Fingerprint information processing device, fingerprint information processing method, and recording medium

Abstract

The present invention provides a fingerprint information processing device, a fingerprint information processing method, and a recording medium that appropriately classify input fingerprint images. [Solution] The fingerprint information processing device (1) includes an output means (11) that outputs a confidence score, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types, using a fingerprint image and a learning model constructed by machine learning using learning data including sample images showing fingerprints, and a processing means (12) that performs processing based on the confidence score.

Description

Technical Field

[0001] This disclosure relates to the technical field of fingerprint information processing apparatuses, fingerprint information processing methods, and recording media.

Background Art

[0002] For example, an apparatus has been proposed that generates a ridge direction pattern from a fingerprint image and classifies fingerprints based on the shape of ridges in the vicinity of the core of the ridge direction pattern and the tendency of the ridge direction (see Patent Document 1). In addition, Patent Documents 2 and 3 are cited as prior art documents related to this disclosure.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0004] This disclosure aims to provide a fingerprint information processing apparatus, a fingerprint information processing method, and a recording medium that improve on the technologies described in the prior art documents.

Means for Solving the Problems

[0005] One embodiment of the fingerprint information processing device of this disclosure includes an output means that outputs a confidence score, which is an index indicating the likelihood that the fingerprint shown by the fingerprint image belongs to at least one of a plurality of pattern types, using a fingerprint image and a learning model constructed by machine learning using learning data including sample images showing a fingerprint; and a processing means that performs processing based on the confidence score. The output means outputs the confidence score using a fingerprint image which is already registered and the learning model. The processing means, as part of the processing, estimates the pattern type of the fingerprint shown by the fingerprint image based on the confidence score, and if the estimated pattern type is different from the pattern type already associated with the fingerprint image, it performs at least one of notification and updating the pattern type already associated with the fingerprint image.

[0006] One embodiment of the fingerprint information processing method of this disclosure is a fingerprint information processing method in which a computer uses a fingerprint image and a learning model constructed by machine learning using training data including sample images showing fingerprints to output a confidence score, which is an index indicating the likelihood that the fingerprint shown by the fingerprint image belongs to at least one of a plurality of pattern types, and the computer performs processing based on the confidence score, wherein the computer uses one already registered fingerprint image as the fingerprint image and the learning model to output the confidence score, and as processing, the computer estimates the pattern type of the fingerprint shown by the one fingerprint image based on the confidence score, and if the estimated pattern type is different from the pattern type already associated with the one fingerprint image, it performs at least one of notification and updating the pattern type already associated with the one fingerprint image.

[0007] One embodiment of the recording medium of this disclosure contains a computer program for causing a computer to execute a fingerprint information processing method that uses a fingerprint image and a learning model constructed by machine learning using training data including sample images showing fingerprints to output a confidence score, which is an index indicating the likelihood that the fingerprint shown by the fingerprint image belongs to at least one of a plurality of pattern types, and performs processing based on the confidence score, wherein the recording medium uses one already registered fingerprint image as the fingerprint image and the learning model to output the confidence score, and as processing, estimates the pattern type of the fingerprint shown by the one fingerprint image based on the confidence score, and if the estimated pattern type is different from the pattern type already associated with the one fingerprint image, notifies the computer and updates the pattern type already associated with the one fingerprint image, or at least one of the above. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing an example of the configuration of an information processing device. [Figure 2] This is a block diagram showing another example of the configuration of an information processing device. [Figure 3] This figure shows an example of an output image. [Figure 4] This figure shows another example of an output image. [Figure 5] This is a flowchart showing the operation according to the second embodiment. [Figure 6] This is a flowchart showing the operation according to the third embodiment. [Figure 7] This is a flowchart showing the operation according to the fourth embodiment. [Figure 8] This is a flowchart showing the operation according to the fifth embodiment. [Figure 9] This flowchart shows the operation according to the sixth embodiment. [Figure 10] This is a flowchart showing the operation according to the seventh embodiment. [Figure 11] This is a flowchart showing the operation according to the eighth embodiment. [Modes for carrying out the invention]

[0009] The following describes embodiments of a fingerprint information processing device, a fingerprint information processing method, and a recording medium with reference to the drawings.

[0010] <First Embodiment> A first embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figure 1. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the first embodiment will be described using the information processing device 1. Figure 1 is a block diagram showing the configuration of the information processing device 1.

[0011] As shown in Figure 1, the information processing device 1 comprises an output unit 11 and a processing unit 12. The output unit 11 uses a fingerprint image and a learning model constructed by machine learning using training data including sample images showing fingerprints to output a confidence score, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. The processing unit 12 performs processing based on the confidence score.

[0012] In the information processing device 1, first, the output unit 11 may output a confidence score using the fingerprint image and the learned model. Next, the processing unit 12 may perform processing based on the confidence score. In other words, the information processing device 1 may output a confidence score using the fingerprint image and the learned model, and perform processing based on the confidence score. Such an information processing device 1 may be realized, for example, by a computer reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the computer to output a confidence score using the fingerprint image and the learned model, and to perform processing based on the confidence score.

[0013] The fingerprint image may include, for example, an image generated by detecting a fingerprint with a sensor, and an image generated by imaging a pressed fingerprint or a latent fingerprint with a camera or reading it with a scanner. As the sensor for detecting a fingerprint, for example, a contact sensor such as an optical method, a capacitance method, an ultrasonic method, etc., or a non-contact sensor such as OCT (Optical Coherence Tomography), a three-dimensional fingerprint scanner, etc. can be applied. The pattern type means a pattern formed by the ridges on the fingertip (i.e., fingerprint), for example, a pattern in which patterns having common forms are grouped based on, for example, the shape of the ridges and the flow direction of the ridges. The pattern type may include, for example, an arch pattern, a loop pattern, a whorl pattern, etc.

[0014] Existing various aspects can be applied to the method of constructing a learning model by machine learning using learning data including a sample image showing a fingerprint. Therefore, the detailed description of the method of constructing the learning model is omitted. The learning model may be constructed by deep learning, which is an aspect of machine learning. The learning model constructed by deep learning may mean a mathematical model constructed by machine learning using a neural network having a multi-layer structure with a plurality of intermediate layers (which may also be referred to as hidden layers). The neural network may be, for example, a convolutional neural network. As the model structure related to the convolutional neural network, for example, VGG, MobileNet, etc. may be used.

[0015] The confidence level is an index indicating the probability that a fingerprint corresponds to at least one of a plurality of pattern types. The higher the probability that a fingerprint corresponds to one pattern type, the higher the confidence level may be. In other words, the lower the probability that a fingerprint corresponds to one pattern type, the lower the confidence level may be. Note that the confidence level may be represented by a numerical value, or may be represented by a grade or class such as A, B, … etc. The confidence level may also be referred to as a probability.

[0016] The output unit 11 may obtain a confidence level for, for example, one of a plurality of pattern types using the fingerprint image and the learning model, and output the obtained confidence level. The output unit 11 may obtain, for example, a plurality of confidence levels respectively corresponding to a plurality of pattern types using the fingerprint image and the learning model, and output the highest confidence level among the obtained plurality of confidence levels. The output unit 11 may obtain, for example, a plurality of confidence levels respectively corresponding to a plurality of pattern types using the fingerprint image and the learning model, and output one or more confidence levels higher than a predetermined value among the obtained plurality of confidence levels. The output unit 11 may obtain, for example, a plurality of confidence levels respectively corresponding to a plurality of pattern types using the fingerprint image and the learning model, and output all of the obtained plurality of confidence levels. Incidentally, the output unit 11 may output the confidence level to, for example, a display device. In this case, the confidence level output from the output unit 11 may be displayed on the screen of the display device.

[0017] The processing unit 12 executes processing based on the confidence level output from the output unit 11. The "processing based on the confidence level" may include processing directly based on the confidence level and processing indirectly based on the confidence level.

