TW201327411A - Life fingerprint authenticity verification technology - Google Patents

Abstract

A steady speckles life fingerprint recognizing technology is disclosed in the invention. A whole set of system includes (1) Laser light source; (2) Beam expander; (3) Image capture unit; (4) Recognizing unit. This invention utilizes one unit of expanding to expand the lighting area of the laser beam. Lighting with the extended laser light beam, then utilize special image capturing unit to capture the fingerprint laser speckle picture, or other surface pattern of the fingerprint. Capture the image continuously, then calculate the correlation between those images serially. Distinguish whether fingerprint is a living subject according to the serial correlations function period. Utilize Local Binary Pattern(LBP) and Support Vector Machine(SVM) artificial intelligence will come to distinguish whether it is the human fingerprint or not. Finally, the laser speckle picture is recognized by the method to discern the identity with a constructed individual identity.

Description

Living fingerprint identification technology

The invention relates to a living body fingerprint identification system, in particular to an object surface spot recognition technology using an image capturing method of a steady state laser spot dynamic image, which uses a laser light beam expanding technique to generate a large area of the illumination beam, and then uses the light beam. The bar-limited light technology produces a steady-state laser spot. After capturing the dynamic image of the steady-state spot, the image processing technology software technology is used to identify the fingerprint. The double-layer identification is used to initially identify the fingerprint and further identify the fingerprint to be recognized. For preliminary grouping, the laser surface spot feature is used to perform accurate identification. The process is shown in Figure 1. The fingerprint stable surface spot identification technology can improve the recognition rate of the fingerprint identification system and ensure that the person to be identified has a pulse. The living fingerprint; the two-stage image processing, the database search and comparison, can accelerate the recognition rate and improve the recognition rate.

Traditional fingerprint recognition uses a variety of methods (optical imaging, capacitance, etc.) to obtain a two-dimensional image of the fingerprint, and then recognizes the fingerprint characteristics of the image to confirm that the captured fingerprint is accepted or rejected. Although this method is widely used in various fields, its acquisition is a two-dimensional still image of the fingerprint, so the fingerprint identification system is easily blinded by the copied fingerprint, and has been criticized by everyone.

The invention relates to a living body fingerprint photoelectric identification system, which adopts a laser light beam expanding method to generate a large-area laser beam which is sufficient for applying to the surface of the illumination fingerprint, and generates a stable state laser spot and a dynamic image image by using the light bar limiting light technology. The photoelectric device of the method, together with the two-level identification process of the invention, can solve the problem that the fingerprint is easily copied in the past, and can improve the recognition rate of the system and shorten the comparison time.

One aspect of the present invention is to provide a fingerprint identification technique.

According to an embodiment, the fingerprint surface spot extraction device comprises (1) a laser light source; (2) a beam expander unit; (3) an image capture unit; and (4) an identification unit. "Laser light source" is the original light source of the illumination fingerprint; "expansion beam unit" is used to enlarge the laser beam area, the laser beam illumination area is enlarged enough to illuminate the fingerprint, and the illuminated fingerprint surface appears fingerprint pattern and laser specific The spot, the spot will change at any time as the pulse beats, and periodically change with the pulse; the "image taking unit" includes a light bar, an imaging lens and an imaging array component, and the general laser spot is even to be identified. Fixed, the spot will change randomly with time, and the "light bar" will form a stable laser spot. The imaging array component can be a CCD, CMOS camera or any two-dimensional imaging component; The unit includes an image signal acquisition component and an image data processing component. The image data processing component analyzes the dynamic image and the laser spot fingerprint related image after obtaining the steady state laser spot fingerprint dynamic image of the fingerprint. Continuously capture images and perform cross-correlation operations in sequence. The calculation results are periodic fluctuations related to the pulse, and the fingerprints are distinguished according to the period of the fluctuations. The local binary patterns (LBP) and support are used. Vector Machine (Support Vector Machine, SVM for short) image processing technology to distinguish whether the fingerprint of the human body, can confirm that the fingerprint is a live fingerprint with a pulse, and then identify the identity by identification method. The working method flow of the present invention is shown in Fig. 2: a fingerprint spot identification flowchart. The identification work is divided into two parts: database establishment and image recognition. After the database is established, the fingerprint spot image recognition technology can be used to identify the identity of the person. In the stage of establishing the fingerprint feature database, the fingerprint surface spot extraction device is used to capture the reference fingerprint surface spot image, then the surface fingerprint spot feature of the reference fingerprint is analyzed, the fingerprint group is determined according to the characteristics, and the fingerprint steady-state laser spot feature is established according to the group. database. In the fingerprint identification stage, the fingerprint surface spot extraction device is used to capture the dynamic surface spot image of the fingerprint to be recognized, and the continuous dynamic images are correlated with each other according to the time series, and it is determined whether the fingerprint person is a living fingerprint with a pulse, and then the region is The image processing technology of binary pattern (LBP) and support vector machine (SVM) classification learning determines whether the person providing the fingerprint is a human fingerprint. Then analyze the surface spot characteristics of the fingerprint to be identified, refer to the traditional fingerprint identification method to judge the fingerprint grouping, and then use the accurate image processing technology to accurately identify the fingerprint steady-state laser spot of the same group of smaller databases. Accurate and fast demand.

Compared with the conventional identification system, the present invention utilizes a steady state laser spot image, and in addition to having the function of a general fingerprint identification system, it can also distinguish a real person fingerprint or a fingerprint copy.

The advantages and spirit of the present invention will be further understood from the following detailed description of the invention.

For a schematic diagram of the identification imaging device, please refer to FIG. 1 : an embodiment of an optoelectronic device that recognizes one of the fingerprints and captures the surface spot of the fingerprint.

