WO2023273616A1 - Image recognition method and apparatus, electronic device, storage medium - Google Patents

Abstract

The present application provides an image recognition method and apparatus, an electronic device, and a storage medium. The method comprises: obtaining an image to be recognized of an object to be recognized; obtaining recognition information according to said image and a template image, the recognition information comprising a similarity score between the image to be recognized and the template image, and the number of matching points and/or an overlap area between the image to be recognized and the template image; and obtaining a recognition result according to the recognition information. The solution provided by the present application improves the recognition accuracy of an image to be recognized such as a fingerprint or palmprint image.

Description

Image recognition method and device, electronic device, storage medium

Cross References to Related Applications

This application claims priority to a Chinese patent application with application number 202110739286.5 entitled "Image recognition method and device, electronic equipment, storage medium" filed with the State Intellectual Property Office of China on June 30, 2021, the entire contents of which are incorporated by reference incorporated in this application.

technical field

The present application relates to the technical field of image processing, and in particular to an image recognition method and device, electronic equipment, and a computer-readable storage medium.

Background technique

As a unique, permanent, and stable biometric feature, fingerprint recognition is widely used in occasions that require biometric authentication, such as unlocking mobile phones, access control attendance, and criminal investigation. The final step of fingerprint recognition is generally to send some features (such as the aligned unlock map and template map) into a binary classification model to obtain a binary classification score that measures the similarity between two fingerprints, and compare the score with a specific threshold. Compare to determine whether the two fingerprints come from the same finger.

However, the scores of related binary classification models often deviate from the actual situation in a few cases. It is often not the most reasonable to judge whether the fingerprint comparison is passed through only one threshold. The threshold directly determines the fingerprint comparison result, and the threshold They are all set based on experience, so the accuracy of the fingerprint comparison result is not high. .

Contents of the invention

The embodiment of the present application provides an image recognition method to at least improve the accuracy of fingerprint or palmprint comparison.

Some embodiments of the present application provide an image recognition method, which may include:

Obtaining the image to be recognized of the object to be recognized;

According to the image to be recognized and the template image, recognition information is obtained, and the recognition information may include a similarity score between the image to be recognized and the template image, a similarity score between the image to be recognized and the template image Number of matching points and/or overlapping area characterization parameters;

A recognition result is obtained according to the recognition information.

In an embodiment, the obtaining the identification result according to the identification information may include:

Determine the similarity threshold according to the number of matching points and/or overlapping area characterization parameters;

The similarity score is compared with the similarity threshold to obtain a recognition result.

In an embodiment, the determining the similarity threshold according to the number of matching points and/or the overlapping area characterization parameters may include:

The larger the number of matching points, the smaller the similarity threshold.

In an embodiment, determining the similarity threshold according to the number of matching points may include:

Obtaining a similarity threshold corresponding to the range of the number of matching points according to the range of the number of matching points in which the number of matching points is located;

Wherein, a matching point number range corresponds to a similarity threshold, and different matching point number ranges correspond to different similarity thresholds.

In an embodiment, the obtaining identification information according to the image to be identified and the template image may include:

Obtaining a similarity score between the image to be recognized and the template image through a binary classification algorithm according to the image to be recognized and the template image;

Before obtaining the similarity threshold corresponding to the range of the number of matching points according to the range of the number of matching points where the number of matching points is located, the method may further include:

Obtain multiple groups of misrecognized image pairs and the number of matching points of multiple groups of said misrecognized image pairs; the two images of said misrecognized image pairs are from different target objects;

Obtain the similarity scores of multiple groups of misrecognized image pairs according to the two classification algorithms;

According to the number of matching points of multiple groups of the misrecognized image pairs and the similarity scores of the multiple groups of the misrecognized image pairs, corresponding relationships between multiple ranges of matching point numbers and multiple similarity thresholds are obtained.

In one embodiment, according to the number of matching points of multiple groups of the misrecognized image pairs and the similarity scores of multiple groups of the misrecognized image pairs, the correspondence between multiple ranges of the number of matching points and multiple similarity thresholds is obtained Relationships can include:

According to the number of matching points of multiple sets of misrecognized image pairs, the segments are divided into a number of matching point ranges;

For each matching point number range, obtain the maximum value of the similarity score of the misrecognized image pair corresponding to the matching point number range;

A similarity threshold corresponding to the matching point quantity range is obtained according to the maximum value of the similarity score.

In one embodiment, according to the number of matching points of multiple groups of the misrecognized image pairs and the similarity scores of multiple groups of the misrecognized image pairs, the correspondence between multiple ranges of the number of matching points and multiple similarity thresholds is obtained Relationships can include:

Obtain multiple groups of false rejection image pairs and the number of matching points of multiple groups of false rejection image pairs; the two images of the false rejection image pairs are from the same target object; obtain multiple groups of false rejection image pairs according to the binary classification algorithm similarity score;

According to the number of matching points of the multiple groups of the falsely recognized image pairs, the similarity scores of the multiple groups of the falsely recognized image pairs, the number of matching points of the multiple groups of the falsely rejected image pairs, and the similarity of the multiple groups of the falsely rejected image pairs degree score, to obtain the corresponding relationship between multiple ranges of matching points and multiple similarity thresholds.