[0018] Processing directly based on the confidence level may include, for example, processing of estimating the pattern type to which the fingerprint indicated by the fingerprint image belongs from a plurality of pattern types based on the confidence level. Processing indirectly based on the confidence level may include, for example, processing of collating the fingerprint indicated by the fingerprint image after limiting the collation target based on the pattern type to which the fingerprint indicated by the fingerprint image estimated based on the confidence level belongs.

[0019] According to the first embodiment, the prior art can be improved.

[0020] <Second Embodiment> A second embodiment of the fingerprint information processing apparatus, the fingerprint information processing method, and the recording medium will be described with reference to FIGS. 2 to 5. Hereinafter, using the information processing apparatus 2, the fingerprint information processing apparatus, the fingerprint information processing method, and the recording medium according to the second embodiment will be described. FIG. 2 is a block diagram showing the configuration of the information processing apparatus �.

[0021] As shown in Figure 2, the information processing device 2 comprises an arithmetic unit 21 and a storage device 22. The information processing device 2 may also include a communication device 23, an input device 24, and an output device 25. However, the information processing device 2 does not have to include at least one of the communication device 23, the input device 24, and the output device 25. In the information processing device 2, the arithmetic unit 21, the storage device 22, the communication device 23, the input device 24, and the output device 25 may be connected via a data bus 26.

[0022] The computing device 21 may include, for example, at least one of a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and an FPGA (Field Programmable Gate Array).

[0023] The storage device 22 may include, for example, at least one of RAM (Random Access Memory), ROM (Read Only Memory), a hard disk drive, a magneto-optical disk drive, an SSD (Solid State Drive), and an optical disk array. In other words, the storage device 22 may include non-temporary recording media. The storage device 22 is capable of storing desired data. For example, the storage device 22 may temporarily store a computer program executed by the arithmetic unit 21. The storage device 22 may temporarily store data that the arithmetic unit 21 uses temporarily when it is executing a computer program.

[0024] The communication device 23 may be able to communicate with devices outside the information processing device 2 via a communication network (not shown). The communication network may be a wide-area network such as the Internet, or a narrow-area network such as a LAN (Local Area Network). The communication device 23 may use wired communication or wireless communication.

[0025] The input device 24 is a device capable of receiving information input to the information processing device 2 from an external source. It may include an operating device (e.g., a keyboard, mouse, touch panel, etc.) that can be operated by the operator of the information processing device 2. The input device 24 may include a recording medium reader capable of reading information recorded on a recording medium that can be attached to and detached from the information processing device 2, such as a USB (Universal Serial Bus) memory. When information is input to the information processing device 2 via the communication device 23 (in other words, when the information processing device 2 acquires information via the communication device 23), the communication device 23 may function as an input device.

[0026] The output device 25 is a device capable of outputting information to the outside of the information processing device 2. The output device 25 may output visual information such as characters or images, auditory information such as sounds, or tactile information such as vibrations. The output device 25 may include at least one of a display, speaker, printer, and vibration motor. The output device 25 may also be capable of outputting information to a recording medium that can be attached to or detached from the information processing device 2, such as a USB memory stick. When the information processing device 2 outputs information via the communication device 23, the communication device 23 may function as an output device.

[0027] The arithmetic unit 21 may have an output unit 211 and a processing unit 212, for example, as a logically implemented functional block or as a physically implemented processing circuit. At least one of the output unit 211 and the processing unit 212 may be implemented in a form that combines a logical functional block and a physical processing circuit (i.e., hardware). If at least a part of the output unit 211 and the processing unit 212 is a functional block, at least a part of the output unit 211 and the processing unit 212 may be implemented by the arithmetic unit 21 executing a predetermined computer program.

[0028] The arithmetic unit 21 may obtain the predetermined computer program from, for example, the storage device 22 (in other words, it may read it). The arithmetic unit 21 may read the predetermined computer program stored on a computer-readable and non-temporary recording medium using a recording medium reader (not shown) provided by the information processing device 2. The arithmetic unit 21 may obtain the predetermined computer program from an external device (not shown) of the information processing device 2 via the communication device 23 (in other words, it may download or read it). The recording medium used to record the predetermined computer program executed by the arithmetic unit 21 may be at least one of an optical disc, a magnetic medium, a magneto-optical disc, a semiconductor memory, and any other medium capable of storing a program.

[0029] The output unit 211 has a learning model constructed by machine learning using training data including sample images showing fingerprints. The output unit 211 inputs the fingerprint image into the learning model and obtains a confidence score from the learning model. The confidence score is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. For this reason, the output unit 211 may obtain the confidence score in association with the pattern type.

[0030] The input device 24 may include, for example, a sensor capable of detecting fingerprints. A fingerprint image may be generated when the sensor detects a fingerprint. The output unit 211 may acquire the generated fingerprint image. The input device 24 may include, for example, a scanner. A fingerprint image may be generated when a embossed fingerprint or a residual fingerprint is read by the scanner. The output unit 211 may acquire the generated fingerprint image. The input device 24 may include, for example, an image acquisition device capable of acquiring an image captured by a camera. A fingerprint image may be generated when the camera captures an embossed fingerprint or a residual fingerprint. The output unit 211 may acquire the fingerprint image via the image acquisition device included in the input device 24.

[0031] The output unit 211 transmits (outputs) a signal indicating the confidence level to the processing unit 212. In this case, the output unit 211 may transmit a signal to the processing unit 212 indicating, for example, the confidence level and the pattern type associated with that confidence level. The output unit 211 may also transmit a signal indicating, for example, the confidence level and the pattern type associated with that confidence level to, for example, an output device 25. In this case, the output device 25 may display (in other words, output) at least one of the characters and images indicating at least one pattern type, and at least one of the characters and images indicating the confidence level associated with that at least one pattern type. As a result, an image like the one shown in Figure 3 may be displayed.

[0032] The processing unit 212 performs processing based on the confidence level. For example, if the output unit 211 transmits signals indicating the confidence level and the pattern type associated with that confidence level to the processing unit 212 and the output device 25, the processing unit 212 may determine the order of the pattern types based on the confidence level. The processing unit 212 may then transmit a signal indicating the determined order of the pattern types to the output device 25. In this case, the output device 25 may display at least one of the characters and images indicating the pattern type and at least one of the characters and images indicating the confidence level associated with the pattern type, according to the determined order of the pattern types. As a result, an image like the one shown in Figure 4 may be displayed.

[0033] For example, if the output unit 211 transmits a signal to the processing unit 212 indicating a confidence level and the pattern type associated with that confidence level, the processing unit 212 may compare the confidence level with a first predetermined value. Here, the confidence level is assumed to be expressed as a numerical value. If the confidence level is higher than the first predetermined value, the processing unit 212 may associate the fingerprint image with the pattern type associated with the confidence level higher than the first predetermined value. In other words, the processing unit 212 may classify the fingerprint shown by the fingerprint image into a pattern type associated with a confidence level higher than the first predetermined value. If there is no pattern type associated with a confidence level higher than the first predetermined value, the processing unit 212 may classify the fingerprint shown by the fingerprint image into an incomplete pattern, for example.

[0034] Furthermore, if the multiple confidence levels associated with each of the multiple pattern types are higher than a first predetermined value, the processing unit 212 may associate the multiple pattern types with the fingerprint image. In this case, the processing unit 212 may set the pattern type associated with the highest confidence level as the main pattern (i.e., the primary pattern type). The processing unit 212 may set the pattern types associated with the confidence levels excluding the highest confidence level among the multiple confidence levels higher than the first predetermined value as secondary patterns (i.e., auxiliary pattern types).

