CN112148908A

CN112148908A - Image database updating method and device, electronic equipment and medium

Info

Publication number: CN112148908A
Application number: CN202011152127.7A
Authority: CN
Inventors: 温圣召
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2020-10-23
Filing date: 2020-10-23
Publication date: 2020-12-29

Abstract

The application discloses an image database updating method, an image database updating device, electronic equipment and a medium, and relates to the technical field of artificial intelligence such as deep learning and computer vision. The image database updating method comprises the following steps: acquiring an input image; determining a similarity between the input image and a plurality of stored images in the database; determining at least one target image for the input image from the plurality of stored images based on the similarity; and updating the database based on the input image in the case that the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than or equal to a preset number.

Description

Image database updating method and device, electronic equipment and medium

Technical Field

The present application relates to the field of computer technologies, and in particular, to the field of artificial intelligence technologies such as deep learning and computer vision, and in particular, to a method and an apparatus for updating an image database, an electronic device, and a medium.

Background

The image recognition technology is widely applied, and in the image recognition process, generally, a current input image is obtained, similarity comparison is performed between the input image and stored images in an image database, and if the similarity reaches a similarity threshold, it can be determined that the input image passes verification. Taking an example that an image recognition technology is applied to a payment scene, a user uploads a face image to an image database in advance, when a payment operation is executed later, the current face image of the user is obtained, similarity comparison is carried out between the current face image and the face image in the image database, if the similarity reaches a similarity threshold value, it is determined that face recognition is verified, and the user is allowed to pay. However, when the image quality of the stored image in the image database is not good, the authentication of the image will fail, thereby degrading the user experience.

Disclosure of Invention

The application provides an updating method and device for an image database, an electronic device and a storage medium.

According to a first aspect, the present application provides an updating method of an image database, including: the method comprises the steps of acquiring an input image, determining the similarity between the input image and a plurality of stored images in a database, determining at least one target image aiming at the input image from the plurality of stored images based on the similarity, and updating the database based on the input image when the similarity between the input image and the at least one target image meets a preset similarity condition and the number of the at least one target image is less than or equal to a preset number.

According to a second aspect, the present application provides an updating apparatus for an image database, comprising: the device comprises an acquisition module, a similarity determination module, a target image determination module and an updating module. The image processing device comprises an acquisition module, a similarity determination module, a target image determination module and an updating module, wherein the acquisition module is used for acquiring an input image, the similarity determination module is used for determining the similarity between the input image and a plurality of stored images in a database, the target image determination module is used for determining at least one target image aiming at the input image from the plurality of stored images based on the similarity, and the updating module is used for updating the database based on the input image under the condition that the similarity between the input image and the at least one target image meets a preset similarity condition and the number of the at least one target image is less than or equal to a preset number.

According to a third aspect, the present application provides an electronic device comprising: at least one processor and a memory communicatively coupled to the at least one processor. Wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method as described above.

According to a fourth aspect, the present application provides a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method as described above.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present application, nor do they limit the scope of the present application. Other features of the present application will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not intended to limit the present application. Wherein:

fig. 1 schematically illustrates an application scenario of image database update according to an embodiment of the present application;

FIG. 2 schematically illustrates a flow chart of a method for updating an image database according to an embodiment of the present application;

FIG. 3 schematically shows a flow chart of a method for updating an image database according to another embodiment of the present application;

FIG. 4 schematically shows a flow chart of a method for updating an image database according to another embodiment of the present application;

FIG. 5 is a schematic diagram that schematically illustrates an update of an image database, in accordance with an embodiment of the present application;

FIG. 6 schematically shows a flow diagram of training a quality assessment model according to an embodiment of the present application;

FIG. 7 schematically shows a diagram of a training quality assessment model according to an embodiment of the present application;

FIG. 8 schematically shows a diagram of a training quality assessment model according to another embodiment of the present application;

fig. 9 is a block diagram schematically showing an updating apparatus of an image database according to an embodiment of the present application; and

fig. 10 is a block diagram of an electronic device for implementing the image database updating method according to the embodiment of the present application.

Detailed Description

The following description of the exemplary embodiments of the present application, taken in conjunction with the accompanying drawings, includes various details of the embodiments of the application for the understanding of the same, which are to be considered exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present application. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The terms "comprises," "comprising," and the like, as used herein, specify the presence of stated features, steps, operations, and/or components, but do not preclude the presence or addition of one or more other features, steps, operations, or components.

All terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined. It is noted that the terms used herein should be interpreted as having a meaning that is consistent with the context of this specification and should not be interpreted in an idealized or overly formal sense.

Where a convention analogous to "at least one of A, B and C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B and C" would include but not be limited to systems that have a alone, B alone, C alone, a and B together, a and C together, B and C together, and/or A, B, C together, etc.).

