WO2022022493A1 - Image authenticity determination method and system - Google Patents

Abstract

Disclosed in embodiments of the present description is an image authenticity determination method and system. The method is applied to a server, and the method comprises: obtaining an original image from a client; extracting, according to a preset extraction rule, multiple images or image parts from the original image as multiple extracted images; determining, by means of a trained machine learning model on the basis of the multiple extracted images, the degree of matching between the multiple extracted images and a preset sequence; and determining the authenticity of the original image on the basis of the degree of matching, the preset sequence corresponding to a camera device of the client, and determining the authenticity of the original image specifically comprising: determining the authenticity of an image from the camera device.

Description

A method and system for judging the authenticity of an image

cross reference

This application claims priority to Chinese Application No. 202010750632.5 filed on July 30, 2020, the entire contents of which are incorporated herein by reference.

technical field

The embodiments of this specification relate to the technical field of image processing, and in particular, to a method and system for judging the authenticity of an image.

Background technique

With the rapid development of science and technology, more and more application scenarios (such as security, finance, and user registration of application platforms, etc.) need to verify user identity through certificate recognition and face recognition. In order to prevent criminals from forging and using other people's identities to commit crimes, it is a key link in identity authentication to identify the authenticity of the identified document images or face images.

To this end, the embodiments of this specification propose a method and system for judging the authenticity of an image, so as to improve the accuracy of identity authentication.

SUMMARY OF THE INVENTION

An aspect of the embodiments of the present specification provides a method for judging the authenticity of an image, which is applied to a server, the method includes: acquiring an original image from a client; extracting multiple images in the original image according to a preset extraction rule Or the image part is used as multiple extracted images; based on the multiple extracted images, the degree of matching between the multiple extracted images and the preset sequence is determined by the trained machine learning model; based on the matching degree, the original image is judged The preset sequence corresponds to the shooting device of the client, and judging the authenticity of the original image is specifically: judging the authenticity of the image from the shooting device.

An aspect of the embodiments of this specification provides a method for judging the authenticity of an image, which is applied to a client, the method includes: acquiring a shooting parameter sequence generated and delivered by a server; generating the original image based on the shooting parameter sequence ; send the original image to the server; obtain the information sent by the server that includes the result of judging the authenticity of the original image.

An aspect of the embodiments of the present specification provides a system for judging the authenticity of an image, which is applied to a server side. The system includes: a first acquisition module for acquiring an original image from a client; an extraction module for acquiring an original image according to a preset The extraction rule extracts multiple images or image parts in the original image as multiple extracted images; the determining module is configured to determine, based on the multiple extracted images, through the trained machine learning model, the multiple extracted images and the preset images. The matching degree of the sequence is set; the judgment module is used to judge the authenticity of the original image based on the matching degree; the preset sequence corresponds to the shooting device of the client, and the judgment of the authenticity of the original image is specifically: : Determine the authenticity of the image from the photographing device.

An aspect of the embodiments of the present specification provides a system for judging the authenticity of an image, which is applied to a client, and the system includes: a second acquisition module for acquiring a shooting parameter sequence generated and delivered by the server; a generation module for using to generate the original image based on the shooting parameter sequence; a sending module is used to send the original image to the server; a third acquisition module is used to obtain the original image sent by the server, including Information about the results of the judgment of authenticity.

One aspect of the embodiments of this specification provides an apparatus for judging the authenticity of an image, the apparatus includes a processor and a memory; the memory is used for storing instructions, and the processor is used for executing the instructions, so as to implement any of the above Operations corresponding to the method for judging the authenticity of an image.

One aspect of the embodiments of this specification provides a computer-readable storage medium, where the storage medium stores computer instructions, and after the computer reads the computer instructions in the storage medium, the computer executes the method for determining the authenticity of an image as described in any of the above corresponding operation.