[0035] The "first predetermined value" is a value that determines whether a fingerprint image can be associated with a single pattern type, or in other words, whether a fingerprint shown by a fingerprint image can be classified into a single pattern type. The first predetermined value may be a fixed value set in advance, or it may be a variable value corresponding to some physical quantity or parameter. The first predetermined value may be set as follows: For example, the confidence level for each pattern type output from the output unit 211 for a single fingerprint image may be linked to the appraisal result obtained by a fingerprint expert who appraised the fingerprint shown by the single fingerprint image. This process may be performed for multiple fingerprint images. The first predetermined value may be set based on the distribution of confidence levels in which the pattern type associated with the highest confidence level matches the pattern type shown by the appraisal result.

[0036] The processing unit 212 may, for example, transmit signals indicating a fingerprint image and the pattern type associated with the fingerprint image to a device capable of performing fingerprint matching, different from the information processing device 2, via the communication device 23. For example, if the storage device 22 contains a fingerprint database, the processing unit 212 may perform fingerprint matching using the fingerprint database. Various existing methods can be applied to fingerprint matching. For this reason, a detailed explanation of fingerprint matching will be omitted, but an overview will be given below.

[0037] In a fingerprint database, each of multiple fingerprint images may be associated with a pattern classification. Various existing methods can be applied to this association. For example, one method is to generate or update table information showing the correspondence between fingerprint images and pattern types. Another method is to add data indicating the pattern type to the header of the image data related to the fingerprint image.

[0038] The processing unit 212 may extract fingerprint images to be compared with a given fingerprint image from the fingerprint database based on the pattern type associated with that fingerprint image. As a result, fingerprint images associated with the same pattern type as the one associated with the given fingerprint image are extracted from the fingerprint database as matching targets for the given fingerprint image. On the other hand, fingerprint images associated with a different pattern type than the one associated with the given fingerprint image do not necessarily need to be extracted from the fingerprint database as matching targets for the given fingerprint image. If a given fingerprint image is associated with both a primary pattern type and a secondary pattern type, fingerprint images associated with the same pattern type as the primary pattern type and fingerprint images associated with the same pattern type as the secondary pattern type may be extracted from the fingerprint database. The processing unit 212 may compare the two fingerprints by comparing a plurality of feature points related to the fingerprint shown in the given fingerprint image with a plurality of feature points related to the fingerprint shown in the matching target fingerprint image. The processing unit 212 may determine that the two fingerprints match if some of the feature points (for example, 12 feature points) match in both fingerprints.

[0039] The processing unit 212 may, for example, store fingerprint images and the pattern types associated with those fingerprint images in the storage device 22, linking them to each other. As a result, a fingerprint database may be constructed or updated. The processing unit 212 may also store the confidence level associated with the pattern types associated with the fingerprint images in the storage device, linking it to the fingerprint images. The processing unit 212 may, for example, transmit signals indicating the fingerprint images and the pattern types associated with those fingerprint images to a device managing the fingerprint database, which is different from the information processing device 2, via the communication device 23. As a result, the fingerprint database may be updated.

[0040] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 5. In Figure 5, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image (step S101). The output unit 211 uses the fingerprint image and the learning model to output a confidence score (step S102). The processing unit 212 of the arithmetic unit 21 performs processing based on the confidence score (step S103).

[0041] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation. The arithmetic unit 21 of the information processing device 2 may correspond to the information processing device 1 according to the first embodiment described above.

[0042] According to the second embodiment, the prior art can be improved.

[0043] <Third Embodiment> A third embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 6. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the third embodiment will be described using the information processing device 2. The third embodiment differs from the second embodiment described above in that the output unit 211 of the arithmetic unit 21 has multiple learning models. Other aspects of the third embodiment may be the same as those of the second embodiment.

[0044] The output unit 211 may have, for example, a first model and a second model constructed by machine learning using training data that includes sample images showing fingerprints. In other words, the output unit 211 may have a first model and a second model as the training models in the second embodiment described above. The output unit 211 may have three or more training models.

[0045] Here, the first model and the second model are learning models whose output tendencies in response to inputs differ from each other. Such first and second models may be constructed, for example, by making the number of hidden layers constituting the neural network different from each other. The first and second models may be constructed, for example, by making the number of nodes included in the hidden layers constituting the neural network different from each other. The first and second models may be constructed, for example, by making the model structures related to the neural network different from each other. The first and second models may be constructed, for example, by making the training data used for machine learning of the neural network different from each other.

[0046] The output unit 211 inputs a fingerprint image to the first model and obtains first confidence data indicating the confidence level as an output result of the first model. The output unit 211 inputs the same fingerprint image to the second model and obtains second confidence data indicating the confidence level as an output result of the second model. The first confidence data and the second confidence data each represent multiple confidence levels corresponding to multiple pattern types. In the third embodiment, the confidence level is expressed numerically.

[0047] The output unit 211 combines the first confidence level data and the second confidence level data. Specifically, the output unit 211 combines multiple confidence levels corresponding to multiple pattern types, each indicated by the first confidence level data and the second confidence level data, for each pattern type. In this case, the output unit 211 may combine the confidence level corresponding to one pattern type indicated by the first confidence level data with the confidence level corresponding to the same pattern type indicated by the second confidence level data to obtain a combined confidence level for one pattern type. The "combined confidence level" may be, for example, an average value or a sum. When obtaining the combined confidence level, for example, the output trends of the first model and the second model with respect to the input may be used as the weights for the combination. For example, the detection accuracy of right-flowing hoof prints of the first model may be better than that of the second model, and the detection accuracy of left-flowing hoof prints of the second model may be better than that of the first model. For example, when calculating a composite confidence value for a right-sided hoof print, the confidence weights corresponding to the right-sided hoof print, as shown by the first confidence data, may be made greater than the confidence weights corresponding to the right-sided hoof print, as shown by the second confidence data, and the confidence values ​​may be combined. Similarly, when calculating a composite confidence value for a left-sided hoof print, the confidence weights corresponding to the left-sided hoof print, as shown by the second confidence data, may be made greater than the confidence weights corresponding to the left-sided hoof print, as shown by the first confidence data, and the confidence values ​​may be combined.

[0048] The first confidence data and the second confidence data are combined to generate a third confidence data that shows the combined confidence level for each pattern type. The output unit 211 transmits a signal indicating the combined confidence level based on the third confidence data to the processing unit 212.

[0049] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 6. In Figure 6, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image (step S101). The output unit 211 acquires first confidence level data by inputting the fingerprint image into the first model (step S201). In parallel with the processing in step S201, the output unit 211 acquires second confidence level data by inputting the fingerprint image into the second model (step S202). Note that the output unit 211 may execute the processing in step S202 on the condition that the first confidence level data has been acquired in the processing of step S201. In other words, the output unit 211 may acquire the second confidence level data after acquiring the first confidence level data. Alternatively, the output unit 211 may acquire the first confidence level data after acquiring the second confidence level data. The output unit 211 combines the first confidence level data and the second confidence level data (step S203). The output unit 211 outputs the combined confidence level, which is represented by the third confidence level data generated by combining the first confidence level data and the second confidence level data (step S102). The processing unit 212 of the arithmetic unit 21 performs processing based on the confidence level (step S103).

[0050] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0051] According to the third embodiment, the accuracy of the confidence level output from the output unit 211 can be improved.

[0052] <Fourth Embodiment> A fourth embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 7. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the fourth embodiment will be described using the information processing device 2. Here, an example will be given in which the information processing device 2 is applied to the review of an existing fingerprint database. In the fourth embodiment, the processing performed by the processing unit 212 (i.e., processing based on confidence level) will be mainly described. Other aspects of the fourth embodiment may be the same as those of the second and third embodiments.

[0053] In fingerprint databases, fingerprints are often classified and registered according to their pattern type. In other words, fingerprint databases often link fingerprint images to the pattern type to which those fingerprints are classified. This is done, for example, to perform fingerprint matching efficiently. Specifically, by limiting the search range of the fingerprint database based on the pattern type, the number of matching targets can be limited (i.e., reduced).