In an embodiment, the determining the similarity threshold according to the number of matching points and/or the overlapping area may include:

When the overlapping area characterization parameter may be greater than a first threshold or may be less than a second threshold, increase the similarity threshold; wherein the first threshold may be greater than the second threshold;

In an embodiment, the determining the similarity threshold according to the number of matching points and/or the overlapping area characterization parameter may also include:

When the overlapping area characteristic parameter is greater than a third threshold, the similarity threshold increases to a maximum value; the third threshold may be greater than the first threshold.

In an embodiment, the obtaining identification information according to the image to be identified and the template image may include:

According to the multiple biological feature points of the image to be recognized and the multiple template feature points of the template image, perform alignment between the multiple biological feature points and the multiple template feature points to obtain matching point pairs; The pair of matching points is used to determine the number of matching points contained in the image to be recognized;

and / or;

performing alignment processing on the image to be recognized and the template image; calculating an overlapping area characteristic parameter between the image to be recognized after the alignment processing and the template image.

In an embodiment, the comparing the similarity score with the similarity threshold to obtain the recognition result may include:

If the similarity score is greater than the similarity threshold, obtain a recognition result that the image to be recognized and the template image belong to the same object.

In an embodiment, the obtaining the identification result according to the identification information may include:

If the number of matching points is greater than a specified threshold, a recognition result that the image to be recognized and the template image belong to the same object is obtained.

In an embodiment, the object to be recognized may be a fingerprint, palmprint or vein, and the image to be recognized may be a fingerprint image, a palmprint image or a vein image.

Other embodiments of the present application provide an image recognition device, which may include:

An image acquisition module, the image acquisition module can be configured to acquire an image to be identified of an object to be identified;

An information obtaining module, the information obtaining module may be configured to obtain recognition information according to the image to be recognized and the template image, and the recognition information may include a similarity score between the image to be recognized and the template image, the The number of matching points and/or the overlapping area between the image to be recognized and the template image;

A result obtaining module, the result obtaining module may be configured to obtain a recognition result according to the recognition information.

In an embodiment, the result obtaining module can also be configured to:

Determine the similarity threshold according to the number of matching points and/or overlapping area characterization parameters;

The similarity score is compared with the similarity threshold to obtain a recognition result.

In an embodiment, the information obtaining module can also be configured to:

According to the multiple biological feature points of the image to be recognized and the multiple template feature points of the template image, perform alignment between the multiple biological feature points and the multiple template feature points to obtain matching point pairs; The pair of matching points is used to determine the number of matching points contained in the image to be recognized;

and / or;

performing alignment processing on the image to be recognized and the template image; calculating an overlapping area characteristic parameter between the image to be recognized after the alignment processing and the template image.

Still other embodiments of the present application provide an electronic device, and the electronic device may include:

processor;

memory for storing processor-executable instructions;

Wherein, the processor is configured to execute the above image recognition method.

Still other embodiments of the present application provide a computer-readable storage medium, where the storage medium stores a computer program, and the computer program can be executed by a processor to complete the above-mentioned image recognition method.

In the technical solution provided by the above-mentioned embodiments of the present application, the image to be recognized and the template image are judged according to the similarity score between the image to be recognized and the template image, the number of matching points and/or the overlapping area between the image to be recognized and the template image belong to the same object. Compared with setting a similarity threshold arbitrarily according to needs and performing image recognition only according to the similarity score, the solution provided by the present application improves the recognition accuracy.

Description of drawings

In order to illustrate the technical solutions of the embodiments of the present application more clearly, the following briefly introduces the drawings that are used in the embodiments of the present application.

FIG. 1 is a schematic structural diagram of an electronic device provided by an embodiment of the present application;

Fig. 2 is a schematic flow chart of an image recognition method provided by an embodiment of the present application;

Fig. 3 is a detailed flowchart of step S220 in the embodiment corresponding to Fig. 2;

Fig. 4 is a schematic flow chart of the alignment process provided by the embodiment of the present application;

Fig. 5 is a detailed flowchart of step S230 in the embodiment corresponding to Fig. 2;

Fig. 6 is a schematic flow chart of determining similarity thresholds in different ranges provided by an embodiment of the present application;

FIG. 7 is a schematic flowchart of the details of step 603 in the embodiment corresponding to FIG. 6;

Fig. 8 is a schematic flow chart of determining similarity thresholds in different ranges provided by another embodiment of the present application;

Fig. 9 is a schematic diagram of the distribution of the number of matching points and similarity scores of falsely rejected image pairs and misrecognized image pairs provided by the embodiment of the present application;

Fig. 10 is a block diagram of an image recognition device provided by an embodiment of the present application.

detailed description

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

Like numbers and letters denote similar items in the following figures, so that once an item is defined in one figure, it does not require further definition and explanation in subsequent figures. Meanwhile, in the description of the present application, the terms "first", "second" and the like are only used to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

