KR102840845B1 - Non-contact type fingerprint certification emthod including distinguable counterfeit fingerfrient using ai - Google Patents

Abstract

The present invention relates to a contactless fingerprint authentication method including a fake fingerprint detection function using AI, and more specifically, to a contactless fingerprint authentication method including a fake fingerprint detection function using AI, which enables the use of fake fingerprints, such as those printed on paper or reproduced on a screen, when recognizing a fingerprint in a contactless manner using a mobile device with a built-in camera, and learns such detection data using AI to quickly and accurately detect fake fingerprints.
According to a non-contact fingerprint authentication method including a fake fingerprint detection function using AI formed as a preferred embodiment of the present invention, fake fingerprints can be detected even in a non-contact fingerprint recognition method using a fingerprint image captured by a camera, fake fingerprints such as those obtained by secretly taking pictures of another person's fingerprints can be detected as fake fingerprints using a noise pattern of the captured fingerprint image, and fake fingerprints can be detected more quickly and accurately by storing and analyzing fake fingerprint data using AI.

Description

Non-contact type fingerprint authentication method including a counterfeit fingerprint detection function using AI {Non-contact type fingerprint certification method including a distinguishable counterfeit finger using AI}

The present invention relates to a contactless fingerprint authentication method including a fake fingerprint detection function using AI, and more specifically, to a contactless fingerprint authentication method including a fake fingerprint detection function using AI, which enables the use of fake fingerprints, such as those printed on paper or reproduced on a screen, when recognizing a fingerprint in a contactless manner using a mobile device with a built-in camera, and learns such detection data using AI to quickly and accurately detect fake fingerprints.

Biometrics refers to a user authentication technology that identifies a user based on human physical or behavioral characteristics. The domestic biometrics forum defines it as a user authentication technology based on the observation of behavioral and biological (anatomical and physiological) characteristics.

Methods for recognizing information about human physical characteristics used in the above biometric technology include fingerprint recognition, iris-scan, retina-scan, hand geometry, and facial recognition, and methods for recognizing information about behavioral characteristics include voice recognition, signature scan, and gait recognition.

Among the above biometric technologies, the most widely used and widespread is fingerprint recognition. Fingerprints formed on human fingers are all different, and not only are they different from one individual to another, but they also have the characteristic of not changing throughout life as they are at birth, so they have high reliability and stability in identification, and information storage and comparison are also easy and convenient.

A conventional fingerprint verification device has a glass surface (sensor) formed on the top, and a camera and light built into the inside. When the user touches the glass surface (sensor) and operates the fingerprint verification device, the camera captures the finger and acquires a fingerprint image.

In addition, the fingerprint recognition device is provided with a function to distinguish between a fake fingerprint made of silicone or the like and an actual human fingerprint (LiveFinger Detection, LFD), and a function to extract and store the characteristic points of the ridges of a fingerprint (optimal recognition fingerprint data generation function, Image Enhancement Feature Extraction Matching).

The above fingerprint recognition device is called an optical (or contact) fingerprint recognition device, and the fingerprint image captured by this optical fingerprint recognition device is adjusted in sharpness and color contrast to obtain a clean, flat fingerprint image.

In addition, the fingerprint image is used to extract the characteristic points (location and directionality) of each fingerprint ridge, such as the broken or split ridges, and to store them as recognition fingerprint data.

The above optimal recognition fingerprint data is stored in a method (hereinafter referred to as the “international standard”) specified in the international standard (ISO19794-2/ANSI378) to ensure compatibility between different products.

However, conventional optical fingerprint recognition devices have problems such as the user feeling uncomfortable because they have to directly touch the sensor with their finger, the shape of the fingerprint may be distorted depending on the strength of the pressure applied to the glass surface (sensor) when touching, the shape of the fingerprint may be distorted when the finger slips on the contact surface of the glass surface (sensor), and it is difficult to obtain a clear fingerprint image depending on the difference in ambient temperature and humidity or the condition of the user's skin (e.g., dryness or moisture of the skin).

To solve the problems of conventional optical fingerprint recognition devices, non-contact fingerprint recognition devices have been developed recently that acquire fingerprint images by taking pictures from a position slightly away from the camera, without the user directly touching the glass surface (sensor).

