I originally wanted to write it myself, but there is too much to say, so I might as well find a COPY. If you still don’t understand anything, you can ask me directly. I work in this industry.
The fingerprints on the fingers represent a person’s identity. In 1788, Mayer first proposed that no two people's fingerprints are exactly the same. In 1823, Purkinie divided fingerprint patterns into nine categories for the first time. In 1889, Henry proposed the fingerprint detail feature identification theory, laying the foundation for modern fingerprint science. However, the manual comparison method is inefficient and slow. In the 1960s, computer image processing and pattern recognition methods began to be used for fingerprint analysis. This is the Automatic Fingerprint Identification System (AFIS for short) [1]. In the late 1970s and early 1980s, the automatic fingerprint identification system (police?AFIS, P?AFIS) for criminal investigation was put into practical use. In the 1990s, AFIS entered civilian use and was called the Civilian Automatic Fingerprint Identification System (civil? AFIS, C?AFIS). This article attempts to start from the analysis of fingerprint characteristics and explain the principles and methods of fingerprints as human body identification, the main technical indicators and testing methods of fingerprint identification, as well as the reality and reliability of practical applications [2-4].
1 Principles and methods of fingerprint identification
1.1 Characteristics and classification of fingerprints
Fingerprint identification is an ancient subject, which is based on human fingerprint characteristics The relative stability and uniqueness of this statistical result were developed. In practical applications, according to different needs, the fingerprint characteristics of the human body can be divided into: permanent characteristics, non-permanent characteristics and vital characteristics [5].
Permanent features include detailed features (center point, triangle point, end point, cross point, bridge point, etc.) and auxiliary features (grain shape, grain density, grain curvature and other elements). It never changes and is most obvious and most evenly distributed in the typical area at the front of the finger [1]. Detailed features are the basis for accurate fingerprint comparison, while pattern features, texture features, etc. are important basis for fingerprint classification and retrieval. The pattern characteristics of human fingerprints can usually be divided into three types according to their different shapes: "bow, skip, and bucket", as well as "orphan, tent, regular skip, reverse skip, ring, spiral, and capsule." There are 9 forms including ", double-skirt shape and hybrid shape" [1]. Texture features are composed of texture parameters such as average grain density, grain density distribution, average grain curvature, and grain curvature distribution. Texture features are mostly used in multi-dimensional classification and retrieval of computer fingerprint recognition algorithms.
Non-permanent features are fingerprint features composed of isolated points, short lines, wrinkles, scars and resulting breakpoints, cross points and other elements. Such fingerprints may appear, heal, develop or even disappear[ 1].
The vital characteristics of fingerprints are closely related to the existence or absence of life of the tested object. However, its relationship and laws with human life phenomena still need to be further understood. At present, it has become one of the hot spots that attracts more and more attention in modern civilian fingerprint identification applications.
1.2 Principles and methods of fingerprint identification
Fingerprint identification technology mainly involves four functions: reading fingerprint images, extracting features, saving data and comparison. The image of the human fingerprint is read by the fingerprint reading device, and then the original image is preliminarily processed to make it clearer, and then the fingerprint characteristic data is established through the fingerprint identification software. The software finds data points called "nodes" (minutiae) from the fingerprint, which are the coordinate positions of the bifurcations, terminations or loops of the fingerprint lines. These points have more than seven unique characteristics at the same time. Typically there are an average of 70 nodes on a finger, so this method will produce about 490 data. These data are usually called templates. Through the computer fuzzy comparison method, the templates of two fingerprints are compared, their similarity is calculated, and the matching result of the two fingerprints is finally obtained [5-6]. Acquisition devices (i.e., imaging devices) are divided into several categories: optical, semiconductor sensors, and others.
2 Main indicators and test methods of fingerprint recognition technology
2.1 Accuracy of the algorithm
The performance indicators of the fingerprint recognition system depend to a large extent on the Algorithm performance. In order to facilitate the use of quantitative methods to express its performance, the following two indicators are introduced.
False rejection rate (FRR): refers to the error probability of rejecting the same fingerprint as being different. FRR = (number of fingerprints rejected/total number of fingerprints examined) × 100%.
False acceptance rate (FAR): refers to the error probability of receiving different fingerprints by mistaking them for the same fingerprint. FAR = (number of incorrectly identified fingerprints/total number of fingerprints examined) × 100%.
For an existing system, by setting different system thresholds, it can be seen that the two indicators are related to each other, and FRR is inversely proportional to FAR. This is easy to understand. The stricter the "checking", the lower the possibility of misrecognition, but the higher the possibility of rejection.
2.2 Testing method of false recognition rate and rejection rate
To test these two indicators, the loop test method is usually used [7]. That is, a set of images is given, then combined in pairs, submitted for comparison, the total number of submissions for comparison and the number of errors occurred are counted, and the proportion of errors is calculated, which is FRR and FAR. For the indicator of FAR=0.0001%, no less than 1,415 different fingerprint images should be used for cyclic testing, and the total number of tests is 1,000,405 times. If an error occurs during the test and the comparison is successful, then FAR=1/1,000 405; For FRR=0.1%, no less than 46 image combinations belonging to the same fingerprint should be used for testing. The total number of test submissions is 1035 times. If an incorrect rejection occurs, FRR=1/1035. The larger the number of samples used in the test, the more accurate the results will be. The fingerprint image used as a test sample should meet the conditions for registration.