[0054] For example, fingerprint databases managed by public institutions may contain fingerprint data collected over several decades. Traditionally, fingerprint patterns are often determined using a rule-based system (i.e., according to rules written by humans). Rule-based systems can determine the fingerprint pattern with relatively high accuracy as long as the pattern matches the rules. On the other hand, they cannot identify fingerprint patterns from aspects that cannot be described as rules. Therefore, for example, if a fingerprint can be interpreted as having multiple patterns, it may be classified into the wrong pattern. For example, in fingerprint matching where the search range of the fingerprint database is limited based on the pattern, fingerprints classified into the wrong pattern will be excluded from the matching.

[0055] If a learning model constructed using deep learning is used as the learning model in the second and third embodiments described above, it is expected that it will be possible to identify fingerprint pattern types that take into account aspects that cannot be described as rules. Therefore, existing fingerprint databases may be reviewed using the method described below.

[0056] The information processing device 2 may perform the following operations, for example, to support the review of the fingerprint database. The output unit 211 of the arithmetic unit 21 acquires a fingerprint image registered in the fingerprint database. The output unit 211 inputs the fingerprint image into a learning model constructed by deep learning to acquire a confidence score for the fingerprint image. In this case, the output unit 211 may acquire multiple confidence scores for the fingerprint image, each corresponding to a plurality of pattern types. The output unit 211 transmits a signal indicating the confidence score for the fingerprint image to the processing unit 212 of the arithmetic unit 21.

[0057] The processing unit 212 compares the confidence level of one fingerprint image with a first predetermined value (see second embodiment). Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image.

[0058] If, for a single fingerprint image, multiple confidence levels corresponding to multiple pattern types include a confidence level higher than a first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to the confidence level higher than the first predetermined value. If multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types.

[0059] If none of the multiple confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0060] The processing unit 212 determines whether the pattern type associated with a fingerprint image in the fingerprint database is the same as the pattern type associated with the fingerprint image based on the confidence level. If the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with the fingerprint image based on the confidence level, the processing unit 212 issues a notification prompting a review of the pattern type.

[0061] The processing unit 212 may, as a notification, send an email to, for example, the administrator of the fingerprint database, prompting them to review the pattern type. The processing unit 212 may, as a notification, display a fingerprint image in which the pattern type associated with a particular fingerprint image in the fingerprint database differs from the pattern type associated with that fingerprint image based on the confidence level. The notification method is not limited to these, and various existing methods can be applied.

[0062] Furthermore, if the pattern type associated with a fingerprint image in the fingerprint database differs from the pattern type associated with the fingerprint image based on the confidence level, the processing unit 212 may provide notification to prompt a review of the pattern type if the confidence level corresponding to the pattern type associated with the fingerprint image based on the confidence level is higher than a second predetermined value.

[0063] The "second predetermined value" is a value that determines whether or not to notify that the pattern type is different. The second predetermined value may be a fixed value set in advance, or it may be a variable value corresponding to some physical quantity or parameter. The second predetermined value may be set as follows: For example, if the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with a fingerprint image based on the confidence level, the relationship between the fingerprint with the corrected pattern type by the fingerprint expert and the confidence level may be determined. The second predetermined value may be set based on the determined relationship.

[0064] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 7. In Figure 7, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image from the fingerprint database (step S101). The output unit 211 inputs the fingerprint image into a learning model constructed by deep learning to obtain a confidence score for the fingerprint image. The output unit 211 outputs the confidence score for the fingerprint image (step S102).

[0065] The processing unit 212 of the arithmetic unit 21 compares each of the multiple confidence levels corresponding to each of the multiple pattern types with a first predetermined value based on the confidence level for one fingerprint image. Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image (step S301).

[0066] In the processing of step S301, if, for a single fingerprint image, the confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image is a pattern type corresponding to a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to a confidence level higher than the first predetermined value. If the multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types. If the multiple confidence levels corresponding to each of the multiple pattern types do not include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by the single fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate the single fingerprint image with an incomplete pattern as a pattern type.

[0067] The processing unit 212 determines whether the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with the fingerprint image based on the confidence level (i.e., the pattern type associated with the fingerprint image in the process of step S301) (step S302). If, in the process of step S302, it is determined that the pattern type associated with a fingerprint image in the fingerprint database is the same as the pattern type associated with the fingerprint image based on the confidence level (step S302: No), the operation shown in Figure 7 is terminated.

[0068] In the process of step S302, if it is determined that the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with a fingerprint image based on the confidence level (step S302: Yes), the processing unit 212 will issue a notification prompting a review of the pattern type (step S303).

[0069] Furthermore, in the processing of step S302, if it is determined that the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with the fingerprint image based on the confidence level (step S302: Yes), the processing unit 212 may determine whether the confidence level corresponding to the pattern type associated with the fingerprint image is higher than a second predetermined value. If the confidence level is determined to be higher than the second predetermined value, the processing unit 212 may provide notification to prompt a review of the pattern type. On the other hand, if the confidence level is determined to be lower than the second predetermined value, the processing unit 212 does not need to provide notification to prompt a review of the pattern type. If the confidence level and the second predetermined value are equal, they may be treated as being included in either.

[0070] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0071] According to the fourth embodiment, it is possible to detect fingerprints that may have been classified into the wrong pattern type among multiple fingerprints registered in the fingerprint database.

[0072] (First variation) Instead of issuing a notification prompting a review of the pattern type, the processing unit 212 may, for example, replace the pattern type associated with a fingerprint image in the fingerprint database with the pattern type associated with the fingerprint image based on the confidence level. In other words, if, in the process of step S302, it is determined that the pattern type associated with a fingerprint image in the fingerprint database is different from the pattern type associated with the fingerprint image based on the confidence level (step S302: Yes), the processing unit 212 may replace the pattern type associated with a fingerprint image in the fingerprint database with the pattern type associated with the fingerprint image based on the confidence level. In this case, the processing unit 212 may issue a notification that the pattern type has been replaced. Note that replacing the pattern type associated with a fingerprint image in the fingerprint database may be considered equivalent to updating the pattern type associated with a fingerprint image in the fingerprint database.

[0073] (Second variation) Alternatively, instead of issuing a notification prompting a review of the pattern type, the processing unit 212 may, for example, register the pattern type associated with one fingerprint image based on the confidence level as a sub-pattern related to one fingerprint image in the fingerprint database. In other words, if, in the processing of step S302, it is determined that the pattern type associated with one fingerprint image in the fingerprint database is different from the pattern type associated with one fingerprint image based on the confidence level (step S302: Yes), the processing unit 212 may associate the pattern type associated with one fingerprint image based on the confidence level with one fingerprint image as a sub-pattern related to one fingerprint image. In this case, the processing unit 212 may issue a notification that the sub-pattern has been registered. Note that the registration of a sub-pattern related to one fingerprint image in the fingerprint database may be considered equivalent to updating the pattern type associated with one fingerprint image in the fingerprint database.

[0074] <Fifth Embodiment> A fifth embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 8. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the fifth embodiment will be described using the information processing device 2. Here, an example will be given in which the information processing device 2 is applied to a fingerprint registration operation. In the fifth embodiment, the processing performed by the processing unit 212 (i.e., processing based on confidence level) will be mainly described. Other aspects of the fifth embodiment may be the same as those of the second to fourth embodiments.

[0075] Fingerprint classification is often performed by individuals with specialized knowledge, such as fingerprint analysts. Therefore, an organization lacking such expertise often requests another organization with specialized knowledge to classify newly acquired fingerprint images. In this case, the organization may be unable to register the newly acquired fingerprint images in its fingerprint database until the other organization completes the classification process. Therefore, the organization may perform fingerprint classification itself using the method described below.

[0076] The information processing device 2 may perform the following operations, for example, to assist in the fingerprint registration process. Here, it is assumed that the information processing device 2 is installed in the organization described above.