Description of drawings

This specification will be further described by way of example embodiments, which will be described in detail with reference to the accompanying drawings. These examples are not limiting, and in these examples, the same numbers refer to the same structures, wherein:

1 is a schematic diagram of an exemplary application scenario of an image authenticity judgment system according to some embodiments of the present specification;

FIG. 2 is an exemplary flowchart of a method for judging the authenticity of an image according to some embodiments of the present specification;

FIG. 3 is an exemplary schematic diagram of extracting a plurality of extracted images according to some embodiments of the present specification;

4 is a schematic diagram of an exemplary structure of a machine learning model according to some embodiments of the present specification;

FIG. 5 is an exemplary flowchart of training a machine learning model according to some embodiments of the present specification;

FIG. 6 is another exemplary structural schematic diagram of a machine learning model according to some embodiments of the present specification;

FIG. 7 is another exemplary flowchart of a method for judging authenticity of an image according to some embodiments of the present specification;

FIG. 8 is a schematic diagram of interaction between a server and a client according to some embodiments of the present specification.

detailed description

In order to illustrate the technical solutions of the embodiments of the present specification more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some examples or embodiments of the present specification. For those of ordinary skill in the art, without creative efforts, the present specification can also be applied to the present specification according to these drawings. other similar situations. Unless obvious from the locale or otherwise specified, the same reference numbers in the figures represent the same structure or operation.

It should be understood that "system", "device", "unit" and/or "module" as used in this specification is a method used to distinguish different components, elements, parts, parts or assemblies at different levels. However, other words may be replaced by other expressions if they serve the same purpose.

As shown in the specification and claims, unless the context clearly dictates otherwise, the words "a", "an", "an" and/or "the" are not intended to be specific in the singular and may include the plural. Generally speaking, the terms "comprising" and "comprising" only imply that the clearly identified steps and elements are included, and these steps and elements do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

Flowcharts are used in this specification to illustrate operations performed by a system according to an embodiment of this specification. It should be understood that the preceding or following operations are not necessarily performed in the exact order. Instead, the various steps can be processed in reverse order or simultaneously. At the same time, other actions can be added to these procedures, or a step or steps can be removed from these procedures.

FIG. 1 is a schematic diagram of an exemplary application scenario of an image authenticity determination system according to some embodiments of the present specification.

In the field of production and life, people often encounter scenarios that require identity authentication. In some embodiments, the identity authentication scenarios may include scenarios in which face recognition is applied, such as face-swiping payment, face-swiping access control, and face-swiping attendance. In some embodiments, the identity authentication scenario may also include a credential identification scenario. For example, a user needs to register as a driver on an online car-hailing platform, and the online car-hailing platform will identify the driver's license, driving license and other relevant documents provided by the user, and determine whether the certificate information is true and whether it complies with relevant regulations.

However, some users will use pre-recorded face action videos for face recognition or use fraudulent ID pictures for ID identification to complete false verification. For example, in the identification of documents, criminals can hijack the camera of the application platform through black production, directly bypass the process of shooting physical documents on the spot, and upload the cheating picture to the application platform for registration, thereby completing false verification. For another example, in face recognition, criminals record in advance actions such as nodding, blinking, and mouth opening, which are commonly used in face recognition, and hijack the camera on the spot during verification and enter the corresponding action video recorded in advance to complete false verification. These false verifications reduce the accuracy of identity authentication and also create some security risks. To this end, this specification proposes a method and system for judging the authenticity of an image, which is used to effectively verify whether a video or image is a video or image shot by a shooting device on-site, that is, a legal video or image, thereby improving the accuracy of identity authentication.

As shown in FIG. 1 , an application scenario of the image authenticity determination system 100 shown in the embodiment of this specification may include a first computing system 140 , a second computing system 170 and a client 110 .

The first computing system 140 may be used to determine the authenticity of the original image. In some embodiments, the first computing system 140 may be used to determine whether the original image is a real image captured by a shooting device on site. For example, it can automatically determine whether the original images such as faces, fingerprints, palm prints, and certificates are real images captured by the shooting equipment on the spot, so as to avoid the camera being hijacked to complete false verification and improve the accuracy of identity authentication.

The first computing system 140 may acquire the extraction image 130 . The extracted image 130 may be obtained from the original image 120 , and the original image 120 may be obtained by the client 110 . In some embodiments, the client 110 may be a photographing device, such as a camera, video recorder, camera, or the like. In some embodiments, the client 110 may be various types of devices having a camera function or including a camera device, such as a mobile phone 110-1, a tablet computer 110-2, a computer 110-3, and the like.

