TWI457841B - Identity recognition system and method - Google Patents

Description

Identity identification system and method

The present invention relates to an identity recognition system and method, and more particularly to an identity recognition system and method that focuses on the identification of palm vein texture.

Biometrics plays an increasingly important role in the current society. From the cash machine, access control system, notebook computer to the pen drive, you can see the application of biometric technology.

In the field of biometrics, palm vein identification is an emerging research focus. The palm vein has more information than fingerprints or palm prints, and it can get a good recognition rate. Its recognition rate is similar to that of the biometric identification system using iris technology, and its advantage of being unable to be counterfeited has gradually become the focus of scientists. Especially in terms of growth rate, it can be expected that the palm vein identification technology will gain market share with other biometric technologies. It is a pity that the current research on palm vein related is very rare.

One aspect of the present invention is to provide an identity recognition system.

According to an embodiment of the present invention, an identity recognition system includes a database, an image capture module, a determination module, a synergy module, a conversion module, and a comparison module. The database is used to store the data of the palm of the plurality of pens. The image capturing module is used to capture an image, and the determining module is used to determine the image. If the determining module determines that the image is a palm image, the synergistic module will perform a synergistic processing on the image. The conversion module is configured to convert the image into a plurality of feature points of constant scale. The comparison module is configured to compare the plurality of feature points with the same scale and the palm data in the database to identify the identity of the image provider.

Another aspect of the present invention is to provide an identification method. According to an embodiment of the invention, the identity identification method comprises the following steps:

(a) storing the palm of the hand in a database;

(b) taking an image;

(c) determining whether the image is a palm image, and if so, proceeding to step (d);

(d) performing a synergistic treatment on the image;

(e) converting the image into a plurality of feature points of constant scale;

(f) comparing the plurality of scale-invariant feature points with the palm data in the database to identify the identity of the image provider.

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

One aspect of the present invention is to provide an identity recognition system. Referring to FIG. 1, a functional block diagram of an identity recognition system 1 in accordance with an embodiment of the present invention is shown.

As shown in FIG. 1 , the identity recognition system 1 includes a database 15 , an image capture module 10 , a determination module 11 , a synergistic module 12 , a conversion module 13 , and a comparison module 14 . The image capturing module 10 is coupled to the judging module 11 , and the judging module 11 is coupled to the jacking module 12 , and the jacking module 12 is coupled to the converting module 13 . The library is coupled to the comparison module 14, and the database 15 is coupled to the comparison module 14.

The database 15 can be used to store a plurality of palm data, and in order to effectively compare the stored data, in one embodiment of the present invention, the palm data can establish a k-d tree data structure.

In an embodiment of the present invention, the image capturing module 10 can include a near infrared camera 100, a filter 102, and an image capture card 104. In general, light having a wavelength between 700 nm and 1000 nm is infrared light, and the light in this band is easily absorbed by the non-oxygenated red blood cells (ie, venous red blood cells) when irradiated to the human body, thereby forming black lines in the image. Therefore, the image capturing module 10 uses a near-infrared light source to illuminate the palm of the user, and the camera 100 captures the image reflected from the palm of the hand to obtain information of the venous vein, and uses the vein venus as the identification feature, wherein The captured image can be captured by the image capture card 104. In a preferred embodiment of the present invention, after filtering through the filter 102, the near-infrared source has a wavelength of about 850 nm.

After the image capturing module 10 captures the image, the determining module 11 determines whether the type of the image is a palm image. If yes, the determining module 11 transmits the image to the effecting module 12 for processing.

It was found in the experiment that the palm image is brighter than the general ambient light due to the reflection. Therefore, in a specific embodiment, the determining module 11 frames a rectangular portion in the image, and then divides the plurality of scanning lines in the vertical and horizontal directions in the rectangular portion. Next, the determining module 11 calculates a point at which each scanning line passes, and if the grayscale value is greater than a preset threshold, it is accumulated, and after a certain amount of accumulation, the image is determined to be a palm image.

If the judging module 11 determines that the image is a palm image, the synergy module 12 can perform the effect processing on the image, which is the image pre-processing. basically. Image pre-processing has a significant impact on image recognition rates. Technically, image pre-processing can use image normalization and histogram equalization methods to increase the contrast of the image, so that the image contains more information for identification. In a preferred embodiment, the present invention employs a histogram equalization method.

It should be noted that after the image pre-processing is completed, the conversion module 13 performs a Scale Invariant Feature Transform (SIFT) program, and converts the image into a plurality of scales with feature descriptions. Feature points.