[0077] The output unit 211 of the arithmetic unit 21 acquires a fingerprint image as a newly acquired fingerprint image. The output unit 211 inputs the fingerprint image into a learning model to obtain a confidence score for the fingerprint image. In this case, the output unit 211 may obtain multiple confidence scores for the fingerprint image, each corresponding to a plurality of pattern types. The output unit 211 transmits a signal indicating the confidence score for the fingerprint image to the processing unit 212 of the arithmetic unit 21.

[0078] The processing unit 212 compares the confidence level of one fingerprint image with a first predetermined value (see second embodiment). Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image.

[0079] If, for a single fingerprint image, multiple confidence levels corresponding to multiple pattern types include a confidence level higher than a first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to the confidence level higher than the first predetermined value. If multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types.

[0080] If none of the multiple confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0081] The processing unit 212 may transmit a signal to the output device 25 indicating the pattern type associated with a fingerprint image. In other words, the processing unit 212 may transmit a signal to the output device 25 indicating the estimated pattern type. As a result, at least one of the characters and image indicating the pattern type associated with a fingerprint image may be displayed.

[0082] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 8. In Figure 8, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image as a newly acquired fingerprint image (step S101). The output unit 211 inputs the fingerprint image into the learning model to obtain the confidence score for the fingerprint image. The output unit 211 outputs the confidence score for the fingerprint image (step S102).

[0083] The processing unit 212 of the arithmetic unit 21 compares each of the multiple confidence values ​​corresponding to each of the multiple pattern types with a first predetermined value based on the confidence value for one fingerprint image (step S401). Based on the comparison result, the processing unit 212 determines whether or not each of the multiple confidence values ​​corresponding to each of the multiple pattern types contains a confidence value higher than the first predetermined value (step S402).

[0084] In the process of step S402, if it is determined that a confidence level higher than the first predetermined value is included (step S402: Yes), the processing unit 212 estimates that the fingerprint pattern type shown by one fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value (step S403). In this case, the processing unit 212 associates one fingerprint image with a pattern type corresponding to a confidence level higher than the first predetermined value. If multiple confidence levels associated with each of multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate one fingerprint image with the multiple pattern types.

[0085] In the process of step S402, if it is determined that no confidence level higher than the first predetermined value is included (step S402: No), the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern (step S404). In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0086] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0087] According to the fifth embodiment, the organization described above can perform the fingerprint registration process relatively quickly by, for example, referring to the pattern type associated with the fingerprint image by the information processing device 2, without having to request the classification of fingerprints from another organization. Because the fingerprint registration process can be performed relatively quickly, for example, a comparison between newly registered fingerprints and previously registered fingerprints can be performed relatively quickly. For example, if various information relating to the individual corresponding to a previously registered fingerprint is linked to that fingerprint, and a previously registered fingerprint that matches a newly registered fingerprint is found during fingerprint matching, the various information relating to the individual corresponding to the newly registered fingerprint can be obtained relatively quickly.

[0088] (modified version) After step S403 or S404, the processing unit 212 may, for example, associate a fingerprint image with a pattern type associated with that fingerprint image and register them in the fingerprint database.

[0089] <Sixth Embodiment> A sixth embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 9. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the sixth embodiment will be described using the information processing device 2. Here, an example will be given in which the information processing device 2 is applied to a fingerprint registration operation. In the sixth embodiment, the processing performed by the processing unit 212 (i.e., processing based on confidence level) will be mainly described. Other aspects of the sixth embodiment may be the same as those of the second to fifth embodiments.

[0090] In the case of residual fingerprints, the ridges may be unclear, only a portion of the fingerprint may remain, or noise may be superimposed on the fingerprint. To properly perform fingerprint matching for residual fingerprints, the matching range may be limited based on a central axis indicating the center position of the fingerprint. The central axis may be set not only for residual fingerprints but for all fingerprints. The central axis may be set, for example, when a newly acquired fingerprint is registered.

[0091] The "central axis" is an axis that passes through the center of the fingerprint (which may also be called the center point) and extends in a specific direction. This specific direction (i.e., the direction in which the central axis extends) is the direction of the fingertip in the case of arched fingerprints, and the direction of the core horseshoe line in the case of fingerprints other than arched fingerprints. The "core horseshoe line" refers to the innermost horseshoe-shaped ridge of the fingerprint. The center of the fingerprint may correspond to the tip of the horseshoe represented by the core horseshoe line. The "direction of the core horseshoe line" refers to the anterior-posterior direction of the horseshoe represented by the core horseshoe line. The direction of the core horseshoe line often differs depending on the type of fingerprint. Therefore, the direction in which the central axis extends often differs depending on the type of fingerprint.

[0092] The information processing device 2 may perform the following operations, for example, to assist in the fingerprint registration process.

[0093] The output unit 211 of the arithmetic unit 21 acquires a fingerprint image as a newly acquired fingerprint image. The output unit 211 inputs the fingerprint image into a learning model to obtain a confidence score for the fingerprint image. In this case, the output unit 211 may obtain multiple confidence scores for the fingerprint image, each corresponding to a plurality of pattern types. The output unit 211 transmits a signal indicating the confidence score for the fingerprint image to the processing unit 212 of the arithmetic unit 21.

[0094] The processing unit 212 compares the confidence level of one fingerprint image with a first predetermined value (see second embodiment). Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image.

[0095] If, for a single fingerprint image, multiple confidence levels corresponding to multiple pattern types include a confidence level higher than a first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to the confidence level higher than the first predetermined value. If multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types.

[0096] If none of the multiple confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0097] The processing unit 212 sets a central axis based on the pattern type associated with a fingerprint image and the fingerprint image itself. If multiple pattern types are associated with a fingerprint image, the processing unit 212 may set multiple central axes, each corresponding to one of the multiple pattern types. In other words, the processing unit 212 may set one central axis for each pattern type associated with a fingerprint image. However, if a fingerprint image is associated with an incomplete pattern, the processing unit 212 does not need to set a central axis.

[0098] If a single fingerprint image is associated with a curved pattern, the processing unit 212 may set a central axis extending in the direction of the fingertip. If a single fingerprint image is associated with a pattern other than a curved pattern, the processing unit 212 may set a central axis extending in the direction of the core hoofline. Various existing methods can be applied to determine the direction of the fingertip and the direction of the core hoofline from the fingerprint shown in the single fingerprint image. Therefore, a detailed explanation of these methods will be omitted.

[0099] The processing unit 212 may transmit a signal to the output device 25 indicating the pattern type associated with a fingerprint image and the central axis corresponding to that pattern type. As a result, at least one of the characters and / or image indicating the pattern type associated with a fingerprint image and the central axis corresponding to that pattern type may be displayed.

[0100] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 9. In Figure 9, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image as a newly acquired fingerprint image (step S101). The output unit 211 inputs the fingerprint image into the learning model to obtain the confidence score for the fingerprint image. The output unit 211 outputs the confidence score for the fingerprint image (step S102).

[0101] The processing unit 212 of the arithmetic unit 21 compares each of the multiple confidence levels corresponding to each of the multiple pattern types with a first predetermined value based on the confidence level for one fingerprint image. Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image (step S501).

[0102] In the processing of step S501, if, for a single fingerprint image, the confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image is a pattern type corresponding to a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to a confidence level higher than the first predetermined value. If the multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types. If the multiple confidence levels corresponding to each of the multiple pattern types do not include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by the single fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate the single fingerprint image with an incomplete pattern as a pattern type.

[0103] Next, the processing unit 212 sets a central axis based on the pattern type associated with a fingerprint image and the fingerprint image itself (step S502). If multiple pattern types are associated with a single fingerprint image, the processing unit 212 may set multiple central axes corresponding to each of the multiple pattern types in step S502.