In recent years, artificial intelligence-based computer vision, deep learning, machine learning, image processing, image recognition and other technologies have made important progress. Artificial Intelligence (AI) is an emerging science and technology that studies and develops theories, methods, technologies and application systems for simulating and extending human intelligence. The subject of artificial intelligence is a comprehensive subject that involves many technologies such as chips, big data, cloud computing, Internet of Things, distributed storage, deep learning, machine learning, and neural networks. As an important branch of artificial intelligence, computer vision is specifically to allow machines to recognize the world. Computer vision technology usually includes face recognition, liveness detection, fingerprint recognition and anti-counterfeiting verification, biometric recognition, face detection, pedestrian detection, target detection, pedestrian detection, etc. Recognition, image processing, image recognition, image semantic understanding, image retrieval, text recognition, video processing, video content recognition, behavior recognition, 3D reconstruction, virtual reality, augmented reality, simultaneous localization and map construction (SLAM), computational photography, robotics Navigation and positioning technologies. With the research and progress of artificial intelligence technology, this technology has been applied in many fields, such as security, urban management, traffic management, building management, park management, face access, face attendance, logistics management, warehouse management, robots , smart marketing, computational photography, mobile imaging, cloud services, smart home, wearable devices, unmanned driving, automatic driving, smart medical care, face payment, face unlock, fingerprint unlock, witness verification, smart screen, smart TV, Cameras, mobile Internet, webcasting, beauty, cosmetics, medical beauty, intelligent temperature measurement and other fields.

FIG. 1 is a schematic structural diagram of an electronic device provided by an embodiment of the present application. The electronic device 100 may be used to execute the image recognition method provided in the embodiment of the present application. As shown in FIG. 1 , the electronic device 100 includes: one or more processors 102 , and one or more memories 104 storing processor-executable instructions. Wherein, the processor 102 is configured to execute the image recognition method provided in the following embodiments of the present application.

The processor 102 may be a gateway, or an intelligent terminal, or a device including a central processing unit (CPU), a graphics processing unit (GPU), or other forms of processing units with data processing capabilities and/or instruction execution capabilities , can process data of other components in the electronic device 100, and can also control other components in the electronic device 100 to perform desired functions.

The memory 104 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and/or cache memory (cache). The non-volatile memory may include, for example, a read-only memory (ROM), a hard disk, a flash memory, and the like. One or more computer program instructions can be stored on the computer-readable storage medium, and the processor 102 can execute the program instructions to implement the image recognition method described below. Various application programs and various data, such as various data used and/or generated by the application programs, may also be stored in the computer-readable storage medium.

In one embodiment, the electronic device 100 shown in FIG. 1 may further include an input device 106, an output device 108, and a data acquisition device 110, and these components are interconnected through a bus system 112 and/or other forms of connection mechanisms (not shown). . It should be noted that the components and structure of the electronic device 100 shown in FIG. 1 are exemplary rather than limiting, and the electronic device 100 may also have other components and structures as required.

The input device 106 may be a device used by a user to input instructions, and may include one or more of a keyboard, a mouse, a microphone, and a touch screen. The output device 108 may output various information (eg, images or sounds) to the outside (eg, a user), and may include one or more of a display, a speaker, and the like. The data acquisition device 110 can acquire images of objects and store the acquired images in the memory 104 for use by other components. Exemplarily, the data collection device 110 may be a camera.

In one embodiment, each device in the example electronic device 100 used to implement the image recognition method of the embodiment of the present application can be integrated or distributed, such as the processor 102, the memory 104, the input device 106 and the output device 108 are integrated in one body, and the data collection device 110 is separately set.

In an embodiment, the example electronic device 100 for implementing the image recognition method of the embodiment of the present application may be implemented as a smart terminal such as a smart phone, a tablet computer, a smart watch, or a vehicle-mounted device.

Fig. 2 is a schematic flowchart of an image recognition method provided by an embodiment of the present application. The method can be executed by the above-mentioned electronic device. As shown in FIG. 2 , the method includes the following steps S210-S230.

Step S210: Obtain the image of the object to be recognized.

Wherein, the object to be recognized may be a fingerprint, a palm print or a vein. The image to be recognized can be a fingerprint image, a palm print image or a vein image.

In one embodiment, the image to be recognized may be collected by the biometric image collection device and then sent to the electronic device, and then the electronic device uses the image recognition method provided in this application to compare the received biometric image with the template image stored in advance. Comparison.

In an embodiment, the electronic device may also include a biometric image collection device, so the image to be recognized is directly collected by the electronic device, and the image to be recognized is recognized.

Wherein, recognizing the image is to judge whether the image to be recognized and the template image belong to the same object. For example, the image to be recognized and the template image stored in advance belong to the same finger or palm, and can be unlocked. Therefore, the image to be recognized can also be called the unlocking image, and the template image can also be called the base library image. Wherein, the template image can be regarded as a biometric image of a known identity stored in advance. However, if the image to be recognized and the template image are compared and determined to belong to the same object, the identity of the user of the image to be recognized can be determined.

Step S220: Obtain identification information according to the image to be identified and the template image.

