Fingerprint registration method

Abstract

A fingerprint registration method includes acquiring a partial image of a fingerprint via a fingerprint sensor; providing the partial image of the fingerprint to an artificial intelligence model, wherein the artificial intelligence model is trained to generate a probability distribution information according to the partial image of the fingerprint, wherein the probability distribution information includes probabilities that the partial image of the fingerprint is located in T sub-regions of a complete fingerprint image; determining which one of the sub-regions the partial image of the fingerprint corresponds to according to the probability distribution information; determining a target registration region according to image quantities of the sub-regions, the target registration region being the sub-region with the least image quantity; and generating a guide signal to indicate a user to move a finger to the target registration region.

Description

Technical Field

[0001] This invention relates to a fingerprint registration method, and more particularly to a fingerprint registration method using an artificial intelligence (AI) model. Background Technology

[0002] More and more electronic devices and systems are using fingerprints to identify users. Most fingerprint sensors in these devices and systems are small and cannot cover the entire finger, thus only capturing a portion of the fingerprint and not the whole image. Traditional fingerprint registration methods typically set a number of presses or a set number of fingerprint images to be captured. However, during the fingerprint registration process, users may press the fingerprint sensor with the same or similar finger area each time, resulting in an insufficiently small effective area for the registered fingerprint, which can easily lead to verification failures. Summary of the Invention

[0003] One of the objectives of this invention is to provide a fingerprint registration method using an artificial intelligence model.

[0004] According to one embodiment, the present invention provides a fingerprint registration method, comprising the following steps: A. acquiring a partial fingerprint image via a fingerprint sensor; B. providing the partial fingerprint image to an artificial intelligence model, wherein the artificial intelligence model is trained to generate probability distribution information based on the partial fingerprint image, wherein the probability distribution information includes the probability that the partial fingerprint image is located in T sub-regions of a complete fingerprint image, where T is an integer greater than 1; C. determining which sub-region the partial fingerprint image corresponds to based on the probability distribution information; D. determining a target region to be registered based on the number of images in each sub-region, wherein the target region to be registered is the sub-region with the fewest number of images; and E. generating a guidance signal to instruct the user to move their finger toward the target region to be registered.

[0005] In one embodiment of the present invention, it further includes: F. When the number of images in the T sub-regions meets a preset condition, the registration process ends.

[0006] In one embodiment of the present invention, the preset condition includes that the number of images in all T regions is greater than a first threshold value, or that the number of images in the T regions that is greater than a second threshold value is greater than a third threshold value.

[0007] In one embodiment of the present invention, step F includes: after satisfying the preset condition, if a total of L fingerprint partial images are obtained and L is greater than a preset value, M fingerprint partial images are selected from the L fingerprint partial images, where L is an integer greater than or equal to 2 and M is an integer greater than 0 and less than or equal to L; and the plurality of fingerprint templates are generated based on the M fingerprint partial images and the registration process is terminated.

[0008] In one embodiment of the present invention, step F includes: generating L fingerprint templates based on the obtained L partial fingerprint images, where L is an integer greater than or equal to 2; and after satisfying the preset condition, if L is greater than a preset value, selecting M fingerprint templates from the L fingerprint templates and ending the registration process, where M is an integer greater than 0 and less than or equal to L.

[0009] In one embodiment of the present invention, step E includes calculating the centroid position of the fingerprint local image based on the probability distribution information corresponding to the fingerprint local image, and generating the guidance signal based on the relative positional relationship between the centroid position and the target area to be registered.

[0010] In one embodiment of the present invention, step E includes generating the guidance signal based on the relative positional relationship between the sub-region corresponding to the fingerprint local image and the target region to be registered.

[0011] In one embodiment of the present invention, step D further includes, when there are two or more sub-regions having the minimum number of images, using the sub-region that is farther away from the sub-region corresponding to the fingerprint local image as the target region to be registered.

[0012] The fingerprint registration method provided by this invention can collect images of various parts of a fingerprint relatively completely during the fingerprint registration process, which helps to identify registered users in the subsequent fingerprint verification process. Attached Figure Description

[0013] Figure 1 This invention illustrates an embodiment of the fingerprint registration method.