[0104] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0105] In the information processing device 2, a central axis is set for each pattern type associated with a single fingerprint image. A person registering a fingerprint can register a central axis for each pattern type for a single fingerprint image by referring to the central axis set in the information processing device 2. If a fingerprint shown by a single fingerprint image can be interpreted as having multiple pattern types, multiple central axes may be registered for that single fingerprint image. According to the sixth embodiment, for example, the fingerprint registration process can be supported. For example, if a fingerprint shown by a single fingerprint image can be interpreted as having multiple pattern types, registering multiple central axes for that single fingerprint image can suppress the occurrence of matching errors in fingerprint matching using a single fingerprint image.

[0106] (modified version) Instead of outputting a signal indicating a pattern type associated with a fingerprint image and a central axis corresponding to that pattern type, the processing unit 212 may register a fingerprint image, a pattern type associated with that fingerprint image, and a central axis corresponding to that pattern type. In this case, the processing unit 212 may perform fingerprint matching on the fingerprint image based on the registered central axis. If multiple central axes are registered for a fingerprint image, the processing unit 212 may perform fingerprint matching on the fingerprint image based on each of the multiple central axes.

[0107] <Seventh Embodiment> A seventh embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 10. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the seventh embodiment will be described using the information processing device 2. Here, an example will be given in which the information processing device 2 is applied to fingerprint registration and editing operations. In the seventh embodiment, the processing performed by the processing unit 212 (i.e., processing based on confidence level) will be mainly described. Other aspects of the seventh embodiment may be the same as those of the second to sixth embodiments.

[0108] For example, in fingerprint databases managed by public institutions, fingerprint data may be registered through the following procedure: A person with specialized knowledge, such as a fingerprint expert, determines the type of fingerprint pattern shown in a single fingerprint image. The single fingerprint image and the determined pattern type are then registered as fingerprint data associated with that single fingerprint image.

[0109] The fingerprint database allows for editing of registered fingerprints. Therefore, when a new fingerprint image is registered, initially only that fingerprint image may be registered in the fingerprint database. Subsequently, when the type of fingerprint pattern shown by the fingerprint image is determined, the determined pattern type may be added (registered) by editing the fingerprint data related to the fingerprint image.

[0110] The information processing device 2 may perform the following operations to support, for example, at least one of the fingerprint registration and editing operations. Here, it is assumed that a fingerprint database is built in the storage device 22 of the information processing device 2.

[0111] The output unit 211 of the arithmetic unit 21 acquires a fingerprint image as a newly acquired fingerprint image. The output unit 211 inputs the fingerprint image into a learning model to obtain a confidence score for the fingerprint image. In this case, the output unit 211 may obtain multiple confidence scores for the fingerprint image, each corresponding to a plurality of pattern types. The output unit 211 transmits a signal indicating the confidence score for the fingerprint image to the processing unit 212 of the arithmetic unit 21.

[0112] The processing unit 212 compares the confidence level of one fingerprint image with a first predetermined value (see second embodiment). Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image.

[0113] If, for a single fingerprint image, multiple confidence levels corresponding to multiple pattern types include a confidence level higher than a first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to the confidence level higher than the first predetermined value. If multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types.

[0114] If none of the multiple confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0115] For example, if a pattern type related to a fingerprint image is registered or edited via the input device 24 (in other words, if a user of the information processing device 2 registers or edits a pattern type related to a fingerprint image), the processing unit 212 determines whether the registered or edited pattern type is the same as the pattern type that the processing unit 212 has associated with the fingerprint image. If the registered or edited pattern type is different from the pattern type that the processing unit 212 has associated with the fingerprint image, the processing unit 212 will, for example, issue a warning prompting the user to reconfirm the pattern type. Furthermore, if the input device 24 receives information indicating the registration or editing of a pattern type (for example, information indicating that a button indicating "Register" or "Update" has been pressed), the processing unit 212 may determine that the pattern type has been registered or edited.

[0116] Furthermore, if the registered or added pattern type differs from the pattern type associated with a fingerprint image by the processing unit 212, the processing unit 212 may determine whether the confidence level for the pattern type associated with the fingerprint image is higher than a second predetermined value (see fourth embodiment). If the confidence level is higher than the second predetermined value, the processing unit 212 may, for example, issue a warning prompting the user to reconfirm the pattern type. On the other hand, if the confidence level is lower than the second predetermined value, the processing unit 212 does not need to issue a warning. Furthermore, if the confidence level is equal to the second predetermined value, it may be treated as being included in either case.

[0117] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 10. In Figure 10, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image (step S101). The output unit 211 inputs the fingerprint image into a learning model to obtain the confidence score for the fingerprint image. The output unit 211 outputs the confidence score for the fingerprint image (step S102).

[0118] The processing unit 212 of the arithmetic unit 21 compares each of the multiple confidence levels corresponding to each of the multiple pattern types with a first predetermined value based on the confidence level for one fingerprint image. Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image (step S601).

[0119] In the processing of step S601, if, for a single fingerprint image, the confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image is a pattern type corresponding to a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to a confidence level higher than the first predetermined value. If the multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types. If the multiple confidence levels corresponding to each of the multiple pattern types do not include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by the single fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate the single fingerprint image with an incomplete pattern as a pattern type.

[0120] The processing unit 212 determines whether the pattern type associated with a fingerprint image has been registered or edited (step S602). If it is determined in step S602 that the pattern type has not been registered or edited (step S602: No), the processing unit 212 repeats the process in step S602. In other words, the processing unit 212 may remain in a waiting state until the pattern type is registered or edited.

[0121] If, in step S602, it is determined that a pattern type has been registered or edited (step S602: Yes), the processing unit 212 determines whether the registered or edited pattern type is the same as the pattern type that the processing unit 212 has associated with a fingerprint image (step S603). If, in step S603, it is determined that the registered or edited pattern type is the same as the pattern type that the processing unit 212 has associated with a fingerprint image (step S603: Yes), the operation shown in Figure 10 is terminated.

[0122] If, during the processing in step S603, it is determined that the registered or edited pattern type and the pattern type associated with a single fingerprint image by the processing unit 212 are not the same (step S603: No), the processing unit 212 will, for example, issue a warning prompting the user to reconfirm the pattern type (step S604).

[0123] Furthermore, in the process of step S604, the processing unit 212 may determine whether the confidence level corresponding to the pattern type associated with a fingerprint image is higher than a second predetermined value. If the confidence level is higher than the second predetermined value, the processing unit 212 may, for example, issue a warning prompting the user to reconfirm the pattern type. On the other hand, if the confidence level is lower than the second predetermined value, the processing unit 212 does not need to issue a warning.

[0124] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0125] According to the seventh embodiment, for example, a warning prompting the user to reconfirm the pattern type is issued, thereby suppressing the occurrence of errors in registering the pattern type when registering or editing fingerprint data.

[0126] <Eighth Embodiment> An eighth embodiment of the fingerprint information processing device, fingerprint information processing method, and recording medium will be described with reference to Figures 2 and 11. Hereinafter, the fingerprint information processing device, fingerprint information processing method, and recording medium according to the eighth embodiment will be described using the information processing device 2. Here, an example will be given in which the information processing device 2 is applied to fingerprint registration and editing operations. In the eighth embodiment, the processing performed by the processing unit 212 (i.e., processing based on confidence level) will be mainly described. Other aspects of the eighth embodiment may be the same as those of the second to seventh embodiments.

[0127] For example, in a fingerprint database managed by a public institution, fingerprint data may be registered using the following procedure: For instance, a person with specialized knowledge, such as a fingerprint expert, determines the type of fingerprint pattern shown in a single fingerprint image. A person different from the one who determined the pattern type determines the central axis of the fingerprint shown in the same fingerprint image. The single fingerprint image, the determined pattern type, and the determined central axis are then registered as fingerprint data related to the single fingerprint image.

[0128] The fingerprint database allows for editing of registered fingerprints. Therefore, when a new fingerprint image is registered, initially only that fingerprint image may be registered in the fingerprint database. Subsequently, when the type of fingerprint pattern shown by the fingerprint image is determined, the determined pattern type may be added (registered) by editing the fingerprint data related to the fingerprint image. Similarly, when the central axis of the fingerprint shown by the fingerprint image is determined, the determined central axis may be added (registered) by editing the fingerprint data related to the fingerprint image.