The embodiment of the application provides an updating method of an image database, which comprises the following steps: an input image is acquired and the similarity between the input image and a plurality of stored images in a database is determined. Next, at least one target image for the input image is determined from the plurality of stored images based on the similarity. And updating the database based on the input image in the case that the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than or equal to a preset number.

Fig. 1 schematically shows an application scenario of image database update according to an embodiment of the present application.

As shown in fig. 1, an application scenario 100 of the embodiment of the present application includes, for example, a user 110, an acquisition device 120, and an image database 130.

In the embodiment of the present application, the input image 111 of the user 110 may be acquired by the acquisition device 120, and the input image 111 of the user may be a face image of the user. After the acquisition device 120 acquires the input image 111 of the user, image recognition may be performed on the input image 111.

The process of image recognition may be performed by an electronic device. The electronic device may be a computer, a mobile phone, a server, etc. The capture device 120 may be installed in an electronic device, and the electronic device captures the input image 111 of the user through the capture device 120 and performs image recognition processing on the input image 111. Or the acquisition device 120 may be an external device independent of the electronic device, the acquisition device 120 is in data connection with the electronic device, and the acquisition device 120 may transmit the acquired input image 111 to the electronic device, so that the electronic device can perform image recognition processing on the received input image 111.

In the image recognition process, the input image 111 may be compared with stored images in the image database 130, which may include, for example, a plurality of

images

131, 132, 133, etc., for similarity. The stored images in the image database 130 may be pre-registered images, which may include facial images of multiple users. When the similarity between the input image 111 and at least one stored image in the image database 130 reaches a similarity threshold, it may be determined that the input image 111 passes the recognition verification, and the user 110 is allowed to perform the correlation operation. For example, taking a payment scenario as an example, when the input image 111 of the user passes the identification verification, the user 110 may be allowed to perform a payment operation.

An embodiment of the present application provides an updating method for an image database, and an updating method for an image database according to an exemplary embodiment of the present application is described below with reference to fig. 2 to 8 in combination with an application scenario of fig. 1.

Fig. 2 schematically shows a flowchart of an update method of an image database according to an embodiment of the present application.

As shown in fig. 2, the method 200 for updating an image database according to the embodiment of the present application may include operations S201 to S204, for example.

In operation S201, an input image is acquired.

In operation S202, a similarity between an input image and a plurality of stored images in a database is determined.

In operation S203, at least one target image for the input image is determined from among the plurality of stored images based on the similarity.

In operation S204, in a case where the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than or equal to a preset number, the database is updated based on the input image.

In an embodiment of the present application, the input image may be an image of a currently captured user, and the determined target image for the input image may be at least one image of a plurality of stored images. For example, when the plurality of stored images include images of a plurality of users, each of the determined at least one target image and the input image may be images belonging to the same user.

According to an embodiment of the present application, the at least one target image may include a plurality of target images. The input image and the plurality of target images having a similarity to each other satisfying the preset similarity condition may be that the similarity of the target image having a large part among the plurality of target images and the input image is high.

In the embodiment of the application, the at least one target image may include a plurality of target images, that is, for one user, a plurality of images of the user may be stored in the database, so that the accuracy of identification and verification may be improved in the process of performing identification and verification on the input image. In addition, by comparing the similarity between the input image and the plurality of target images, the input image can be added into the database or the input image is used for replacing one target image under the condition that the similarity meets the preset similarity condition and the number of the target images is less than or equal to the preset number, so that the database can be updated in real time, the image quality of the images in the database is improved, and the identification accuracy of image identification verification is improved.

Fig. 3 schematically shows a flowchart of an update method of an image database according to another embodiment of the present application.

As shown in fig. 3, the method 300 for updating an image database according to the embodiment of the present application may include operations S301 to 304, for example. Wherein, the operation S304 may include operations S304A to S304C.

In operation S301, an input image is acquired.

In operation S302, a similarity between an input image and a plurality of stored images in a database is determined.

In operation S303, at least one target image for the input image is determined from the plurality of stored images based on the similarity.

In operation S304, in a case where the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than or equal to a preset number, the database is updated based on the input image.

In an embodiment of the present application, the preset similarity condition may include: in the at least one target image, the ratio between the number of target images with the similarity to the input image being greater than the first similarity threshold and the number of the at least one target image is greater than a preset ratio.

For example, the first similarity threshold may be 90%, 95%, etc. The preset ratio may be, for example, 50%, 60%, etc. Taking the first similarity threshold as 80% and the preset ratio as 50%, taking the at least one target image including 10 target images as an example, when the similarity between each target image and the input image in more than 50% of the target images (e.g., 6 target images) is greater than the first similarity threshold of 80%, determining that the similarity between the input image and the at least one target image satisfies the preset similarity condition.

In operation S304A, it is determined that the number of at least one target image is less than or equal to a preset number.

