Fingerprints have always been one of the most reliable ways to identify a person. Nowadays, with services like Aadhar, your fingerprint can hold an entire database of information, making a "profile" that can be used for verification. Traditionally, fingerprints are captured by pressing your finger against a surface, like when you unlock your phone or when using biometric scanners at work. However, new research is exploring a touch-free way to capture fingerprints aiming to make this process cleaner, easier, and more accurate. A collaboration between researchers from Sardar Patel Institute of Technology, Indian Institute of Technology Bombay, and Indian Institute of Technology Indore is looking to use a photograph of your finger to capture fingerprints.
A fingerprint is composed of unique patterns of ridges and valleys on the skin of your fingers. Features called "minutiae," where ridges end, or split, are unique to an individual and play a crucial role in distinguishing one fingerprint from another. In this study, the scientists devised a method that begins with capturing a picture of your finger instead of pressing it against a scanner. They then use a range of image processing techniques to highlight and enhance the intricate details of your fingerprint in the photo.
A method called Adaptive Thresholding helps adjust the brightness levels to make the fingerprint patterns more distinct. Next, a "Gabor Filter," which sharpens the textures, emphasizing those all-important ridges and valleys, is applied. Once the minutiae are clearly visible, they use something called a K-means clustering algorithm to remove the background from the image, much like focusing a camera lens to blur out everything except the main subject – your fingerprint. Once the image is processed, it undergoes a thinning process to make the fingerprint one pixel wide. This skeletonized version of the fingerprint is then used to extract minutiae.
Additionally, the researchers used an innovative combination of machine learning, specifically a kind of artificial intelligence called a "Siamese network," with traditional techniques. The Siamese network aids in learning patterns by comparing more than one fingerprint image and recognizing similar features, making the system very accurate. Using this combination method, the system achieves impressive accuracy, with the error rate dropping as low as 2.5% or 3.76% depending on the datasets used for testing.
While the current research is promising, there are still areas that need further exploration. Researchers are interested in testing different types of wavelets (a mathematical function used to analyze the details at different frequencies) that could potentially enhance the process and results. Additionally, they aim to develop even more advanced techniques for fusing all the gathered information, making the fingerprint identification process even more robust.
This touchless method has some immediate perks. For one, it’s more hygienic since you don’t have to physically touch a scanner. Imagine not having to worry about transferring germs and viruses, as experienced during the COVID-19 pandemic. Moreover, it overcomes problems like sensor wear and tear, which can happen very quickly when a device gets pressed repeatedly. In the future, these advancements could even allow for systems that recognize other biometric data like palm prints or face characteristics, integrating them into a singular system for even more secure identification.
This research news was partly generated using artificial intelligence and edited by an editor at Research Matters.