2.3 System parameters
Error registration rate (ERR): refers to the probability that the fingerprint device will have fingerprints that cannot be logged in and processed. If ERR is too high, it will seriously affect The usage range of the equipment is usually required to be less than 1%.
Login time: The time required for a fingerprint device to log in a fingerprint. Usually the time required for a single login is no more than 2 seconds.
Comparison time: The time it takes for a fingerprint device to compare two sets of fingerprint feature templates is usually required to be no more than 1 s.
Operating temperature: The temperature change range allowed when the fingerprint device is working normally, generally 0 ~ 40 ℃.
Operating humidity: The relative humidity change range allowed when the fingerprint device is working normally is generally 30% to 95%.
3 Applications of Fingerprint Identification Technology
Fingerprint identification technology has matured and its applications are becoming increasingly common. In addition to criminal investigation, it is also widely used in civilian applications, such as fingerprint access control systems, Fingerprint attendance system, bank fingerprint savings system, bank fingerprint safe deposit box, fingerprint medical insurance system, family planning fingerprint management system, child pick-up and drop-off fingerprint management system, fingerprint blood donation management system, securities trading fingerprint system, fingerprint firearms management system, smart building fingerprint access control management system, driver fingerprint management system, etc.
Fingerprint access control system and fingerprint attendance system are the earliest access management systems developed and used, including intercom fingerprint access control, online fingerprint access control, offline fingerprint access control, etc. Press an individual's finger on the fingerprint collector at the entrance. The system will compare the fingerprints registered in the fingerprint database (called registered). If the two match (i.e. match), the comparison will be successful and the door will automatically open. Open. If there is no match, "Unsuccessful" or "No such fingerprint" will be displayed, and the door will not open. In the fingerprint access control system, it can be one-to-one matching (one?to?one matching) or one-to-several matching (one?to?few matching). The former can be a company or department, and the latter can be a member of a family, a bank business hall, a treasury, a financial department, a warehouse and other confidential places. In these applications, fingerprint recognition systems will replace or supplement many mass-use photo and ID systems.
Combining fingerprint recognition technology with IC cards is one of the most promising applications at present.
This technology stores the fingerprint of the card owner (encrypted) on the IC card, and installs a fingerprint recognition system on the IC card reader. When the card reader reads the information on the card, it also reads the card holder. The fingerprint can be compared to confirm whether the cardholder is the real owner of the card, so as to proceed with the next transaction. Fingerprint IC cards can replace current ATM cards and produce anti-counterfeiting documents. ATM card holders do not need a password to avoid the difficulty of the elderly and children remembering passwords.
In recent years, the Internet has brought convenience and benefits to people, but it also has security issues. Fingerprint feature data can be transmitted and verified on a computer network through email or other transmission methods. Through fingerprint identification technology, only designated people can access relevant information, which can greatly improve the security of online information. A series of online business activities such as online banking, online trade, and e-commerce are guaranteed to be safe.
The application of the fingerprint social insurance system has played a very effective role in the accurate payment of pensions. It avoids the problem that someone else uses a stamp or a copy of an ID card to collect the pension on your behalf, and the disbursing officer cannot be sure that the person is deceased. Only his or her living fingerprints can be used to accurately distribute the pension.
4 Reliability of fingerprint identification
Fingerprint identification technology is a mature biometric technology. Because each person's skin texture, including fingerprints, is unique in patterns, breakpoints and intersections, and remains unchanged throughout life. By comparing his fingerprints with pre-saved fingerprints, his true identity can be verified. Automatic fingerprint recognition is a method that uses computers to perform fingerprint recognition. It benefits from modern electronic integrated manufacturing technology and fast and reliable algorithm theoretical research. Although fingerprints are only a small part of human skin, the amount of data used for identification is quite large, and comparing these data requires fuzzy matching algorithms that require a large amount of operations. Small fingerprint image reading devices and faster computers produced using modern electronic integrated manufacturing technology provide the possibility of performing fingerprint comparison operations on microcomputers. In addition, the reliability of the matching algorithm is also continuously improved. Therefore, fingerprint recognition technology has become very simple and practical. Since the computer only involves some limited information when processing fingerprints, and the comparison algorithm is not a very precise match, the results cannot be guaranteed to be 100% accurate.
An important measurement indicator for a specific application of a fingerprint identification system is the recognition rate. It mainly includes the rejection rate and the misrecognition rate, which are inversely proportional to each other. Adjust these two values ??according to different uses. Although the fingerprint recognition system has reliability issues, its security is much higher than the "user ID + password" solution with the same reliability level. The rejection rate is actually an important indicator of system ease of use. In the design of application systems, ease of use and security must be weighed. Usually, two or more fingerprints are compared to greatly improve the security of the system without losing the ease of use.