In operation S304B, in a case where the similarity between the input image and the at least one target image satisfies the preset similarity condition and the number of the at least one target image is less than the preset number, the input image is added to the database.

In operation S304C, in a case where the similarity between the input image and the at least one target image satisfies the preset similarity condition and the number of the at least one target image is equal to the preset number, one of the at least one target image is replaced with the input image based on at least one of the input image and each target image.

In an embodiment of the application, the number of images stored in the database for the same user does not exceed a preset number, for example, which may be 9, 10, etc. Wherein each of the at least one target image belongs to the same user. In updating the database based on the input images, the input images may be added to the database when the number of the at least one target image is less than a preset number, at which time the number of images of the user stored in the database is increased by 1. In case that the number of at least one target image is equal to the preset number, one target image may be replaced with the input image, for example, one target image is deleted from the database, and the input image is stored in the database, while the number of images of the user stored in the database remains unchanged.

In an example of the present application, one of the at least one target image may be replaced with the input image based on the input image and each target image. For example, one target image to be updated is determined from at least one target image based on the image quality evaluation of the input image and the image quality evaluation of each target image, the image quality evaluation of the target image to be updated is, for example, lower than the image quality evaluation of the input image, and the target image to be updated is replaced with the input image.

For example, an image quality rating of the input image and an image quality rating of each target image may be determined and compared to determine one or more target images from the plurality of target images that are lower than the image quality rating of the input image. Then, one of the one or more target images with low image quality evaluation is taken as an image to be updated to replace the image to be updated with the input image, for example, the image to be updated in the database is deleted, and the input image is stored in the database. Among them, the image to be updated may be the one with the lowest quality evaluation among the one or more target images with low image quality evaluation.

In an embodiment of the application, the image quality assessment characterizes, for example, the quality of the image. The image quality is related to factors such as the resolution, the blur degree, etc. of the image. The image quality evaluation may be expressed in terms of an evaluation score, and the higher the evaluation score is, the higher the image quality evaluation is, i.e., the higher the image quality is. For example, the higher the image resolution and the lower the degree of blur, the higher the image quality evaluation. The lower the image resolution and the higher the degree of blurring, the lower the image quality evaluation.

In the embodiment of the present application, whether to update the database is determined by comparing the similarity between the input image and each target image and according to the image quality evaluation of the input image and the image quality evaluation of each target image. When the image quality evaluation of the input image is greater than the image quality evaluation of the one or more target images, the input image can be used for replacing the image to be updated in the one or more target images so as to update the database in real time, improve the image quality of the image in the database and further improve the identification accuracy of image identification verification.

In another example of the present application, one of the at least one target image may be replaced with the input image on a per target image basis. For example, a target image to be updated is determined from at least one target image, the storage time of the target image to be updated in the database meets a preset time condition, and the target image to be updated is replaced by the input image.

In the embodiment of the present application, the storage time satisfying the preset time condition may be a target image with the earliest storage time. For example, taking an example that at least one target image includes three target images, the storage time of the three target images stored in the database is 2018 year 1 month 1 day, 2019 year 1 month 1 day, and 2020 year 1 month 1 day in this order, and if the image with the earliest storage time among the three target images is the target image with the storage time of 2018 year 1 month 1 day, the target image is determined as the image to be updated, and the image to be updated is replaced with the input image.

In the embodiment of the present application, an image whose storage time is earlier is an image whose image quality is lower with a high probability, and therefore, an image may be replaced according to the storage time to update the database. The timeliness of the images can be improved by updating the database based on the storage time, so that the images in the database are updated along with the time, the image quality of the images in the database is improved, and the identification accuracy of image identification and verification is improved.

In another example of the present application, when a plurality of target images lower in image quality evaluation than the input image are determined from the plurality of target images, one target image whose storage time is the earliest may be determined from the plurality of target images as an image to be updated, and the image to be updated may be replaced with the input image. That is, the embodiment of the application can determine the image to be updated based on the image quality evaluation and the storage time at the same time, thereby improving the image quality of the image in the database.

In another example of the present application, the at least one target image includes a registration image, and the embodiment of the present application may replace one of the at least one target image with an input image on a per target image basis. For example, one of the at least one target image other than the registered image is replaced with the input image.

In the embodiment of the present application, the registered image is, for example, the most original image of the user. When the image is replaced, other images except the registered image can be replaced, so that the registered image which is representative of the user can be reserved, and the identification accuracy of image identification verification is improved. For example, one or more target images having an image quality lower than the input image are determined from the target images other than the registered image, and then one of the one or more target images having the image quality lower than the input image may be replaced with the input image. When the image quality of a plurality of target images determined from the target images other than the registered image is lower than the image quality evaluation of the input image, one of the target images whose storage time is the earliest may be determined as an image to be updated, and the image to be updated may be replaced with the input image.