Wherein, the identification information includes the similarity score between the image to be identified and the template image, the number of matching points between the image to be identified and the template image and/or the overlapping area characterization parameters;

In one embodiment, identification information may include similarity score and number of matching points; in another embodiment, identification information may include similarity score and overlapping area characterization parameters; in other embodiments, identification information may include similarity Score, number of matching points, and overlapping area characterize parameters. In other embodiments, the identification information may also only include the number of matching points. The identification information may include one or more of the similarity score, the number of matching points, and the overlapping area characterization parameters.

Among them, the similarity score is used to characterize the similarity between the image to be recognized and the template image. The number of matching points refers to the number of feature point pairs matched between the image to be recognized and the template image obtained during feature point pairing. The overlapping area characterization parameter can be the ratio of the area of the overlapping area to the total area of the image to be recognized, or the area size of the overlapping area. The process of how to obtain the similarity score, the number of matching points and the characteristic parameters of overlapping area will be described in detail below.

In an embodiment, the above step S220 specifically includes: according to the image to be recognized and the template image, obtaining a similarity score between the image to be recognized and the template image through a binary classification algorithm.

The similarity score is used to characterize the similarity between the image to be recognized and the template image, for example, 1 may represent the highest similarity score, and 0 may represent the lowest similarity score. In one embodiment, the image to be recognized may be aligned with the template image, and then the aligned image to be recognized and the template image may be input into a trained binary classification algorithm to obtain a similarity score output by the binary classification algorithm. Binary classification algorithms can be trained using a large number of biometric images with known similarity scores.

In another embodiment, as shown in FIG. 3 , the above step S220 may include the following steps S221 to S222.

Step S221: According to the multiple biological feature points of the image to be recognized and the multiple template feature points of the template image, perform alignment between the multiple biological feature points and the multiple template feature points to obtain matching point pairs .

Among them, the biological feature point refers to the feature point and its descriptor in the image to be recognized. The template feature points refer to the feature points and their descriptors in the template image. The method for extracting image feature points and their descriptors can use ORB method, SIFT method or method based on deep neural network.

As shown in Figure 4, feature points and descriptors corresponding to each point are extracted from the image to be recognized and the template image respectively. Then, according to the feature points and descriptors, the two images are aligned to obtain the matching point pairs, and the brute force matching or RANSAC method can be used. It should be noted that a biological feature point corresponds to a template feature point, and these two feature points can be considered to form a matching point pair.

Step S222: Determine the number of matching points contained in the image to be recognized according to the matching point pairs.

The number of matching points contained in the image to be recognized can be obtained by counting the number of matching point pairs.

In other embodiments, the above step S220 may include the following steps: performing alignment processing on the image to be recognized and the template image; calculating an overlapping area characteristic parameter between the image to be recognized and the template image after the alignment processing.

Among them, the alignment process refers to detecting the feature points of the image to be recognized and the template image, pairing the feature points, and calculating the transformation matrix between the feature points of the image to be recognized and the matching points of the template image, and transforming the image to be recognized according to this The matrix is transformed. The transformation matrix can be a rigid transformation matrix or an affine transformation matrix.

As shown in Figure 4, based on the rigid transformation matrix M, the aligned image to be recognized can be obtained. In other embodiments, alignment may be performed based on feature points and descriptors to obtain aligned images to be recognized. The aligned image is equivalent to a trapezoid. Therefore, the overlapping area between the aligned image to be recognized and the template image can be calculated by calculating the intersection area of two convex quadrilaterals. In one embodiment, the overlapping area can be directly used as the above-mentioned overlapping area characterization parameter. In another embodiment, the overlapping area is divided by the total area of the aligned images to be recognized to obtain the overlapping area ratio, and the overlapping area ratio can be calculated as as the overlapping area characterization parameter.

Step S230: Obtain a recognition result according to the recognition information.

Wherein, the recognition result is used to indicate whether the image to be recognized and the template image belong to the same object.

In one embodiment, when the number of matching points exceeds a specified threshold, it can be directly determined that the image to be recognized and the template image belong to the same object, without calling the binary classification algorithm to calculate the similarity score, thereby reducing the comparison time. When the number of matching points is less than or equal to the specified threshold, then calculate the similarity score between the image to be recognized and the template image to improve the recognition accuracy, and when the similarity score is greater than the similarity threshold, determine that the image to be recognized and the template image belong to the same object.

In an embodiment, as shown in FIG. 5 , the above step S230 specifically includes the following steps S231 to S232.

Step S231: Determine the similarity threshold according to the number of matching points and/or the overlapping area characteristic parameter.

Step S232: Compare the similarity score with the similarity threshold to obtain a recognition result.

Wherein, if the similarity score is greater than or equal to the similarity threshold, a recognition result that the image to be recognized and the template image belong to the same object can be obtained. On the contrary, if the similarity score is less than the similarity threshold, the comparison result that the image to be recognized and the template image do not belong to the same object can be obtained.

It should be noted that if the similarity score is calculated, the similarity threshold is not a fixed value manually set according to experience. The similarity threshold may be determined according to the number of matching points and the overlapping area, or determined solely according to the number of matching points, or determined solely according to the characteristic parameter of the overlapping area.

How to determine the similarity threshold is described in detail below.