[0014] Figure 2 This invention illustrates an embodiment of the fingerprint registration system.

[0015] Figure 3 This illustrates an embodiment that divides a complete fingerprint image into T sub-regions.

[0016] Figure 4 This illustrates another embodiment of dividing a complete fingerprint image into T sub-regions.

[0017] Figure 5 This is a diagram showing the user's finger moving on the fingerprint sensor.

[0018] Explanation of reference numerals in the attached figures: 10-Fingerprint registration system; 11-Fingerprint sensor; 12-Artificial intelligence model; 13-Processor; 14-Memory; 20-Complete fingerprint image; 21~29-Sub-region; 30-Complete fingerprint image; 31~34-Sub-region; 40-Finger; 41-Upper vibrator; 42-Lower vibrator; M-Partial fingerprint image; P-Probability distribution information; S10~S16-Steps. Detailed Implementation

[0019] Figure 1 This invention illustrates an embodiment of the fingerprint registration method. Figure 2 This invention illustrates an embodiment of the fingerprint registration system. For example... Figure 2 As shown, the fingerprint registration system 10 of the present invention includes a fingerprint sensor 11, an artificial intelligence model 12, a processor 13, and a memory 14. The fingerprint sensor 11 is connected to the artificial intelligence model 12 and the processor 13, and transmits the sensed partial fingerprint image to the artificial intelligence model 12 and the processor 13. The artificial intelligence model 12 is connected to the fingerprint sensor 11 and the processor 13, and generates probability distribution information based on the partial fingerprint image, which is then sent to the processor 13. The processor 13 is connected to the fingerprint sensor 11, the artificial intelligence model 12, and the memory 14, and determines and marks the fingerprint sub-region corresponding to the partial fingerprint image based on the probability distribution information. In one embodiment, the artificial intelligence model 12 may be integrated into the processor 13.

[0020] Reference Figure 1 and Figure 2 In step S10, the fingerprint sensor 11 of the fingerprint registration system 10 senses the fingerprint of the user's finger and acquires a partial fingerprint image M. Then, the fingerprint sensor 11 provides the partial fingerprint image M to the artificial intelligence model 12. The artificial intelligence model 12 is trained to generate probability distribution information P based on the partial fingerprint image M, as shown in step S11. The probability distribution information P includes the probability that the partial fingerprint image M is located in T sub-regions of a complete fingerprint image, where T is an integer greater than 1. For example, as... Figure 3 As shown, a complete fingerprint image 20 is divided into T=9 sub-regions 21~29. When the fingerprint local image M obtained by the fingerprint sensor 11 is located in... Figure 3 When the fingerprint image is between sub-regions 24, 25, 27 and 28, the artificial intelligence model 12 provides a probability distribution information P=[0%, 0%, 0%, 20%, 10%, 0%, 55%, 15%, 0] based on the fingerprint local image M, where from left to right are the probabilities of the fingerprint local image M being located in sub-regions 21 to 29.

[0021] The processor 13 of the fingerprint registration system 10 acquires a partial fingerprint image M from the fingerprint sensor 11 and obtains probability distribution information P from the artificial intelligence model 12. The processor 13 can determine the sub-region corresponding to the partial fingerprint image M based on the probability distribution information P, as shown in step S12. For example, when the probability distribution information P of the partial fingerprint image M is [0%, 0%, 0%, 20%, 10%, 0%, 55%, 15%, 0], the processor 13 determines the sub-region 27 with the highest probability as the sub-region corresponding to the partial fingerprint image M.