Fig. 4 schematically shows a flowchart of an update method of an image database according to another embodiment of the present application.

As shown in fig. 4, the method 400 for updating an image database according to the embodiment of the present application may include, for example, operations S401 to 405. Wherein, the operation S403 may include operations S403A through S403C, and the operation S404 may include operations S404A through S404D.

In operation S401, an input image is acquired.

In operation S402, a similarity between an input image and a plurality of stored images in a database is determined.

In operation S403, at least one target image for the input image is determined from among the plurality of stored images based on the similarity.

In operation S403A, a stored image having the greatest similarity with the input image among the plurality of stored images is determined as a reference target image based on the similarity.

In operation S403B, identification information of the reference target image is determined. The identification information may be, for example, an identity of a user, each image stored in the database includes identification information, and images with the same identification information belong to the same user.

In operation S403C, at least one target image is determined from the plurality of stored images, the identification information of each of the at least one target image being identical to the identification information of the reference target image. Wherein the determined at least one target image comprises a reference target image.

In operation S404, in a case where the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than or equal to a preset number, the database is updated based on the input image.

In operation S404A, it is determined that the number of at least one target image is less than or equal to a preset number.

In operation S404B, in a case where the similarity between the input image and the at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than a preset number, the input image is added to the database.

In operation S404C, in a case where the similarity between the input image and the at least one target image satisfies the preset similarity condition and the number of the at least one target image is equal to the preset number, one of the at least one target image is replaced with the input image based on at least one of the input image and each target image.

Operation S404D may be performed before operation S404A to determine whether the similarity satisfies a preset similarity condition. If so, operation S404A is performed, otherwise, it ends.

After operation S402, operation S405 may be performed to determine that the input image passes the recognition verification in a case where the similarity between the input image and one of the stored images is greater than a second similarity threshold, where the second similarity threshold is less than the first similarity threshold. The one stored image may be a stored image having the greatest similarity with the input image among the plurality of stored images.

According to the embodiment of the present application, after determining the stored image with the largest similarity with the input image, the similarity between the input image and the stored image may be compared, and it is determined whether the similarity between the input image and the stored image is greater than a second similarity threshold, which may be a preset threshold, for example, the second similarity threshold may be 80%, 85%, or the like. Under the condition that the similarity between the input image and the stored image is larger than the second similarity threshold value, the input image can be determined to pass the identification verification, and therefore the efficiency and the verification accuracy of the image identification verification are improved. Taking the payment scenario as an example, when the input image passes the identification verification, the user may be allowed to perform a payment operation. One skilled in the art will appreciate that a variety of methods may be used to determine the similarity between the input image and the stored image. Furthermore, as will be understood by those skilled in the art, the input image may be preprocessed, for example, normalized, before calculating the similarity between the input image and the stored image, which is not limited by the embodiments of the present application.

Fig. 5 schematically shows a schematic diagram of updating of an image database according to an embodiment of the present application.

As shown in fig. 5, the

input images

511A, 511B and 7 stored

images

521A, 521B, 521C, 521D, 522A, 522B, 522C in the database 520 are described as an example. Each stored image has, for example, identification information, e.g., stored

images

521A, 521B, 521C, 521D have the same identification information "user 1", indicating that the stored

images

521A, 521B, 521C, 521D belong to the same user. The stored

images

522A, 522B, 522C have the same identification information "user 2," indicating that the stored

images

522A, 522B, 522C belong to the same user.

When the input image is the image 511A, one having the highest similarity to the input image is determined from all stored images stored in the database as the reference target image, for example, the image 521A has the highest similarity to the image 511A, and the image 521A is determined as the reference target image. Then, a plurality of

target images

521A, 521B, 521C, 521D for the input image are determined based on the identification information "user 1" of the reference target image. Taking the preset number of 4 as an example, the number (e.g., the number of 4) of the plurality of

target images

521A, 521B, 521C, 521D is equal to the preset number, and when the similarity between the input image 511A and the at least one target image satisfies the preset similarity condition, one of the plurality of

target images

521A, 521B, 521C, 521D may be replaced with the input image 511A. In replacing the plurality of

target images

521A, 521B, 521C, 521D, the target image to be updated of the plurality of

target images

521A, 521B, 521C, 521D may be determined based on at least one of the above-mentioned image quality evaluation of the input image and the image quality evaluation of the plurality of target images, the storage time of the plurality of target images, and a registered image of the plurality of target images, for example, the target image 521D is determined as the target image to be updated. Next, the target image 521D is replaced with the input image 511A. That is, the target image 521D is deleted from the database 520, and the input image 511A is stored into the database 520.