The laser source is a single-mode 3mW semiconductor laser with a wavelength of 633 nm. The beam expander unit first expands the beam with a negative lens, and then condenses the light into parallel light by a collecting lens to form an integrated illumination source module.

The imaging unit is close to the 4f architecture, using a positive focal length lens as the imaging mirror, and placing the light barrier in front of the lens. The CCD is the interface IEEE 1394. The distance between the components needs to be optimally adjusted to obtain a stable state of suitable scale. Light spot is the basis.

The optical axis of the illumination source module and the imaging unit should avoid exactly the specular reflection angle, so as to avoid the light intensity being too strong, and it is relatively impossible to obtain a better contrast spot image. The embodiment deviates from the reflection angle by about 10°.

The "identification unit" uses a personal computer to capture image data through the IEEE 1394 interface, and then uses a specific software to identify the surface spot image obtained.

Figure 3 shows the 24 fingerprint patterns captured in sequence, including the traditional fingerprint pattern and the spot image. The spot image will change due to the pulse beat. The continuous dynamic images are cross-correlated according to the time series, and the calculation result is as shown in FIG. 4, and the period of the cross-correlation timing change is used to distinguish whether the fingerprint to be recognized is a suitable living body.

The SVM method is used to copy fingerprints and human fingerprints of different materials, and an image processing technology for distinguishing human fingerprints from copying fingerprints is established through a learning process. The captured fingerprint image is processed by the LBP method (as shown in Figure 5).

Each sample is a p-dimensional vector, and the SVM image processing technology learning method is used to determine the appropriate number of fingerprint data. The data is divided into two groups: non-biological fingerprint and human fingerprint. The sixth picture shows the p-dimension. A simplified schematic diagram of the classification. Using ten people's one hundred fingerprints and ten fake fingerprints and one hundred arbitrary spots to establish a judgment mechanism, this mechanism is used to test another fifteen fingerprints and five fake fingerprints. In the limited sample test, it is possible to correctly determine whether it is a human fingerprint. No error has occurred.

The identification work of the present invention is divided into two parts: the establishment of the database and the image recognition. After the database is established, the fingerprint spot image recognition technology can be used to identify the identity of the person. The spot fingerprint image adopts the traditional fingerprint identification method for preliminary group processing, and then uses the accurate image processing technology to compare and identify a small number of databases in the same group, and at the same time achieve accurate and rapid requirements. Using two-stage image processing, the database is searched and compared in groups, which can speed up the recognition rate and improve the recognition rate. Compared with the conventional fingerprint identification technology, the present invention utilizes a steady state laser spot image to distinguish a real person fingerprint or a fingerprint copy.

The features and spirit of the present invention will be more apparent from the detailed description of the preferred embodiments. On the contrary, the intention is to cover various modifications and equivalents within the scope of the invention as claimed.

1. . . Steady state laser spot fingerprint identification system

12. . . Laser source

13. . . Expanding unit

14. . . Image capture unit

141. . . Imaging lens

142. . . Light bar

143. . . Imaging array component

15. . . Identification unit

2. . . finger

Figure 1: Schematic diagram of the identification device for identifying fingerprints.

Figure 2: Fingerprint identification flow chart.

Figure 3: Sequence of fingerprint spot images.

Figure 4: Dynamic image cross-correlation timing diagram

Figure 5: The fingerprint image is processed by the LBP method.

Figure 6: Schematic diagram of the simplified classification of the p-dimension.

Claims (8)

A fingerprint surface spot feature recognition technology includes: 1.1 extracting a fingerprint spot; 1.2 analyzing a fingerprint surface spot feature, and establishing a fingerprint surface spot feature database; 1.3 continuously extracting a fingerprint surface spot to be recognized, the spot will periodically follow a pulse beat Change to determine whether it is a living fingerprint; 1.4 Capture the surface spot of the fingerprint to be identified, and use the image processing technology of the classified learning to judge whether it is a human fingerprint; 1.5 Analyze the surface spot characteristics of the fingerprint to be recognized; 1.6 The surface spot characteristics and data of the fingerprint to be identified The reference fingerprint surface spot features in the library are compared, and the fingerprint to be recognized is confirmed as a specific reference fingerprint in the database, or the fingerprint to be recognized is not present in the established database. For example, the fingerprint surface spot features described in item 1.2 of the patent application are divided into two categories, one of which is a visible fingerprint or surface spot information under general light source illumination, and the other is a steady state laser spot information. If the characteristics of the fingerprint surface spot are analyzed as described in item 1.2 of the patent application, the reference fingerprint can be grouped and processed by using the visible fingerprint or surface spot information under the illumination of the general light source. As described in item 1.3 of the patent application, continuous dynamic images are cross-correlated according to time series, and the period of the cross-correlation time series is used to identify whether the fingerprint is a suitable living body. As described in item 1.4 of the patent application, the image processing technique of regional binary pattern and support vector machine classification learning is used to distinguish whether the fingerprint is provided by the human body. As described in claim 5, the regional binary pattern uses a video processing technique of the following equation to convert each sample into a p-dimensional vector. As described in item 1.5 of the patent application scope, the fingerprints or surface spot information under the illumination of the general light source on the surface of the object to be identified are analyzed, and the fingerprints to be identified are initially determined to be grouped in the associated database. If the surface spot feature of the fingerprint to be identified is analyzed as described in item 1.5 of the patent application scope, the obtained spot feature is compared with the fingerprint surface spot feature of the group in the database, and the specific fingerprint corresponding to the fingerprint to be recognized is confirmed, or the Identify fingerprints that are not in the database.