The confidence of the similarity score is often positively correlated with the number of matching points. In one embodiment, the larger the number of matching points, the smaller the similarity threshold. Therefore, the larger the number of matching points, the lower the false recognition rate, so the similarity threshold can be appropriately reduced. In some other embodiments, the targeted similarity threshold may also be adjusted according to the overlapping area characterization parameter. In an embodiment, the initial value of the similarity threshold may be determined according to the number of matching points, or may be a preset initial value. Afterwards, the similarity threshold is adjusted according to the overlapping area characterization parameter.

In an embodiment, when the overlapping area characteristic parameter is larger than a first threshold or smaller than a second threshold, the similarity threshold is increased; wherein, the first threshold is larger than the second threshold. In another embodiment, when the overlapping area characteristic parameter is greater than a third threshold, the similarity threshold may be increased to a preset maximum value; the third threshold is greater than the first threshold.

For example, the first threshold may be 0.95, the second threshold may be 0.2, and the third threshold may be 0.98.

When the overlapping area ratio is large (such as greater than 0.95) or small (such as less than 0.2), the confidence of the similarity score decreases, and the binary classification threshold can be appropriately restricted, that is, the similarity threshold can be increased. In particular, if the overlapping area ratio of the image is extremely large (such as greater than 0.98), it is likely that the algorithm fails to align to non-biological texture (such as non-fingerprint) areas such as scratches, cracks, and foreign objects. At this time, the similarity threshold It is often necessary to be stuck to an extreme size to prevent misrecognition in such extreme situations.

In one embodiment, determining the similarity threshold according to the number of matching points specifically includes the following steps: obtaining a similarity threshold corresponding to the range of the number of matching points according to the range of the number of matching points in which the number of matching points is located.

Wherein, a matching point number range corresponds to a similarity threshold, and different matching point number ranges correspond to different similarity thresholds. For example, when the range of the number of matching points is in the first interval (for example, 0-100), the corresponding similarity threshold is 0.57. When the number of matching points falls within the second interval (for example, 100-140), the corresponding similarity threshold is 0.56; when the number of matching points falls within the third interval (for example, 140-200), the corresponding similarity threshold is 0.55. That is, different matching point quantity ranges correspond to different similarity thresholds, and the larger the number range, the smaller the similarity threshold.

For example, according to the number of matching points between the image to be recognized and the template image, the range of the number of matching points corresponding to the number of matching points can be determined. For example, if the number of matching points is in the first interval, the similarity threshold of the first interval can be obtained as 0.57.

In an embodiment, as shown in FIG. 6 , similarity thresholds corresponding to different matching point quantity ranges may be determined in the manner of the following steps S601 to S604 .

Step S601: Obtain multiple sets of misrecognized image pairs and the number of matching points of the multiple sets of misrecognized image pairs; the two images of the misrecognized image pairs are from different target objects.

For example, in a fingerprint identification system, the user is generally required to enter several fingers first, and these entered finger pictures form a "bottom database". In the subsequent unlocking process, the unlocking map and the bottom library map form a "finger pairing". If you use a finger that does not exist in the bottom library to unlock, the pairing of this unlocking map with a certain bottom library is called "attacking finger pairing", and the successful identification of the attacking finger pairing is called "false acceptance" (False Accept, that is, FA).

Therefore, the "misrecognized image pair" can be considered as two images belonging to different fingers or palms. In a recognition system with a preset similarity threshold, two images are considered to belong to the same finger or palm.

Step S602: Obtain the similarity scores of multiple sets of misrecognized image pairs according to the binary classification algorithm.

It should be noted that the similarity score between the image to be recognized and the template image is also calculated by using the binary classification algorithm.

Step S603: According to the number of matching points of multiple groups of the misrecognized image pairs and the similarity scores of the multiple groups of the misrecognized image pairs, obtain the corresponding relationship between multiple ranges of matching point numbers and multiple similarity thresholds.

In an embodiment, as shown in FIG. 7 , the above step S603 may include steps S701 to S703.

Step S701: According to the number of matching points of multiple groups of the misrecognized image pairs, the segments are divided into multiple ranges of matching point numbers.

The number range of each matching point can be the same or different. If the number of matching points is different and the similarity score changes greatly, the span of each matching point number range can be reduced and multiple matching point number ranges can be set. For example, the number of matching points may range from 0-50, 51-100, 101-150, 151-200 and other intervals.

Step S702: For each range of the number of matching points, obtain the maximum value of the similarity scores of the misrecognized image pairs corresponding to the range of the number of matching points.

Specifically, for each matching point number range, the misrecognized image pairs corresponding to the matching point number range can be found, and the maximum similarity score can be found according to the similarity scores of these misrecognized image pairs. Therefore, for each range of the number of matching points, the maximum value of the corresponding similarity score can be found.

Step S703: Obtain a similarity threshold corresponding to the matching point quantity range according to the maximum value of the similarity score.

For each range of the number of matching points, the similarity threshold corresponding to the range of the number of matching points may be slightly greater than the maximum value of the similarity score corresponding to the range of the number of matching points. Thus, the similarity threshold corresponding to each matching point quantity range can be obtained.

In other embodiments, as shown in FIG. 8 , similarity thresholds corresponding to different matching point quantity ranges may also be determined by the following steps S801 to S803 .