When the input image is the image 511B, one having the highest similarity to the input image is determined from all stored images stored in the database as a reference target image, for example, the image 522A has the highest similarity to the image 511B, and the image 522A is determined as the reference target image. Then, a plurality of

target images

522A, 522B, 522C for the input image are determined based on the identification information "user 2" of the reference target image. Taking the preset number 4 as an example, the number (e.g., the number is 3) of the plurality of

target images

522A, 522B, 522C is smaller than the preset number, and when the similarity between the input image 511B and the plurality of

target images

522A, 522B, 522C satisfies the preset similarity condition, the input image 511B may be added to the database 520.

According to an embodiment of the application, determining similarity between the input image and the plurality of stored images in the database with respect to each other comprises: the method includes processing an input image and a plurality of stored images using an image recognition model to obtain image features of the input image and image features of each stored image, and determining a similarity between the input image and each stored image based on the image features of the input image and the image features of each stored image.

In an embodiment of the present application, the input image and the plurality of stored images may be processed using an image recognition model to obtain image features of the input image and image features of each stored image, and then a similarity between the input image and each stored image may be determined based on the image features of the input image and the image features of each stored image. The image features may be represented as feature vectors of the image, and the similarity may be determined by calculating a distance between the feature vector of the input image and the feature vector of the stored image, where a smaller distance indicates a higher similarity, and a larger distance indicates a lower similarity. Those skilled in the art will understand that the distance between the feature vectors includes, but is not limited to, euclidean distance, cosine distance, manhattan distance, which is not limited by the embodiments of the present application.

In embodiments of the present application, the image recognition model may include, but is not limited to, a convolutional neural network model.

According to the embodiment of the application, the input image and each target image can be subjected to quality evaluation by utilizing the trained quality evaluation model so as to obtain the image quality evaluation of the input image and the image quality evaluation of each target image. The quality evaluation model may include, but is not limited to, a convolutional neural network model, and the image quality evaluation of the input image and the image quality evaluation of each target image are obtained by inputting the input image and each target image into the convolutional neural network model for processing. The image quality evaluation by using the trained quality evaluation model can improve the accuracy of the quality evaluation.

In an example, the embodiment of the application may perform image quality evaluation on stored images in a database in advance, and store the image quality evaluation for each stored image and the stored image in association with each other in the database, so that after a target image is determined subsequently, comparison may be performed based on the image quality evaluation of an input image and the image quality evaluation of the predetermined target image, thereby improving the contrast efficiency of the image quality evaluation.

In another example, when comparing the image quality evaluation of the input image with the image quality evaluation of the target image, the embodiment of the application may process the input image and the target image by using the quality evaluation model in real time to obtain the image quality evaluation of the input image and the image quality evaluation of the target image, so as to perform the quality evaluation on the input image and the target image which need to be compared in a targeted manner.

Fig. 6 schematically shows a flow chart of training a quality assessment model according to an embodiment of the application.

As shown in fig. 6, the method 600 for updating an image database according to the embodiment of the present application may include operations S620 to S660, for example.

In operation S620, the plurality of sample images are processed using an image recognition model to obtain a similarity between a reference image of the plurality of sample images and other sample images except the reference image.

In operation S640, an image quality evaluation of each of the other sample images is obtained based on the similarity between the other sample images and the reference image with the image quality of the reference image as a reference.

In operation S660, a quality evaluation model is trained based on the plurality of sample images and an image quality evaluation of each of the plurality of sample images.

In the embodiment of the present application, a sample image with the highest image quality may be specified as a reference image from among a plurality of sample images. And then processing the reference image and other sample images by using the image recognition model to obtain the similarity between the reference image and other sample images. Next, an image quality evaluation of each of the other sample images is determined based on the similarity between the other sample images and the reference image with the image quality of the reference image as a reference. When the degree of similarity between the other sample image and the reference image is higher, it indicates that the image quality evaluation of the sample image is higher, i.e., the image quality evaluation of the sample image is closer to the image quality evaluation of the reference image.

After obtaining the image quality evaluation of each sample image, the image quality evaluation of each sample image may be taken as the label information of the sample image, and the quality evaluation model may be trained based on the sample image having the label information such that the image quality evaluation output by the quality evaluation model processing the sample image is closer to the image quality evaluation indicated by the label information of the sample image.

In the embodiment of the application, the image quality evaluation of the sample image is obtained by processing the sample image by using the image recognition model, so that the accuracy of the image quality evaluation of the sample image is improved. The quality evaluation model is trained based on the sample image and the image quality evaluation corresponding to the sample image, so that the evaluation accuracy of the quality evaluation model is improved, and the image quality evaluation obtained by evaluating the quality of the input image and the stored image in the database based on the trained quality evaluation model is more accurate.

