or omissions in fingerprint comparison. A representative example is the train bombing terror attack in Madrid, Spain, on March 11, 2004. At that time, the FBI mistakenly arrested an American lawyer as the culprit based on fingerprints from the crime scene provided by Spanish police authorities. The reliability of fingerprint identification was questioned not only worldwide but also within the U.S. legal community.

The author of "The Present and Future of Artificial Intelligence" explains, "When determining matches using limited fingerprint features, errors can occur. Especially as the number of fingerprints entered into the database skyrockets, the number of similar fingerprints also increases, raising the risk of errors." The ever-expanding database is also an obstacle. In China, the amount of fingerprint data increases by more than 20% annually compared to the previous year. As data volume grows, computer processing speed slows and accuracy decreases. Because of this, the existing computer-aided method was once sidelined amid the flood of new technologies like DNA and smart image processing.

▲Artificial micro fingerprint that cannot be duplicated. [Photo by Korea Research Foundation]

Fingerprint identification technology, which seemed to be on the decline, has taken a new leap forward by meeting AI. The existing fingerprint comparison work, which abstracts and describes fingerprint images by combining features, requires a lot of empirical knowledge to recognize even shallow layers like partially concealed layers. The computer developed by Chinese public security authorities in 2017 greatly compensated for these shortcomings. It recognizes fingerprints as ordinary images and underwent deep learning training on the features of all elements and the natural layered structure. Thanks to this, it can compare and identify not only pixels and ridge shapes but also patterns and images as a whole. By mimicking the brain of a fingerprint analysis expert, the computer understands the different features of each fingerprint and determines whether two fingerprints match.

Chief Xu is preparing a large amount of computing resources for more accurate comparisons. By utilizing the largest and most powerful supercomputer center in the public security field and integrating AI technology, a major innovation in traditional fingerprint identification technology is anticipated. "The Present and Future of Artificial Intelligence" predicts this future as follows: "The task of experts analyzing collected fingerprints will soon disappear into the annals of history. (Omitted) Using the mysterious and beautiful fingerprints bestowed by nature on humans as a powerful weapon to solve cases, it will ensure the safety of the people and make criminals disappear."

This content was produced with the assistance of AI translation services.