Technologies related to conventional non-contact fingerprint recognition devices and fingerprint recognition methods have been disclosed in Korean Intellectual Property Office Registered Patent Publications Nos. 0604267, 1274260, and 1596298.

However, since there are people who want to use fake fingerprints for authentication for various purposes, a function that can detect fake fingerprints is needed in the conventional fingerprint recognition and authentication method.

These fake fingerprints include fake fingerprints printed on paper (printed fake fingerprints), fake fingerprints displayed on a screen (screen fake fingerprints), fake fingerprints made by wearing a silicone-like material that imitates another person's fingerprint and then recognizing it (silicon fake fingerprints), and fake fingerprints made by taking a non-contact photograph of a person's fingers without their consent (secretly photographed fake fingerprints).

To prevent such fake fingerprints, the Korean Intellectual Property Office provides a technology for identifying fake fingerprints using neural network learning in Patent Publication No. 2017-0112302, Patent No. 874688, and Patent No. 1179559.

However, conventional fake fingerprint detection technology had the following problems.

(1) Various technologies for detecting fake fingerprints are provided for contact-type fingerprint recognition devices, but they are not provided for non-contact fingerprint recognition methods that use the camera of a mobile device with a built-in camera.

(2) It cannot address issues such as special fake fingerprints that may be misrecognized when taking non-contact photos or acquiring someone else's fingerprints through secret photography.

(3) There is no method available to quickly and accurately identify fake fingerprints by digitizing them and analyzing them.

In order to solve the above-mentioned problem, the present invention provides a non-contact fingerprint recognition method for obtaining a fingerprint by taking a picture of a subject's finger with a mobile device having a built-in camera.

A fingerprint recognition program is formed in a mobile device having the above camera built in, and a fake fingerprint detection program is formed in the fingerprint recognition program.

The above fake fingerprint detection program uses a noise pattern method to detect fake fingerprints by recognizing noise patterns;

Histogram method for determining fake fingerprints through RGB histogram;

Simultaneously adopting the contour method that recognizes the outline of the finger and determines whether it is a fake fingerprint,

The discovered fake fingerprints are stored as fake fingerprint data, including the noise pattern, RGB histogram pattern, and outline errors.

It is characterized by deep learning of the above fake fingerprint data through AI.

According to a non-contact fingerprint authentication method including a fake fingerprint detection function using AI formed as a preferred embodiment of the present invention, the following effects occur.

(1) Fake fingerprints can also be identified using a non-contact fingerprint recognition method that uses a fingerprint image captured by a camera.

(2) Forged fingerprints, such as those obtained by taking pictures of other people's fingerprints, can be identified as forged fingerprints by utilizing the noise pattern of the captured fingerprint image.

(3) Using AI, fake fingerprint data can be stored and analyzed to identify fake fingerprints more quickly and accurately.

Figure 1 is a photo showing an example of a fake fingerprint.
Figure 2 is a flowchart showing a non-contact fingerprint authentication method including a fake fingerprint detection function using AI formed as a preferred embodiment of the present invention.
FIG. 3 is a photograph showing an example of a histogram method in a non-contact fingerprint authentication method including a fake fingerprint detection function using AI formed as a preferred embodiment of the present invention.

Before describing specific embodiments of the present invention, the drawings illustrated in this specification may be somewhat exaggerated or simplified in size or shape of components in order to more clearly explain the present invention.

Since the terms and symbols defined in the present invention are terms arbitrarily defined or selectively used by users, operators, and authors, these terms should be interpreted as meanings and concepts that conform to the technical idea of the present invention based on the entire contents of this specification, and should not be limited to the meaning of the terms themselves.

The present invention relates to a non-contact fingerprint recognition method for obtaining a fingerprint by taking a picture of a subject's finger using a mobile device having a built-in camera.

A fingerprint recognition program is formed in a mobile device having the above camera built in, and a fake fingerprint detection program is formed in the fingerprint recognition program.

The above fake fingerprint detection program uses a noise pattern method to detect fake fingerprints by recognizing noise patterns;

Histogram method for determining fake fingerprints through RGB histogram;

Simultaneously adopting the contour method that recognizes the outline of the finger and determines whether it is a fake fingerprint,

The discovered fake fingerprints are stored as fake fingerprint data, including the noise pattern, RGB histogram pattern, and outline errors.