According to the embodiment of the application, after the image recognition model is optimized and updated, the sample image can be reprocessed based on the updated image recognition model to obtain the image quality evaluation of the updated sample image. Then, the quality evaluation model is retrained based on the sample image and the updated image quality evaluation thereof, so that the quality evaluation of the input image and the plurality of stored images is performed again based on the updated quality evaluation model, and the quality evaluation process is optimized to improve the updating effect of the database.

Fig. 7 schematically shows a schematic diagram of a training quality evaluation model according to an embodiment of the present application.

As shown in fig. 7, an embodiment of the present application includes a plurality of

sample images

711, 712, 713, 714, and includes an image recognition model 720, a quality evaluation model 730.

For example, one sample image is designated as a reference image from among the plurality of

sample images

711, 712, 713, 714, the sample image 711 is designated as a reference image, for example, and the image quality evaluation of the reference image is designated, for example, the image quality evaluation of the reference image is designated as 100 points. The plurality of

sample images

711, 712, 713, 714 are input into the image recognition model 720 for processing, and the similarity between the

sample images

712, 713, 714 and the sample image 711 is obtained. For example, the similarity between the sample image 712 and the sample image 711 is 95%, the similarity between the sample image 713 and the sample image 711 is 90%, and the similarity between the sample image 714 and the sample image 711 is 85%.

After the similarity between the

specimen images

712, 713, 714 and the specimen image 711 to each other is obtained, the image quality evaluation of each of the

specimen images

712, 713, 714 is obtained with the image quality score of the specimen image 711 as a reference. The sample image having the greater similarity to the sample image 711 has the higher image quality evaluation of the sample image, that is, the closer to the image quality evaluation of the sample image 711. For example, the image quality evaluation of the sample image 712 may be 95 points, the image quality evaluation of the sample image 713 may be 90 points, and the image quality evaluation of the sample image 714 may be 85 points. In one example, the evaluation score of the image quality evaluation of the other sample image may be, for example, the similarity between the sample image and the reference image multiplied by the evaluation score of the image quality evaluation of the reference image. Taking the sample image 712 as an example, the evaluation score of the image quality evaluation of the sample image 712 is, for example, 95% by 100 to 95.

After obtaining the image quality evaluations of the plurality of

sample images

711, 712, 713, 714, the plurality of

sample images

711, 712, 713, 714 and their respective corresponding image quality evaluations may be input to the quality evaluation model 730 for training the quality evaluation model 730.

Fig. 8 schematically shows a schematic diagram of a training quality assessment model according to another embodiment of the present application.

As shown in fig. 8, the plurality of sample images of the embodiment of the present application include, for example, a plurality of sets of

sample images

810A, 810B, 810C, and an image recognition model 820 and a quality evaluation model 830. Where each of the multiple sets of

sample images

810A, 810B, 810C is for one user. For example, a first set of images 810A includes a plurality of

sample images

811A, 812A, 813A, 814A for user 1, a second set of images 810B includes a plurality of

sample images

811B, 812B, 813B, 814B for user 2, and a third set of images 810C includes a plurality of

sample images

811C, 812C, 813C, 814C for user 3.

For each of the plurality of sets of sample images, the sample image includes a reference image and other sample images except the reference image.

Taking the first group of sample images 810A as an example, one sample image is designated as a reference image from among the plurality of

sample images

811A, 812A, 813A, 814A, for example, the sample image 811A is designated as the reference image, and the image quality evaluation of the reference image, for example, the image quality evaluation of the reference image is 100 points. The plurality of

sample images

811A, 812A, 813A, and 814A are input to the image recognition model 820 and processed, and the similarity between the

sample images

812A, 813A, and 814A and the sample image 811A is obtained. For example, the similarity between the sample image 812A and the sample image 811A is 95%, the similarity between the sample image 813A and the sample image 811A is 90%, and the similarity between the sample image 814A and the sample image 811A is 85%.

After the similarity between the

sample images

812A, 813A, 814A and the sample image 811A is obtained, the image quality evaluation of each of the

sample images

812A, 813A, 814A is obtained with the image quality score of the sample image 811A as a reference. The sample image having the greater similarity to the sample image 811A has the higher image quality evaluation of the sample image, i.e., closer to the image quality evaluation of the sample image 811A. For example, the image quality evaluation of the sample image 812A may be 95 points, the image quality evaluation of the sample image 813A may be 90 points, and the image quality evaluation of the sample image 814A may be 85 points. In one example, the evaluation score of the image quality evaluation of the other sample image may be, for example, the similarity between the sample image and the reference image multiplied by the evaluation score of the image quality evaluation of the reference image. Taking the sample image 812A as an example, the evaluation score of the image quality evaluation of the sample image 812A is, for example, 95% by 100 to 95.