Step S801: Obtain multiple groups of falsely rejected image pairs and the number of matching points of the multiple groups of falsely rejected image pairs; the two images of the falsely rejected image pairs are from the same target object.

Among them, the "falsely rejected image pair" can be considered as two images belonging to the same finger or palm. In a recognition system where two images are at a preset similarity threshold, they are considered to belong to different fingers or palms.

Step S802: Obtain the similarity scores of multiple groups of falsely rejected image pairs according to the binary classification algorithm;

It should be noted that both the similarity score of the falsely recognized image pair and the similarity score of the falsely rejected image pair can be obtained by using the same binary classification algorithm.

Step S803: According to the number of matching points of the multiple groups of the falsely recognized image pairs, the similarity scores of the multiple groups of the falsely recognized image pairs, the number of matching points of the multiple groups of the falsely rejected image pairs, and the multiple groups of the falsely rejected image pairs The similarity score of the pair is obtained to obtain the correspondence between multiple ranges of matching points and multiple similarity thresholds.

Specifically, according to the number of matching points of multiple groups of misrecognized image pairs, it can be segmented into multiple ranges of matching point numbers; for each range of matching point numbers, the similarity of misrecognized image pairs corresponding to the range of matching point numbers is obtained The maximum value of the score, and the similarity score of the falsely rejected image pair corresponding to the range of the number of matching points. For each matching point number range, according to the maximum value of the similarity score of the falsely rejected image pair and the similarity score of the misrecognized image pair corresponding to the matching point number range, the similarity corresponding to the matching point number range is obtained threshold.

Ideally, it is hoped that all misrecognized image pairs will fail to compare, and all falsely rejected image pairs will be compared successfully. The rejected image pair can be compared successfully. Then according to the maximum value of the similarity score of the misrecognized image pair, the maximum value can be appropriately increased as the threshold, so that the similarity score of the misrecognized image pair is less than the threshold, and at the same time, the similarity score of as few falsely rejected image pairs as possible is less than the threshold . Or appropriately lower the maximum value as the threshold, so that the similarity scores of most misrecognized image pairs are less than the threshold, and the similarity scores of as few falsely rejected image pairs as possible are less than the threshold.

If misrecognition will have more serious consequences than false rejection, you can choose the former method, and appropriately increase the maximum value as the threshold according to the maximum value of the similarity score of misrecognized image pairs, so that the similarity scores of all misrecognized image pairs is less than the threshold, and at the same time let the similarity scores of as few falsely rejected image pairs as possible be less than the threshold.

As shown in FIG. 9 , 901 represents the number of matching points and the similarity score of the falsely rejected image pair, and 902 represents the number of matching points and the similarity score of the falsely recognized image pair. Since misrecognition may bring greater security risks, it is possible to control the misrecognition rate while appropriately reducing the similarity threshold, thereby reducing the false rejection rate.

As shown in FIG. 9 , when the similarity score is above 0.5, there are relatively few misrecognized image pairs, so the minimum similarity threshold can be 0.5. The similarity threshold may be greater than the maximum similarity score of the misrecognized image pair, for example greater than 0.57. Confidence according to the similarity score is often positively correlated with the number of matching points, that is, the larger the number of matching points, the lower the false positive rate. Therefore, the larger the number of matching points, the lower the similarity threshold can be.

As shown in FIG. 9 , five matching point quantity ranges can be configured, and when the matching point quantity is in the first quantity range (for example, 0-100), the corresponding similarity threshold is 0.57. When the number of matching points is in the second number section (for example, 100-140), the corresponding similarity threshold is 0.56; when the number of matching points is in the third number section (for example, 140-200), the corresponding similarity threshold is 0.55, when the number of matching points is in the fourth number section (for example, 200-230), the corresponding similarity threshold is 0.53, when the number of matching points is in the fifth number section (for example, 230-330), the corresponding similarity The threshold is 0.5. If the number of matching points is greater than 330, it can be directly considered that the image to be recognized and the template image belong to the same object.

In an embodiment, according to the number of matching points between the image to be recognized and the template image, the range of the number of matching points corresponding to the number of matching points can be determined first, and then the similarity threshold corresponding to the range of the number of matching points can be obtained. Then adjust the similarity threshold according to the overlapping area ratio between the image to be recognized and the template image. If the overlapping area ratio is large or small, the similarity threshold can be appropriately increased. Then, based on the finally determined similarity threshold, it is compared with the similarity score to determine whether the image to be recognized and the template image belong to the same object. Assuming that the database stores a large number of template images, the method provided by the embodiment of the present application can be used to compare the image to be recognized with each template image until a template image that belongs to the same object as the image to be recognized is found.

For example, by calculating the number of matching points between the image to be recognized and the template image, and according to the range of the number of matching points in which the number of matching points is located, the similarity threshold corresponding to the range of the number of matching points can be determined. If the image to be recognized and the template image If the overlapping area ratio is less than or equal to the first threshold (for example, 0.95) and greater than or equal to the second threshold (for example, 0.2), then the similarity threshold is used as the similarity threshold for judging whether the image to be recognized and the template image belong to the same object.