[0022] Processor 13 counts the number of fingerprint local images corresponding to each sub-region 21-29, obtaining T image counts. Processor 13 determines a target region to be registered based on the number of images in the T regions, as shown in step S13. In one embodiment, the target region to be registered is the sub-region with the fewest number of images among the T sub-regions 21-29. Specifically, assuming that after executing steps S10 to S12 multiple times, the number of images in sub-region 21 is 1, the number of images in sub-region 22 is 3, the number of images in sub-region 23 is 5, the number of images in sub-region 24 is 2, the number of images in sub-region 25 is 1, the number of images in sub-region 26 is 0, the number of images in sub-region 27 is 4, the number of images in sub-region 28 is 3, and the number of images in sub-region 29 is 2, among which sub-region 26 has the fewest images, therefore processor 13 sets sub-region 26 as the target region to be registered. In one embodiment, the conditions for determining the target region to be registered further include that if there are multiple sub-regions with the minimum number of images, the sub-region that is farther away from the sub-region corresponding to the current fingerprint partial image is taken as the target region to be registered. Specifically, assuming that the number of images of sub-regions 21 and 25 is 1, while the number of images of the remaining sub-regions 22-24 and 26-29 is greater than 1, if the currently acquired fingerprint partial image M corresponds to sub-region 24, since sub-region 21 is farther away from sub-region 24, the processor 13 will set sub-region 21 as the target region to be registered.

[0023] After determining the target area to be registered, step S14 is performed. In step S14, the processor 13 generates a guidance signal based on the target area to instruct the user to move their finger toward the target area to obtain more of the fingerprint local image. In one embodiment, the processor 13 calculates the centroid position of the fingerprint local image M based on the probability distribution information P corresponding to the fingerprint local image M, and generates the guidance signal based on the relative positional relationship between the centroid position and the target area to be registered. In one embodiment, the processor 13 generates the guidance signal based on the relative positional relationship between the sub-region corresponding to the fingerprint local image M and the target area to be registered. In one embodiment, the guidance signal may be, but is not limited to, an arrow pattern or a sound. In one embodiment, the fingerprint registration system 10 may provide the guidance signal using, but is not limited to, a display or a speaker.

[0024] by Figure 3 This example illustrates how to calculate the centroid position of a fingerprint partial image M based on the probability distribution information P corresponding to that image. Assume the probability distribution information P of the fingerprint partial image M is [0%, 0%, 0%, 20%, 10%, 0%, 55%, 15%, 0], where sub-regions 24, 25, 27, and 28 have probability values. Therefore, the centroid position of the fingerprint partial image M can be calculated by substituting these four probability values ​​into a preset formula. In one embodiment, the preset formula includes:

[0025] ,

[0026] ,

[0027] Where Xc is the X coordinate of the centroid position, Yc is the Y coordinate of the centroid position, PT is the probability value of each sub-region, XT is the X coordinate of each sub-region, and YT is the Y coordinate of each sub-region. Assuming the coordinates of the fingerprint local image M are set to the origin (0,0), and the coordinates of sub-regions 24, 25, 27, and 28 are (-1 / 3, 2 / 3), (1 / 3, 2 / 3), (-1 / 3, -2 / 3), and (1 / 3, -2 / 3), respectively, we can calculate Xc = (-1 / 3)×0.2 + (1 / 3)×0.1 + (-1 / 3)×0.55 + (1 / 3)×0.15 = -0.167, Yc = (2 / 3)×0.2 + (2 / 3)×0.1 + (-2 / 3)×0.55 + (-2 / 3)×0.15 = -0.267. That is, the coordinates of the centroid of the fingerprint local image M are (-0.167, -0.267).

[0028] After generating the guidance signal, step S15 is performed. In step S15, the processor 13 determines whether the number of images in the T sub-regions meets a preset condition. If the number of images in the T sub-regions does not meet the preset condition, the process returns to step S10. If the number of images in the T sub-regions meets the preset condition, step S16 is performed to end the registration process. In one embodiment, the preset condition includes, but is not limited to, the number of images in all T sub-regions being greater than a first threshold, or the number of images in the T sub-regions greater than a second threshold being greater than a third threshold. For example, T is 9, the first threshold is 10, the second threshold is 10, and the third threshold is 6. When the number of images obtained in each of these 9 sub-regions is greater than 10, step S16 is performed to end the registration process. Alternatively, if the number of images in these 9 sub-regions are 11, 12, 11, 13, 8, 10, 15, 12, and 11 respectively, the processor determines that among these 9 image counts, 7 are greater than the second threshold of 10 and 6 are greater than the third threshold, and then proceeds to step S16 to end the registration process.