The above fake fingerprint data is configured to be able to quickly and accurately identify fake fingerprints by performing deep learning on the AI.

The non-contact fingerprint recognition method of the present invention is employed in a mobile device having a built-in camera, and a fake fingerprint determination program is formed within the fingerprint recognition program while a fingerprint recognition program is installed in the mobile device having a built-in camera.

The above fake fingerprint detection program applies noise pattern method, histogram method, and outline method simultaneously.

The above noise pattern method is a method that utilizes the noise pattern that occurs when a camera photographs a finger. This method utilizes the fact that the quality of finger images photographed from a distance is poor because more noise occurs in the image.

That is, when taking a picture of a normal fingerprint, strong lighting is applied to capture it clearly, but when taking a picture from a distance, a noise pattern is created and the image quality deteriorates. If such a noise pattern occurs, it can be determined as a fake fingerprint.

In addition to the above noise pattern method, the histogram method is used to determine fake fingerprints. When taking a picture of a normal fingerprint with a mobile device equipped with a built-in camera, it is taken with strong lighting, so it is not only taken very clearly, but also taken in color, so the RGB histogram can be confirmed in RGB colors, so the fake fingerprint can be determined from the difference in the RGB histograms of the normal fingerprint and the fake fingerprint.

That is, when a normal fingerprint is photographed, the RGB histogram is clearly revealed, whereas for a fake fingerprint that is printed, displayed on a screen, or printed on paper, the RGB histogram is not clear and the colors do not appear well as they overlap.

Figure 3 compares the RGB histograms of a normal fingerprint and a fake fingerprint. The fingerprint formed on the left is a normal fingerprint, and the fingerprint formed on the right is a fake fingerprint (a photograph of a printed fingerprint).

In Figure 3, the RGB histogram on the left has sharp peaks and colors appear clearly, while the RGB histogram on the right has blurred peaks and not all colors appear.

In the above histogram method, it is possible to sufficiently determine the ridge by selecting 2 to 3 ridge areas and analyzing the RGB histogram of the ridges.

Also, when a normal fingerprint is photographed, the ridge area has a higher light reflectance than the valley area between the ridges, and because of the sense of perspective, the center of the fingerprint area has a higher light reflectance than the outer area of the fingerprint area, so the difference in illumination can be confirmed. However, this difference in illumination does not appear in a fake fingerprint, so it can be determined.

The present invention can sufficiently identify fake fingerprints using only the noise pattern method and the histogram method, but the outline method can also be employed to enable more accurate identification.

The above outline method is a method of determining whether a photographed fingerprint image is a fake fingerprint by checking for newly created or shaded areas that deviate from the expected normal fingerprint ridges and outline of the finger.

When a fake fingerprint is identified by a fake fingerprint identification program as described above, the fake fingerprint data is saved as fake fingerprint data, including the noise pattern level, RGB histogram distribution shape, and contour error data of a unique shape.

The above fake fingerprint data is classified into printed fake fingerprints, screen fake fingerprints, silicon fake fingerprints, and burglarized fake fingerprints.

The above stored fake fingerprint data is formed to quickly and accurately identify fake fingerprints by performing deep learning using AI.

The artificial intelligence neural network of the AI that analyzes the above-mentioned fake fingerprint data includes a state value transmission step in which the fingerprint state value (s) is transmitted to a deep reinforcement learning neural network, a discriminant prediction neural network, and a discriminant neural network;

The above discriminant prediction neural network is composed of a normal fingerprint prediction neural network and a fake fingerprint prediction neural network, and the normal fingerprint prediction neural network and the fake fingerprint prediction neural network have an additional discriminant value transmission step of determining and transmitting an additional discriminant value (γ) after learning about the fingerprint status value (s);

The above deep reinforcement learning neural network is composed of an Actor neural network and a Critic neural network, and the Critic neural network calculates a value value (Q) through a state-value function along with the transmitted fingerprint state value (s) and the expected discriminant value (a) and an additional discriminant value (γ) and transmits the value value to the Actor neural network, and the Actor neural network calculates an optimal discriminant value (A) by utilizing the value value (Q) and the state value (s), in an optimal discriminant value calculation step;

The above discriminant neural network is composed of a fingerprint discrimination value calculation step that stores the discrimination value (a) as learning data, calculates the optimal fingerprint discrimination value (K) using a machine learning technique, and transmits it as a discrimination means of a fake fingerprint discrimination program.