In an embodiment, if the overlapping area ratio is greater than a first threshold (for example, 0.95) or smaller than a second ratio threshold (for example, 0.2), the similarity threshold corresponding to the obtained matching point quantity range may be increased. For example, the similarity threshold is adjusted from 0.57 to 0.8, and the increased similarity threshold is used as the similarity threshold for judging whether the image to be recognized and the template image belong to the same object.

In other embodiments, whether the image to be recognized and the template image belong to the same object may also be determined solely according to the overlapping area characteristic parameter. For example, when the overlapping area characterization parameter is in the range of 0.2-0.95, it is considered whether the image to be recognized and the template image belong to the same object.

In the technical solution provided by the above-mentioned embodiments of the present application, the similarity threshold is determined according to the number of matching points and/or overlapping area between the image to be recognized and the template image, and the similarity score between the image to be recognized and the template image is compared with the similarity threshold to obtain the recognition result of whether the image to be recognized and the template image belong to the same object. Compared with arbitrarily setting a similarity threshold as required, the solution provided by the present application improves the accuracy of recognition of the image to be recognized.

The following is an embodiment of the device of the present application, which can be used to implement the above-mentioned embodiment of the image recognition method of the present application. For the details not disclosed in the device embodiment of the present application, please refer to the embodiment of the image recognition method of the present application.

Fig. 10 is a block diagram of an image recognition device according to an embodiment of the present application. As shown in FIG. 10 , the device includes: an image acquisition module 910 , an information acquisition module 920 , and a result acquisition module 930 .

The image acquisition module 910. The image acquisition module 910 may be configured to acquire an image to be identified of the object to be identified.

An information obtaining module 920, the information obtaining module 920 may be configured to obtain recognition information according to the image to be recognized and the template image, where the recognition information includes a similarity score between the image to be recognized and the template image, The number of matching points and/or the overlapping area between the image to be recognized and the template image.

The result obtaining module 930. The result obtaining module 930 may be configured to obtain a recognition result according to the recognition information.

For the implementation process of the functions and effects of each module in the above-mentioned device, please refer to the implementation process of the corresponding steps in the above-mentioned image recognition method for details, and will not be repeated here.

In the several embodiments provided in this application, the disclosed devices and methods may also be implemented in other ways. The device embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functions and possible implementations of devices, methods and computer program products according to multiple embodiments of the present application. operate. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or portion of code that contains one or more executable instruction. In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions.

In addition, each functional module in each embodiment of the present application may be integrated to form an independent part, each module may exist independently, or two or more modules may be integrated to form an independent part.

If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the essence of the technical solution of the present application or the part that contributes to the related technology or the part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several The instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods in various embodiments of the present application. The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disc, etc., which can store program codes. .

Industrial Applicability

The present application provides an image recognition method and device, electronic equipment, and a storage medium. The method includes: acquiring an image to be recognized of an object to be recognized; obtaining recognition information according to the image to be recognized and a template image; the recognition information includes The similarity score between the image to be recognized and the template image, the number of matching points and/or the overlapping area between the image to be recognized and the template image; obtaining a recognition result according to the recognition information. The scheme provided by the application improves the accuracy of fingerprint or palmprint image recognition waiting to be recognized.

In addition, it can be understood that the image recognition method and device, electronic equipment, and storage medium of the present application are reproducible and can be used in various industrial applications. For example, the image recognition method and device, electronic equipment, and storage medium of the present application may be used in the technical field of image processing.

Claims (18)

1. An image recognition method, characterized in that, comprising:

   Obtaining the image to be recognized of the object to be recognized;

   According to the image to be recognized and the template image, recognition information is obtained, and the recognition information includes a similarity score between the image to be recognized and the template image, a match between the image to be recognized and the template image Number of points and/or overlapping area characterizing parameters;

   A recognition result is obtained according to the recognition information.
2. The method according to claim 1, wherein said obtaining the identification result according to the identification information comprises:

   Determine the similarity threshold according to the number of matching points and/or overlapping area characterization parameters;

   The similarity score is compared with the similarity threshold to obtain a recognition result.
3. The method according to claim 2, wherein said determining the similarity threshold according to the number of matching points and/or the overlapping area characterization parameters comprises:

   The larger the number of matching points, the smaller the similarity threshold.
4. The method according to claim 2 or 3, wherein determining the similarity threshold according to the number of matching points includes:

   Obtaining a similarity threshold corresponding to the range of the number of matching points according to the range of the number of matching points in which the number of matching points is located;

   Wherein, a matching point number range corresponds to a similarity threshold, and different matching point number ranges correspond to different similarity thresholds.
5. The method according to claim 4, wherein the obtaining identification information according to the image to be identified and the template image comprises:

   Obtaining a similarity score between the image to be recognized and the template image through a binary classification algorithm according to the image to be recognized and the template image;

   Before obtaining the similarity threshold corresponding to the range of the number of matching points according to the range of the number of matching points where the number of matching points is located, the method further includes:

   Obtain multiple groups of misrecognized image pairs and the number of matching points of multiple groups of said misrecognized image pairs; the two images of said misrecognized image pairs are from different target objects;