[0029] In one embodiment, the processor 13 further includes performing a post-processing step after step S15. After completing this post-processing, step S16 is performed to terminate the registration process. This post-processing includes, but is not limited to, performing calculations, filtering, converting a partial fingerprint image into a fingerprint template, and storing at least one of the fingerprint templates.

[0030] In one embodiment, the post-processing includes, after satisfying the preset condition, if a total of L fingerprint partial images are obtained and L is greater than a preset value, the processor 13 filters out M fingerprint partial images from the L fingerprint partial images, where L is an integer greater than or equal to 2, and M is an integer greater than 0 and less than or equal to L. After filtering out the M fingerprint partial images, the processor 13 generates multiple fingerprint templates based on the M fingerprint partial images and stores the multiple fingerprint templates in the memory 14. After storing the multiple fingerprint templates in the memory 14, the processor 13 performs step S16 to end the registration process.

[0031] In the aforementioned embodiments, the fingerprint partial image is first converted into a fingerprint template after determining whether the preset condition is met. However, the fingerprint registration method of the present invention is not limited to this. For example, the fingerprint registration method of the present invention can generate L fingerprint templates based on the obtained L fingerprint partial images before performing step S15, where L is an integer greater than or equal to 2. Then, after determining that the preset condition is met in step S15, the subsequent processing is performed. In this subsequent processing, if the processor 13 determines that L is greater than a preset value, the processor 13 will select M fingerprint templates from the L fingerprint templates and store them in the memory 14, where M is an integer greater than 0 and less than or equal to L. Finally, step S16 is performed to end the registration process.

[0032] In one embodiment, the aforementioned filtering method includes reducing the number of images in the T regions from the largest number to the same number as the second largest number, or reducing it by a preset number. Specifically, assuming the preset value is 130 images, the T regions are divided into [0, 0, 20, 20, 40, 80, 40, 0, 0], where T=9, and the total number of images in the T regions is L=200, which is greater than the preset value. Therefore, the largest 80 images are first reduced to the second largest number of 40 images. The reduction method can be First-In-First-Out (FIFO) or random reduction. After the reduction, the T regions are divided into [0, 0, 20, 20, 40, 40, 40, 0, 0]. Since the total number of images is L=160, which is still greater than the preset value, the largest 40 images are then reduced by 10 images each (the preset number). After deletion, the number of images in the T regions is divided into [0, 0, 20, 20, 30, 30, 30, 0, 0]. Since the total number of images L=130 meets the preset value, the filtering process ends.

[0033] exist Figure 1 In the fingerprint registration method shown, some steps can be replaced or omitted. For example, in one embodiment, steps S13 and S14 can be omitted, that is, the step of guiding the user can be omitted.

[0034] Figure 4 This illustrates another embodiment of dividing a complete fingerprint image into T sub-regions. Figure 5 This diagram illustrates the movement of a user's finger on the fingerprint sensor. Figure 4 In the image, the complete fingerprint image 30 is divided into T=4 sub-regions 31~34, which are arranged vertically. Figure 4 The embodiments are applicable to narrow and elongated fingerprint sensors 11, such as Figure 5 As shown. (Refer to...) Figure 1 The processor 13 counts the number of fingerprint local images corresponding to each sub-region 31-34 to obtain the number of images for the four regions. After performing steps S10-S12, the processor 13 determines a target region to be registered based on the number of images for the four sub-regions, such as... Figure 1 As shown in step S13, the processor 13 generates a guidance signal based on the target area to be registered, instructing the user to move their finger 40 up or down toward the target area to be registered, such as... Figure 1The process is described in step S14. After step S14, the processor 13 executes steps S15-S16. In one embodiment, the guidance signal can be implemented by the upper vibrator 41 and the lower vibrator 42 of the fingerprint sensor 11. When the upper vibrator 41 vibrates, it indicates that the user's finger 40 is moving upward, and when the lower vibrator 42 vibrates, it indicates that the user's finger 40 is moving downward. In one embodiment, the guidance signal can be, but is not limited to, an arrow pattern or a sound.