The image quality rating for each sample image in each set of sample images may be obtained the same as or similar to the first set of sample images 810A, and is shown in fig. 8 as an example for each sample image in the first set of sample images 810A. After obtaining an image quality evaluation for each sample image of the plurality of sets of sample images, each sample image and the corresponding image quality evaluation may be input to the quality evaluation model 830 to train the quality evaluation model 830. The quality evaluation model is trained by utilizing multiple groups of sample data, so that the evaluation accuracy of the quality evaluation model is improved.

The method for updating the image database can be applied to the field of face recognition. For example, the input image comprises a face image, the plurality of stored images comprise a plurality of pre-stored face images, and performing recognition verification on the input image comprises performing face recognition verification on the face image.

Fig. 9 is a block diagram schematically showing an updating apparatus of an image database according to an embodiment of the present application.

As shown in fig. 9, the image database updating apparatus 900 according to the embodiment of the present application includes, for example, an obtaining module 901, a similarity determining module 902, a target image determining module 903, and an updating module 904.

The acquisition module 901 may be used to acquire an input image. According to the embodiment of the present application, the obtaining module 901 may, for example, perform the operation S201 described above with reference to fig. 2, which is not described herein again.

The similarity determination module 902 may be configured to determine a similarity between the input image and a plurality of stored images in the database. According to the embodiment of the present application, the similarity determining module 902 may, for example, perform operation S202 described above with reference to fig. 2, which is not described herein again.

The target image determination module 903 may be configured to determine at least one target image for the input image from the plurality of stored images based on the similarity. According to the embodiment of the present application, the target image determining module 903 may perform, for example, operation S203 described above with reference to fig. 2, which is not described herein again.

The updating module 904 may be configured to update the database based on the input image in a case where a similarity between the input image and the at least one target image satisfies a preset similarity condition and a number of the at least one target image is less than or equal to a preset number. According to the embodiment of the present application, the updating module 904 may, for example, perform the operation S204 described above with reference to fig. 2, which is not described herein again.

According to an embodiment of the present application, an electronic device and a readable storage medium are also provided.

As shown in fig. 10, the electronic device 1000 is a block diagram of an image database updating method according to an embodiment of the present application. The electronic device 1000 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application that are described and/or claimed herein.

As shown in fig. 10, the electronic apparatus 1000 includes: one or more processors 1010, memory 1020, and interfaces for connecting the various components, including a high-speed interface and a low-speed interface. The various components are interconnected using different buses and may be mounted on a common motherboard or in other manners as desired. The processor may process instructions for execution within the electronic device 1000, including instructions stored in or on the memory to display graphical information of a GUI on an external input/output apparatus (such as a display device coupled to the interface). In other embodiments, multiple processors and/or multiple buses may be used, along with multiple memories and multiple memories, as desired. Also, multiple electronic devices 1000 may be connected, with each device providing portions of the necessary operations (e.g., as a server array, a group of blade servers, or a multi-processor system). Fig. 10 illustrates an example of a processor 1010.

Memory 1020 is a non-transitory computer readable storage medium as provided herein. The memory stores instructions executable by the at least one processor, so that the at least one processor executes the method for updating the image database provided by the application. The non-transitory computer-readable storage medium of the present application stores computer instructions for causing a computer to execute the updating method of the image database provided by the present application.

The memory 1020, which is a non-transitory computer readable storage medium, may be used to store non-transitory software programs, non-transitory computer executable programs, and modules, such as program instructions/modules corresponding to the updating method of the image database in the embodiment of the present application (for example, the acquiring module 901, the similarity determining module 902, the target image determining module 903, and the updating module 904 shown in fig. 9). The processor 1010 executes various functional applications of the server and data processing, namely, implements the method for updating the image database in the above-described method embodiment, by executing the non-transitory software programs, instructions, and modules stored in the memory 1020.

The memory 1020 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required for at least one function; the storage data area may store data created from use of the electronic device 1000 according to updating of the image database, and the like. Further, the memory 1020 may include high speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid state storage device. In some embodiments, memory 1020 optionally includes memory located remotely from processor 1010, which may be connected to the updated electronic device 1000 of the image database via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The electronic device 1000 of the method for updating an image database may further include: an input device 1030 and an output device 1040. The processor 1010, the memory 1020, the input device 1030, and the output device 1040 may be connected by a bus or other means, and fig. 10 illustrates an example of connection by a bus.

The input device 1030 may receive input numeric or character information and generate key signal inputs related to user settings and function controls of the electronic device 1000 for updating of the image database, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, a pointing stick, one or more mouse buttons, a track ball, a joystick, or other input devices. The output devices 1040 may include a display device, auxiliary lighting devices (e.g., LEDs), and tactile feedback devices (e.g., vibrating motors), among others. The display device may include, but is not limited to, a Liquid Crystal Display (LCD), a Light Emitting Diode (LED) display, and a plasma display. In some implementations, the display device can be a touch screen.