   Obtain the similarity scores of multiple groups of misrecognized image pairs according to the two classification algorithms;

   According to the number of matching points of multiple groups of the misrecognized image pairs and the similarity scores of the multiple groups of the misrecognized image pairs, corresponding relationships between multiple ranges of matching point numbers and multiple similarity thresholds are obtained.
6. The method according to claim 5, characterized in that, according to the number of matching points of multiple sets of misrecognized image pairs and the similarity scores of multiple sets of misrecognized image pairs, the range of the number of multiple matching points and the number of matching points are obtained. Correspondence of multiple similarity thresholds, including:

   According to the number of matching points of multiple sets of misrecognized image pairs, the segments are divided into a number of matching point ranges;

   For each matching point number range, obtain the maximum value of the similarity score of the misrecognized image pair corresponding to the matching point number range;

   A similarity threshold corresponding to the matching point quantity range is obtained according to the maximum value of the similarity score.
7. The method according to claim 5, characterized in that, according to the number of matching points of multiple sets of misrecognized image pairs and the similarity scores of multiple sets of misrecognized image pairs, the range of the number of multiple matching points and the number of matching points are obtained. Correspondence of multiple similarity thresholds, including:

   Obtain multiple groups of false rejection image pairs and the number of matching points of multiple groups of false rejection image pairs; the two images of the false rejection image pairs are from the same target object;

   Obtain the similarity score of multiple groups of wrongly rejected image pairs according to the two classification algorithms;

   According to the number of matching points of the multiple groups of the falsely recognized image pairs, the similarity scores of the multiple groups of the falsely recognized image pairs, the number of matching points of the multiple groups of the falsely rejected image pairs, and the similarity of the multiple groups of the falsely rejected image pairs degree score, to obtain the corresponding relationship between multiple ranges of matching points and multiple similarity thresholds.
8. The method according to any one of claims 2 to 7, wherein said determining the similarity threshold according to the number of matching points and/or the overlapping area comprises:

   When the overlapping area characteristic parameter is larger than a first threshold or smaller than a second threshold, increase the similarity threshold; wherein, the first threshold is larger than the second threshold.
9. The method according to claim 8, wherein said determining the similarity threshold according to the number of matching points and/or the overlapping area characterization parameter further comprises:

   When the overlapping area characteristic parameter is greater than a third threshold, the similarity threshold increases to a maximum value; the third threshold is greater than the first threshold.
10. The method according to any one of claims 1 to 9, wherein the obtaining identification information according to the image to be identified and the template image comprises:

    According to the multiple biological feature points of the image to be recognized and the multiple template feature points of the template image, perform alignment between the multiple biological feature points and the multiple template feature points to obtain matching point pairs; The pair of matching points is used to determine the number of matching points contained in the image to be recognized;

    and / or;

    performing alignment processing on the image to be recognized and the template image; calculating an overlapping area characteristic parameter between the image to be recognized after the alignment processing and the template image.
11. The method according to any one of claims 2 to 9, wherein said comparing said similarity score with said similarity threshold to obtain a recognition result comprises:

    If the similarity score is greater than the similarity threshold, obtain a recognition result that the image to be recognized and the template image belong to the same object.
12. The method according to any one of claims 1 to 11, wherein said obtaining the identification result according to the identification information comprises:

    If the number of matching points is greater than a specified threshold, a recognition result that the image to be recognized and the template image belong to the same object is obtained.
13. The method according to any one of claims 1 to 12, wherein the object to be recognized is a fingerprint, palmprint or vein, and the image to be recognized is a fingerprint image, a palmprint image or a vein image.
14. An image recognition device, characterized in that it comprises:

    an image acquisition module, the image acquisition module is configured to acquire an image to be identified of the object to be identified;

    An information obtaining module, the information obtaining module is configured to obtain recognition information according to the image to be recognized and the template image, the recognition information includes a similarity score between the image to be recognized and the template image, the The number of matching points and/or the overlapping area between the image to be recognized and the template image;

    A result obtaining module configured to obtain a recognition result according to the recognition information.
15. The image recognition device according to claim 14, wherein the result obtaining module is further configured to:

    Determine the similarity threshold according to the number of matching points and/or overlapping area characterization parameters;

    The similarity score is compared with the similarity threshold to obtain a recognition result.
16. The image recognition device according to claim 14 or 15, wherein the information obtaining module is further configured to:

    According to the multiple biological feature points of the image to be recognized and the multiple template feature points of the template image, perform alignment between the multiple biological feature points and the multiple template feature points to obtain matching point pairs; The pair of matching points is used to determine the number of matching points contained in the image to be recognized;

    and / or;

    performing alignment processing on the image to be recognized and the template image; calculating an overlapping area characteristic parameter between the image to be recognized after the alignment processing and the template image.
17. An electronic device, characterized in that the electronic device comprises:

    processor;

    memory for storing instructions to be executed by the processor;

    Wherein, the processor is configured to execute the image recognition method according to any one of claims 1-13.
18. A computer-readable storage medium, characterized in that the storage medium stores a computer program, and the computer program is executed by a processor to complete the image recognition method according to any one of claims 1-13.

Images