[0035] From the above description, it can be understood that the fingerprint registration method of the present invention includes:

[0036] Step A: Acquire a partial image of a fingerprint using a fingerprint sensor;

[0037] Step B: Provide the partial fingerprint image to an artificial intelligence model, wherein the artificial intelligence model is trained to generate a probability distribution information based on the partial fingerprint image, wherein the probability distribution information includes the probability that the partial fingerprint image is located in T sub-regions in a complete fingerprint image, where T is an integer greater than 1;

[0038] Step C: Based on the probability distribution information, determine which sub-region the fingerprint image corresponds to;

[0039] Step D: Determine a target region to be registered based on the number of images in each sub-region; the target region to be registered is the sub-region with the fewest images.

[0040] Step E: Generate a guiding signal to instruct the user to move their finger toward the target area to be registered.

[0041] Figure 2 The training methods for the AI ​​model 12 include, but are not limited to, acquiring a large number of complete fingerprint images of different fingers and dividing each complete fingerprint image into T sub-regions (e.g., Figure 3 (The nine sub-regions are shown). Based on the size of the fingerprint sensor 11, several partial fingerprint images are acquired in each sub-region, and the sub-region corresponding to each partial fingerprint image is marked. If the acquired partial fingerprint images cover multiple sub-regions, such as... Figure 3 As shown in the fingerprint local image M, the sub-region 27 with the largest coverage area is marked as the sub-region corresponding to the fingerprint local image M. Finally, each fingerprint local image and its corresponding sub-region are input into the artificial intelligence model 12. The artificial intelligence model 12 can use, but is not limited to, a cross-entropy loss function for analysis and learning.

[0042] The above description is merely an embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been provided above with embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A fingerprint registration method, characterized in that, Includes the following steps: A. Acquire a partial image of a fingerprint using a fingerprint sensor; B. Provide the partial fingerprint image to an artificial intelligence model, wherein the artificial intelligence model is trained to generate a probability distribution information based on the partial fingerprint image, wherein the probability distribution information includes the probability that the partial fingerprint image is located in T sub-regions of a complete fingerprint image, where T is an integer greater than 1; C. Based on the probability distribution information, determine which sub-region the fingerprint image corresponds to; D. Determine a target region to be registered based on the number of images in each sub-region; the target region to be registered is the sub-region with the fewest number of images. E. Generate a guiding signal to instruct the user to move their finger toward the target area to be registered.

2. The fingerprint registration method as described in claim 1, characterized in that, Also includes: F. When the number of images in the T sub-regions meets a preset condition, the registration process ends.

3. The fingerprint registration method as described in claim 2, characterized in that, The preset conditions include either the number of images in all T regions being greater than a first threshold, or the number of images in the T regions being greater than a second threshold being greater than a third threshold.

4. The fingerprint registration method as described in claim 2, characterized in that, Step F includes: After meeting the preset condition, if a total of L partial fingerprint images are obtained and L is greater than a preset value, M partial fingerprint images are selected from the L partial fingerprint images, where L is an integer greater than or equal to 2, and M is an integer greater than 0 and less than or equal to L; and The multiple fingerprint templates are generated based on the M partial images of the fingerprint, and the registration process is then completed.

5. The fingerprint registration method as described in claim 2, characterized in that, Step F includes: Based on the obtained L partial fingerprint images, L fingerprint templates are generated, where L is an integer greater than or equal to 2; and After the preset condition is met, if L is greater than a preset value, M fingerprint templates are selected from the L fingerprint templates and the registration process ends, where M is an integer greater than 0 and less than or equal to L.

6. The fingerprint registration method as described in claim 1, characterized in that, Step E includes calculating the centroid position of the fingerprint local image based on the probability distribution information corresponding to the fingerprint local image, and generating the guidance signal based on the relative positional relationship between the centroid position and the target area to be registered.

7. The fingerprint registration method as described in claim 1, characterized in that, Step E includes generating the guidance signal based on the relative positional relationship between the sub-region corresponding to the fingerprint partial image and the target region to be registered.

8. The fingerprint registration method as described in claim 1, characterized in that, Step D further includes, when there are two or more sub-regions with the minimum number of images, using the sub-region that is farther away from the sub-region corresponding to the fingerprint local image as the target region to be registered.

Images