The scale invariant feature conversion procedure is divided into the following four steps:

1. Detect extreme values in the scale space;

2. Screening feature points;

3. Determine the direction of the feature points;

4. Construct a description vector of feature points.

The previous steps turn the image into a collection of feature points with 128 dimensions. Then, the comparison module 14 can perform similarity calculation, that is, the plurality of scale-invariant feature points and the description vectors carried by them are compared with the palm data in the database 15 to identify the image provider. The identity. In order to effectively compare these feature points, the palm data stored in the database 15 can establish a k-d tree data structure and speed up the search with the BBF algorithm.

The following is an example to illustrate how to perform similarity calculations. The matching module 14 compares the image with the palm data in the database 15 one by one, and finds an individual matching feature point set. It is assumed that after the feature point alignment is completed, a set of points matching the feature points in the two images as shown in FIG. 2A and FIG. 2B can be obtained. In FIG. 2A, the feature point set matched by the image conversion is shown, and FIG. 2B shows the feature point set matching one of the palm data in the database 15, wherein A1 matches A2, B1 matches B2, and C1 matches. C2, D1 matches D2.

In order to calculate the similarity, the present invention calculates the two-point concentration distance similarity as the identification basis, and the method of the method is taken as a reference (usually taking the first feature point; for example, taking A1 as the reference point in FIG. 2A, Figure 2B takes A2 as the reference point) and measures the distance between this point and other points on the graph.

Suppose that there are n matching points in the two graphs, so that the distance between the reference point A1 in Fig. 2A and other matching feature points (ie, B1, C1, D1) is xi, and the reference point A2 in Fig. 2B and other matching feature points ( That is, the distance between B2, C2, and D2) is yi, and the distance similarity of the two points set can be expressed by the following formula:

Where d represents the distance similarity and n represents the number of matching feature points and is greater than one.

According to the distance similarity, the present invention can match in each image and find the most similar image. However, if the two images match too little, it will cause the difference between the similarity and the actual result to be too large. For example, if only two matching points exist after comparing two images, the distance similarity at this time does not reflect the actual similarity. Therefore, in a preferred embodiment of the present invention, after finding the matching point, the image with the matching point less than 5 points is ignored to improve the matching speed. At this time, for the above formula, the number of matching feature points is n≧5.

Table 1 and Table 2 below show exemplary experimental data of one of the identity recognition systems of the present invention.

In the biometric system, the biometric method is generally evaluated by the error acceptance rate and the false rejection rate. The false acceptance rate (FAR) is the ratio of non-legitimate users successfully accepted by the system; the false rejection rate (FRR) is the rate at which legitimate users cannot be successfully accepted by the system.

A total of 80 palm images were taken from this experimental example, and four palms were taken from each palm for a total of 320 images. For the FRR assessment, we took an image of each of the 80 people and used the other three images as a database for a total of 320 comparisons. For FAR, we excluded one person's image at a time, and took one of the images for identification, a total of 320 times. The identification result is controlled by the similarity threshold value. If the comparison result is greater than the threshold value, the identification is successful. If the comparison result is lower than the threshold value, it is regarded as one identification failure. It is known from the above two tables that when the threshold value is 85, the FAR is 0 and the FRR is 1.86%, which is the best.

Another aspect of the present invention is to provide an identification method. FIG. 3 is a flow chart of a method for identifying an identity according to an embodiment of the present invention, which includes the following steps. Please refer to Figure 1 to Figure 2B and related descriptions to fully understand the concept of this method.

Step S10 is executed to store the palm data of the plurality of pens in a database. Then, step S12 is performed to capture an image. In one embodiment of the invention, step S12 uses a near-infrared light source to illuminate the palm and then captures the image reflected from the palm. In addition, step S12 can use an image capture card to capture the captured image.

Next, step S14 is performed to determine whether the image is a palm image. In a specific embodiment of the present invention, step S14 forms a rectangular portion in the image, and then divides a plurality of scanning lines in the vertical and horizontal directions in the rectangular portion, and then calculates points passing through the scanning lines. If the grayscale value is greater than a preset threshold, it is accumulated, and after a certain amount of accumulation, the image is judged to be a palm image.

If the determination in the step S14 is NO, the identity recognition method is ended; if yes, the step S16 is executed to perform a synergistic process on the image. In a preferred embodiment of the invention, the synergistic process may use a histogram equalization method to increase the contrast of the image.

Next, step S18 is executed to convert the image into a plurality of feature points whose dimensions are unchanged. The steps of the scale invariant feature conversion procedure have been previously mentioned and will not be repeated here.

Then, step S20 is performed to compare the plurality of feature points whose dimensions are invariant with the palm data in the database to identify the identity of the image provider. Please refer to the previously introduced formula and related description for details of the comparison.

In summary, the present invention focuses on the identification of the palm vein texture, and uses the scale invariant feature transformation to convert the captured image into feature points, and then uses these feature points to calculate the similarity. It should be specially stated that the feature points obtained by the scale invariant feature transformation are quite resistant to scale change and rotation, and can also resist some image illumination changes and noise interference, so that the identification system and method can have Quite a good recognition rate.