[0129] As described in the sixth embodiment, the direction in which the central axis extends often differs depending on the pattern type. When a fingerprint shown by a single fingerprint image can be interpreted as having multiple pattern types, multiple central axes corresponding to each of the multiple pattern types may be registered for that single fingerprint image. As mentioned above, the person who determines the pattern type and the person who determines the central axis may be different, so for example, a central axis that does not correspond to one of the multiple pattern types may be associated and registered. In that case, there is an error in the central axis associated with one pattern type, and fingerprint matching may not be performed properly for that single fingerprint image.

[0130] For example, if two central axes are registered for a single fingerprint image, it is conceivable that, without considering the correspondence between the pattern type and the central axes, when matching a single fingerprint image with a matching target limited based on a pattern type associated with that fingerprint image, both fingerprint matching within the matching range limited by one of the two central axes and fingerprint matching within the matching range limited by the other central axis could be performed. With such a configuration, fingerprint matching can be performed appropriately for a single fingerprint image. However, this would increase the processing load related to fingerprint matching.

[0131] The information processing device 2 may perform the following operations to support, for example, at least one of the fingerprint registration and editing operations. Here, it is assumed that a fingerprint database is built in the storage device 22 of the information processing device 2.

[0132] The output unit 211 of the arithmetic unit 21 acquires a single fingerprint image. The output unit 211 inputs the single fingerprint image into a learning model to acquire a confidence score for the single fingerprint image. In this case, the output unit 211 may acquire multiple confidence scores for the single fingerprint image, each corresponding to a plurality of pattern types. The output unit 211 transmits a signal indicating the confidence score for the single fingerprint image to the processing unit 212 of the arithmetic unit 21.

[0133] The processing unit 212 compares the confidence level of one fingerprint image with a first predetermined value (see second embodiment). Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image.

[0134] If, for a single fingerprint image, multiple confidence levels corresponding to multiple pattern types include a confidence level higher than a first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image corresponds to a pattern type with a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to the confidence level higher than the first predetermined value. If multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types.

[0135] If none of the multiple confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by one fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate one fingerprint image with an incomplete pattern as a pattern type.

[0136] The processing unit 212 sets a central axis based on the pattern type associated with a fingerprint image and the fingerprint image itself. If multiple pattern types are associated with a single fingerprint image, the processing unit 212 may set multiple central axes, each corresponding to one of the multiple pattern types. In other words, the processing unit 212 may set one central axis for each pattern type associated with a single fingerprint image.

[0137] For example, if multiple pattern types and multiple central axes are registered or edited for a single fingerprint image via the input device 24 (in other words, if a user of the information processing device 2 registers or edits multiple pattern types and multiple central axes for a single fingerprint image), the processing unit 212 determines whether the multiple central axes associated with each of the multiple pattern types are correct. In this case, the processing unit 212 may, for example, compare the central axis associated with one pattern type with the central axis set by the processing unit 212 for that pattern type. Based on the comparison result, the processing unit 212 may determine whether the multiple central axes associated with each of the multiple pattern types are correct. If it is determined that the central axis associated with at least one of the multiple pattern types is incorrect, the processing unit 212 may, for example, issue a warning prompting the user to reconfirm the central axes.

[0138] Furthermore, it is assumed that the pattern type associated with a registered or edited fingerprint image and the pattern type associated with a fingerprint image by the processing unit 212 based on the confidence level are the same. If, however, the pattern type associated with a registered or edited fingerprint image and the pattern type associated with a fingerprint image by the processing unit 212 based on the confidence level are different, the processing unit 212 may, for example, issue a warning prompting the user to reconfirm the pattern type, as described in the seventh embodiment.

[0139] The operation of the information processing device 2 will be explained with reference to the flowchart in Figure 11. In Figure 11, the output unit 211 of the arithmetic unit 21 acquires a fingerprint image (step S101). The output unit 211 inputs the fingerprint image into a learning model to obtain the confidence score for the fingerprint image. The output unit 211 outputs the confidence score for the fingerprint image (step S102).

[0140] The processing unit 212 of the arithmetic unit 21 compares each of the multiple confidence levels corresponding to each of the multiple pattern types with a first predetermined value based on the confidence level for one fingerprint image. Based on the comparison result between each of the multiple confidence levels corresponding to each of the multiple pattern types and the first predetermined value, the processing unit 212 estimates the pattern type of the fingerprint shown by one fingerprint image (step S701).

[0141] In the processing of step S701, if, for a single fingerprint image, the confidence levels corresponding to each of the multiple pattern types include a confidence level higher than the first predetermined value, the processing unit 212 estimates that the fingerprint pattern type shown by the single fingerprint image is a pattern type corresponding to a confidence level higher than the first predetermined value. In this case, the processing unit 212 associates the single fingerprint image with the pattern type corresponding to a confidence level higher than the first predetermined value. If the multiple confidence levels associated with each of the multiple pattern types are higher than the first predetermined value, the processing unit 212 may associate the single fingerprint image with the multiple pattern types. If the multiple confidence levels corresponding to each of the multiple pattern types do not include a confidence level higher than the first predetermined value, the processing unit 212 may estimate that the fingerprint pattern type shown by the single fingerprint image is an incomplete pattern. In this case, the processing unit 212 may associate the single fingerprint image with an incomplete pattern as a pattern type.

[0142] Next, the processing unit 212 sets a central axis based on the pattern type associated with a fingerprint image and the fingerprint image itself (step S702). If multiple pattern types are associated with a single fingerprint image, the processing unit 212 may set multiple central axes corresponding to each of the multiple pattern types in step S702.

[0143] The processing unit 212 determines whether at least one of the pattern type and the central axis has been registered or edited for a single fingerprint image (step S703). If, in the process of step S703, it is determined that neither the pattern type nor the central axis has been registered or edited (step S703: No), the processing unit 212 repeats the process of step S703. In other words, the processing unit 212 may remain in a waiting state until at least one of the pattern type and the central axis has been registered or edited.

[0144] If, in step S703, it is determined that at least one of the pattern type and the central axis has been registered or edited (step S703: Yes), the processing unit 212 determines whether there are two or more pattern types associated with a single fingerprint image (step S704). If, in step S704, it is determined that there are not two or more pattern types (step S704: No), the operation shown in Figure 11 is terminated.

[0145] In step S704, if it is determined that there are two or more pattern types (step S704: Yes), the processing unit 212 determines whether the multiple central axes associated with each of the multiple pattern types are correct (step S705). In step S705, if it is determined that the multiple central axes associated with each of the multiple pattern types are correct (step S705: Yes), the operation shown in Figure 11 is terminated.

[0146] If, during the process in step S705, it is determined that the central axis associated with at least one of the multiple pattern types is incorrect (step S705: No), the processing unit 212 issues a warning prompting the user to reconfirm the central axis (step S706).

[0147] The above-described operation may be achieved by the information processing device 2 reading a computer program recorded on a recording medium. In this case, it can be said that the recording medium contains a computer program that causes the information processing device 2 to perform the above-described operation.

[0148] According to the eighth embodiment, for example, a warning is issued prompting the user to reconfirm the central axis associated with the pattern type, thereby suppressing the occurrence of errors in registering the central axis when registering or editing fingerprint data. In the fingerprint database, the central axis is associated with the pattern type. For example, if two pattern types and two central axes associated with each of the two pattern types are registered for a single fingerprint image, fingerprint matching for that single fingerprint image is performed as follows: When matching a single fingerprint image with a matching target limited based on one of the two pattern types, the matching range is limited by the central axis associated with the one pattern type before fingerprint matching is performed. Also, when matching a single fingerprint image with a matching target limited based on the other of the two pattern types, the matching range is limited by the central axis associated with the other pattern type before fingerprint matching is performed. Therefore, fingerprint matching can be performed appropriately for a single fingerprint image without increasing the processing load related to fingerprint matching.