Various implementations of the systems and techniques described here can be realized in digital electronic circuitry, integrated circuitry, application specific ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

These computer programs (also known as programs, software applications, or code) include machine instructions for a programmable processor, and may be implemented using high-level procedural and/or object-oriented programming languages, and/or assembly/machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus, and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and/or data to a programmable processor.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present application may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present application can be achieved, and the present invention is not limited herein.

The above-described embodiments should not be construed as limiting the scope of the present application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An updating method of an image database comprises the following steps:

acquiring an input image;

determining a similarity between the input image and a plurality of stored images in a database;

determining at least one target image for the input image from the plurality of stored images based on the similarity; and

and updating the database based on the input image when the similarity between the input image and at least one target image meets a preset similarity condition and the number of the at least one target image is less than or equal to a preset number.

2. The method of claim 1, wherein the preset similarity condition comprises:

in the at least one target image, the ratio of the number of target images with the similarity to the input image being greater than a first similarity threshold to the number of the at least one target image is greater than a preset ratio.

3. The method of claim 1, wherein the updating the database based on the input image comprises at least one of:

adding the input image to the database in the case that the similarity between the input image and at least one target image satisfies a preset similarity condition and the number of the at least one target image is less than a preset number; and

and replacing one of the at least one target image with the input image based on at least one of the input image and each target image in the case that the similarity between the input image and the at least one target image satisfies the preset similarity condition and the number of the at least one target image is equal to a preset number.

4. The method of claim 3, wherein said replacing one of the at least one target image with the input image based on at least one of the input image and each target image comprises:

determining a target image to be updated from the at least one target image based on the image quality evaluation of the input image and the image quality evaluation of each target image, wherein the image quality evaluation of the target image to be updated is lower than the image quality evaluation of the input image; and

and replacing the target image to be updated with the input image.

5. The method of claim 3, wherein said replacing one of the at least one target image with the input image based on at least one of the input image and each target image comprises;

determining a target image to be updated from the at least one target image, wherein the storage time of the target image to be updated in the database meets a preset time condition; and

and replacing the target image to be updated with the input image.

6. The method of claim 3, wherein the at least one target image includes a registration image; wherein said replacing one of the at least one target image with the input image based on at least one of the input image and each target image comprises:

replacing one other target image of the at least one target image except the registered image with the input image.

7. The method of claim 2, wherein the determining, based on the similarity, at least one target image for the input image from the plurality of stored images comprises:

determining, as the reference target image, a stored image of the plurality of stored images having a maximum similarity with the input image, based on the similarity;

determining identification information of the reference target image; and

determining the at least one target image from the plurality of stored images, the identification information of each of the at least one target image being consistent with the identification information of the reference target image.

8. The method of claim 7, further comprising:

determining that the input image passes recognition verification if a similarity between the input image and the reference target image is greater than a second similarity threshold, wherein the second similarity threshold is less than the first similarity threshold.

9. The method of claim 4, wherein the determining a similarity between the input image and a plurality of stored images in a database with respect to each other comprises:

processing the input image and the plurality of stored images by using an image recognition model to obtain image characteristics of the input image and image characteristics of each stored image; and

determining a similarity between the input image and each stored image based on image features of the input image and image features of each stored image.

10. The method of claim 9, further comprising:

and performing quality evaluation on the input image and each target image by using the trained quality evaluation model to obtain the image quality evaluation of the input image and the image quality evaluation of each target image.

11. The method of claim 10, wherein the quality assessment model is trained by:

processing a plurality of sample images by using the image recognition model to obtain the similarity between a reference image in the plurality of sample images and other sample images except the reference image;

taking the image quality of the reference image as a reference, and obtaining the image quality evaluation of each sample image in other sample images based on the similarity between the other sample images and the reference image; and

training the quality assessment model based on the plurality of sample images and an image quality assessment of each of the plurality of sample images.

12. The method of claim 11, wherein the plurality of sample images comprises a plurality of sets of sample images, each set of sample images for one user;

wherein the reference image and the other sample images except the reference image belong to the same group of sample images among the plurality of groups of sample images.

13. The method of any of claims 1-12, wherein the input image comprises a face image; the plurality of stored images comprises a plurality of pre-stored face images; and the step of identifying and verifying the input image comprises the step of identifying and verifying the face of the face image.

14. An updating apparatus of an image database, comprising:

the acquisition module is used for acquiring an input image;

a similarity determining module for determining similarity between the input image and a plurality of stored images in a database;

a target image determination module to determine at least one target image for the input image from the plurality of stored images based on the similarity; and

an updating module, configured to update the database based on the input image if a similarity between the input image and at least one target image satisfies a preset similarity condition and a number of the at least one target image is less than or equal to a preset number.

15. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 13.

16. A non-transitory computer readable storage medium storing computer instructions for causing a computer to perform the method of any one of claims 1 to 13.