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. . . Identity identification system

10. . . Image capture module

11. . . Judging module

12. . . Efficiency module

13. . . Conversion module

14. . . Alignment module

15. . . database

100. . . Near infrared camera

102. . . Filter

104. . . Image capture card

A1, B1, C1, D1, A2, B2, C2, D2. . . Matching feature points

Xi, yi. . . distance

S10~S20. . . Process step

1 is a functional block diagram of an identity recognition system in accordance with an embodiment of the present invention.

FIG. 2A and FIG. 2B are schematic diagrams showing the concentrated distance between the reference point and other matching feature points in the matching feature points.

FIG. 3 is a flow chart showing a method for identifying an identity according to an embodiment of the present invention.

S10~S20. . . Process step

Claims (18)

An identity recognition system includes: a database for storing a plurality of palm data; an image capture module for capturing an image; and a determination module for determining the image The judging module frames a rectangular portion in the image, and then divides a plurality of scanning lines in the vertical and horizontal directions in the rectangular portion, and then calculates a point where each scanning line passes, if the grayscale value is greater than A preset threshold is accumulated, and after a certain amount of accumulation, the image is determined to be a palm image; and a synergistic module, if the determining module determines that the image is a palm image, the synergy The module performs a synergistic processing on the image; a conversion module is configured to convert the image into a plurality of feature points having constant dimensions; and a comparison module is configured to use the plurality of modules The feature points with the same scale are compared with the palm data in the database to identify the identity of the image provider. For example, in the identity identification system described in claim 1, wherein the conversion module first detects the extreme value in a scale space, then performs screening of the feature points, and then determines the direction of the feature points, and then constructs the feature points. Description vector. The identity recognition system of claim 1, wherein the image capturing module is a camera including a near-infrared light source for illuminating a palm, and the camera is photographed from the camera. The image reflected by the palm of the hand. The identity recognition system of claim 3, wherein the near-infrared source has a wavelength of about 850 nm. The identity recognition system described in claim 3, wherein the image capture mode The group also includes an image capture card for capturing the image being captured. The identity recognition system of claim 1, wherein the synergistic processing performed by the synergistic module is used to increase the contrast of the image. The identity recognition system of claim 6, wherein the image enhancement processing uses a histogram equalization method. The identity identification system of claim 1, wherein the comparison module compares the image with the palm data in the database one by one, and finds an individual matching feature point set, and then the comparison The module will recognize the image according to the following formula of distance similarity: Where d represents the distance similarity, n represents the number of matching feature points and the system is greater than 1, x _i represents the distance between the reference point of the matching feature point in the image and other matching feature points, and y _i represents a match in one of the palm data. The distance between the reference point of the feature point and other matching feature points. For example, the identity identification system described in claim 8 of the patent scope, wherein n≧5. An identification method includes the following steps: (a) storing a plurality of palm data in a database; (b) capturing an image; (c) determining whether the image is a palm image, and if so, performing the step (d) The step (c) includes the sub-steps of: framing a rectangular portion in the image, and then dividing the scanning lines in the vertical and horizontal directions in the rectangular portion, and then calculating the scanning lines. The point is accumulated if the grayscale value is greater than a predetermined threshold, and after a certain amount of accumulation, the image is determined to be a palm image; (d) a synergistic processing is performed on the image; (e) converting the image into a plurality of feature points of constant scale; and (f) comparing the plurality of feature points having the same scale to the palm data in the database to identify the identity of the image provider . The method for identifying an identity according to claim 10, wherein the step (e) comprises the steps of: (e1) detecting an extreme value in a scale space; (e2) performing a screening feature point; (e3) determining a feature point; And (e4) construct a description vector of the feature points. The method for identifying an identity according to claim 10, wherein the step (b) comprises the steps of: (b1) using a near-infrared light source to illuminate a palm; and (b2) capturing the image reflected from the palm. The method for identifying an identity according to claim 12, wherein the near-infrared source has a wavelength of about 850 nm. The method for identifying an identity according to claim 12, wherein the step (b) uses an image capture card to capture the captured image. The method for identifying an identity according to claim 10, wherein the synergistic processing performed in step (d) is to increase the contrast of the image. The method for identifying an identity as described in claim 15 wherein the image enhancement processing uses a histogram equalization method. For example, the method for identifying the identity described in claim 10, wherein the step (f) compares the image with the palm data in the database one by one and finds an individual matching feature point set, and then according to the following one The similarity formula is used to identify the image: Where d represents the distance similarity, n represents the number of matching feature points and is greater than 1, x _i represents the distance between the reference point of the matching feature point in the image and other matching feature points, and y _i represents the matching feature in one of the palm data. The distance between the reference point of the point and other matching feature points. For example, the identity identification method described in claim 17 of the patent application, wherein n≧5.