The above artificial intelligence neural network uses a supervised learning algorithm and operates in an optimized state using a training set (learning data).

The above supervised learning algorithm operates in a state where the basic algorithm is optimized to find the optimal fingerprint discrimination value (K) by using fingerprint status values of many types of fake fingerprint data.

An example of the discriminant function used in the above discriminant prediction neural network is

It can be used to learn and output the discriminant value (γ) as the expected value when the state value (s) and discriminant value (a) are at a specific location (t).

The above discriminant prediction neural network is composed of a normal fingerprint prediction neural network and a fake fingerprint prediction neural network. After the normal fingerprint prediction neural network and the fake fingerprint prediction neural network learn about the fingerprint status value (s), the average of the evaluated discriminant values is determined as the final additional discriminant value (γ) and transmitted to the deep reinforcement learning neural network.

The above deep reinforcement learning neural network is composed of an Actor neural network and a Critic neural network. The Critic neural network learns and produces a value value (Q) by utilizing the state value (s) and the discriminant value (γ).

An example of an equation used for the above value (Q) is

The Bellman expectation equation can be used.

The above Actor neural network learns and outputs the optimal discriminant value (A) using the state value (s) and value value (Q).

The optimal decision value (A) generated from the above Actor neural network is transmitted to the Critic neural network and stored as a new previous state value.

By updating the discrimination value using deep learning using AI as described above, it is possible to accurately determine fake fingerprints predicted from captured fingerprint images and learn from this to quickly and accurately determine fake fingerprints in the future.

According to a non-contact fingerprint authentication method including a fake fingerprint detection function using AI formed as a preferred embodiment of the present invention, fake fingerprints can be detected even in a non-contact fingerprint recognition method using a fingerprint image captured by a camera, fake fingerprints such as those obtained by secretly taking pictures of another person's fingerprints can be detected as fake fingerprints using a noise pattern of the captured fingerprint image, and fake fingerprints can be detected more quickly and accurately by storing and analyzing fake fingerprint data using AI.

Although the present invention has been described with reference to the attached drawings and focusing on preferred embodiments, it will be apparent to those skilled in the art that various modifications are possible without departing from the scope of the present invention encompassed by the claims set forth below.

Claims (2)

In a general non-contact fingerprint recognition method, a fingerprint is acquired by taking a picture of the subject's finger with a mobile device equipped with a camera and learning it with an artificial intelligence neural network,
The above artificial intelligence neural network includes a state value transmission step in which a fingerprint state value (s) is transmitted to a deep reinforcement learning neural network, a discriminant prediction neural network, and a discriminant neural network to analyze fake fingerprint data;
The above discriminant prediction neural network is composed of a normal fingerprint prediction neural network and a fake fingerprint prediction neural network, and the normal fingerprint prediction neural network and the fake fingerprint prediction neural network have an additional discriminant value transmission step of determining and transmitting an additional discriminant value (γ) after learning about the fingerprint status value (s);
The above deep reinforcement learning neural network is composed of an Actor neural network and a Critic neural network, and the Critic neural network calculates a value value (Q) through a state-value function along with the transmitted fingerprint state value (s) and the expected discriminant value (a) and an additional discriminant value (γ) and transmits the value value to the Actor neural network, and the Actor neural network calculates an optimal discriminant value (A) by utilizing the value value (Q) and the state value (s), in an optimal discriminant value calculation step;
A contactless fingerprint authentication method including a fake fingerprint detection function using AI, characterized in that the above-mentioned discriminant neural network is configured with a fingerprint detection value calculation step that stores the discrimination value (a) as learning data, calculates the optimal fingerprint discrimination value (K) using a machine learning technique, and transmits it to the discrimination means of a fake fingerprint detection program. In the first paragraph,
A contactless fingerprint authentication method including a fake fingerprint identification function using AI, characterized in that the above fake fingerprint data is classified into printed fake fingerprints, screen fake fingerprints, silicon fake fingerprints, and surreptitious fake fingerprints.