[0149] (modified version) In the process of step 706 described above, the processing unit 212 may, instead of or in addition to issuing a warning prompting the user to reconfirm the central axis, link together the multiple pattern types associated with a single fingerprint image in the process of step S701 described above and the multiple central axes corresponding to each of the multiple pattern types set in the process of step S702 described above, and register them in the fingerprint database.

[0150] <Note> The following additional information is disclosed regarding the embodiments described above.

[0151] (Note 1) An output means that outputs a confidence score, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types, using a learning model constructed by machine learning using a fingerprint image and training data including sample images showing fingerprints, Processing means for performing processing based on the aforementioned confidence level, A fingerprint information processing device equipped with the following features.

[0152] (Note 2) The output means outputs the confidence score by combining the output result of the first model when the fingerprint image is input to the first model as a learning model and the output result of the second model when the fingerprint image is input to the second model as a learning model. The first model and the second model differ from each other in their output tendencies in response to input. The fingerprint information processing device described in Appendix 1.

[0153] (Note 3) The output means outputs the confidence score using a fingerprint image which is already registered and the learning model. The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the first fingerprint image is estimated. If the estimated pattern type differs from the pattern type already associated with the first fingerprint image, at least one of the following is performed: notification and / or updating the pattern type already associated with the first fingerprint image. A fingerprint information processing device as described in Appendix 1 or 2.

[0154] (Note 4) The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the fingerprint shown in the aforementioned fingerprint image corresponds to two or more of the aforementioned pattern types, then multiple central axes corresponding to each of the two or more pattern types are set. A fingerprint information processing device as described in any of Appendix 1 to 3.

[0155] (Note 5) If the two or more pattern types include an arched pattern and one pattern type different from the arched pattern, the processing means sets a central axis corresponding to the arched pattern that extends in the direction of the fingertip of the fingerprint shown in the fingerprint image, and sets a central axis corresponding to the one pattern type that extends in the direction of the core horseshoe line of the fingerprint shown in the fingerprint image. The fingerprint information processing device described in Appendix 4.

[0156] (Note 6) The processing means performs fingerprint matching on the fingerprint image using each of the multiple central axes corresponding to the two or more pattern types. A fingerprint information processing device as described in Appendix 4 or 5.

[0157] (Note 7) The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the pattern type entered by the user for the fingerprint shown in the aforementioned fingerprint image differs from the estimated pattern type, notification is issued. A fingerprint information processing device as described in any of Appendix 1 to 6.

[0158] (Note 8) The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the fingerprint shown by the fingerprint image corresponds to two or more of the multiple pattern types, then multiple central axes corresponding to each of the two or more pattern types are set. Notification is given if the correspondence between the two or more pattern types and the set number of central axes differs from the correspondence between the pattern type and central axis entered by the user for the fingerprint shown in the fingerprint image. A fingerprint information processing device as described in any of Appendix 1 to 7.

[0159] (Note 9) Using a learning model constructed by machine learning with a fingerprint image and training data including sample images showing fingerprints, a confidence score is output, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. The process is executed based on the aforementioned confidence level. Fingerprint information processing method.

[0160] (Note 10) On the computer, Using a learning model constructed by machine learning with a fingerprint image and training data including sample images showing fingerprints, a confidence score is output, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. The process is executed based on the aforementioned confidence level. A recording medium on which a computer program for executing a fingerprint information processing method is stored.

[0161] This disclosure is not limited to the embodiments described above. For example, if pattern classification is possible for palm prints, this disclosure may be applied to palm prints in addition to fingerprints. This disclosure may be modified as appropriate, as long as it does not contradict the gist or idea of ​​the invention as can be inferred from the claims and the specification as a whole. Fingerprint information processing devices, fingerprint information processing methods and recording media with such modifications are also within the technical scope of this disclosure.

[0162] To the extent permitted by law, this application claims priority based on Japanese Patent Application No. 2022-120344, filed on 28 July 2022, and incorporates all disclosures thereof. To the extent permitted by law, all published gazettes and articles described herein are also incorporated herein. [Explanation of symbols]

[0163] 1, 2 Information Processing Devices 11, 211 Output section 12,212 Processing Unit 21 Arithmetic unit 22 Storage device 23 Communication equipment 24 Input Devices 25 Output device

Claims

1. An output means that outputs a confidence score, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types, using a learning model constructed by machine learning using a fingerprint image and training data including sample images showing fingerprints, Processing means for performing processing based on the aforementioned confidence level, Equipped with, The output means outputs the confidence score using a fingerprint image which is already registered and the learning model. The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the first fingerprint image is estimated. If the estimated pattern type differs from the pattern type already associated with the first fingerprint image, at least one of the following is performed: notification and / or updating the pattern type already associated with the first fingerprint image. Fingerprint information processing device.

2. The output means outputs the confidence score by combining the output result of the first model when the fingerprint image is input to the first model as a learning model and the output result of the second model when the fingerprint image is input to the second model as a learning model. The first model and the second model differ from each other in their output tendencies in response to input. The fingerprint information processing device according to claim 1.

3. The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the fingerprint shown in the aforementioned fingerprint image corresponds to two or more of the aforementioned pattern types, then multiple central axes corresponding to each of the two or more pattern types are set. The fingerprint information processing device according to claim 1.

4. If the two or more pattern types include an arched pattern and one pattern type different from the arched pattern, the processing means sets a central axis corresponding to the arched pattern that extends in the direction of the fingertip of the fingerprint shown in the fingerprint image, and sets a central axis corresponding to the one pattern type that extends in the direction of the core horseshoe line of the fingerprint shown in the fingerprint image. The fingerprint information processing device according to claim 3.

5. The processing means performs fingerprint matching on the fingerprint image using each of the multiple central axes corresponding to the two or more pattern types. The fingerprint information processing device according to claim 3.

6. The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the pattern type entered by the user for the fingerprint shown in the aforementioned fingerprint image differs from the estimated pattern type, notification is issued. The fingerprint information processing device according to claim 1.

7. The processing means, as the processing, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the fingerprint image is estimated. If the fingerprint shown by the fingerprint image corresponds to two or more of the multiple pattern types, then a plurality of central axes corresponding to each of the two or more pattern types are set. Notification is given if the correspondence between the two or more pattern types and the set number of central axes differs from the correspondence between the pattern type and central axis entered by the user for the fingerprint shown in the fingerprint image. The fingerprint information processing device according to claim 1.

8. The computer uses a fingerprint image and a learning model constructed by machine learning using training data including sample images showing fingerprints to output a confidence score, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. The computer performs processing based on the confidence level. A method for processing fingerprint information, The computer outputs the confidence score using a fingerprint image that has already been registered and the learning model. The computer, as part of the process, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the first fingerprint image is estimated. If the estimated pattern type differs from the pattern type already associated with the first fingerprint image, at least one of the following is performed: notification and / or updating the pattern type already associated with the first fingerprint image. Fingerprint information processing method.

9. On the computer, Using a learning model constructed by machine learning with a fingerprint image and training data including sample images showing fingerprints, a confidence score is output, which is an index indicating the likelihood that the fingerprint shown in the fingerprint image belongs to at least one of a plurality of pattern types. The process is executed based on the aforementioned confidence level. A method for processing fingerprint information, Using the aforementioned fingerprint image, which is one already registered fingerprint image, and the learning model, the confidence level is output. As the aforementioned process, Based on the aforementioned confidence level, the type of fingerprint pattern shown by the first fingerprint image is estimated. If the estimated pattern type differs from the pattern type already associated with the first fingerprint image, at least one of the following is performed: notification and / or updating the pattern type already associated with the first fingerprint image. A recording medium on which a computer program for executing a fingerprint information processing method is stored.

Images