The extracted image 130 may enter the first computing system 140 through various common means (eg, a network). Through the model 141 in the first computing system 140, the degree of matching 150 can be output. The first computing system 140 further obtains a judgment result of the authenticity of the original image based on the matching degree 150 .

The parameters of the model 141 can be obtained through training. The second computing system 170 may acquire multiple sets of training samples 160 , and each set of training samples includes sample image frames and corresponding labels. The second computing system 170 updates the parameters of the initial model 171 through multiple sets of training samples 160 to obtain a trained model. The parameters of the model 141 come from the trained model 171 . Among them, parameters can be passed in any common way.

A model (eg, model 141 or/and model 171 ) may refer to a collection of several methods performed based on a processing device. These methods can include a large number of parameters. When executing the model, the parameters used can be preset or can be dynamically adjusted. Some parameters can be obtained through training, and some parameters can be obtained during execution. For the specific description of the models involved in this specification, please refer to the relevant parts of this specification (FIG. 4, FIG. 5 and their related descriptions).

The first computing system 140 and the second computing system 170 may be the same or different. The first computing system 140 and the second computing system 170 refer to systems with computing capabilities, which may include various computers, such as servers, personal computers, or computing platforms composed of multiple computers connected in various structures.

Processing devices may be included in the first computing system 140 and the second computing system 170, and the processing devices may execute program instructions. The processing device may include various common general-purpose central processing units (CPUs), graphics processing units (Graphics Processing Units, GPUs), microprocessors, application-specific integrated circuits (ASICs), or other types of integrated circuits.

The first computing system 140 and the second computing system 170 may include storage media, and the storage media may store instructions and may also store data. The storage medium may include mass memory, removable memory, volatile read-write memory, read-only memory (ROM), and the like, or any combination thereof.

The embodiments of this specification provide an image authenticity judgment system, the system is applied to the server side, and the image authenticity judgment system may include a first acquisition module, an extraction module, a determination module, and a judgment module.

The fetch module can be used to fetch raw images from the client. In some embodiments, the first obtaining module is further configured to: generate a shooting parameter sequence; send the shooting parameter sequence to the client; obtain the original image from the client, the original image Generated by the client based on the shooting parameter sequence.

In some embodiments, the first obtaining module is further configured to: determine identification information of the shooting device; determine a shooting parameter set of the shooting device based on the identification information; generate the shooting parameter set based on the shooting parameter set Describe the shooting parameter sequence.

In some embodiments, the first obtaining module is further configured to: randomly select a preset number of shooting parameters from the shooting parameter set, and generate the shooting parameter sequence based on the shooting parameters. In some embodiments, the sequence of shooting parameters includes a sequence of color temperature parameters.

The extraction module may be configured to extract multiple images or image parts in the original image as multiple extracted images according to preset extraction rules.

The determining module may be configured to determine, based on the multiple extracted images, the degree of matching between the multiple extracted images and the preset sequence by using a trained machine learning model. In some embodiments, the machine learning model includes at least a plurality of convolutional neural network units and a sequence-to-sequence unit; the determining module is further configured to: pass through each of the plurality of convolutional neural network units, respectively processing each of the multiple extracted images to obtain an image representation vector corresponding to each extracted image; processing the image representation vector through the sequence-to-sequence unit to obtain predictions of the multiple extracted images A change sequence of shooting parameters; the matching degree is determined based on the predicted change sequence of shooting parameters and the preset sequence.

In some embodiments, the machine learning model may be trained by the following method: acquiring a plurality of training samples carrying labels, the training samples including a plurality of sample image frames obtained based on sample shooting parameters, the labels including the The variation relationship of the sample shooting parameters between the multiple sample image frames; the initial machine learning model is trained based on the multiple training samples carrying the labels, and the machine learning model is obtained.

The judging module can be used to judge the authenticity of the original image based on the matching degree; the preset sequence corresponds to the shooting device of the client, and judging the authenticity of the original image is specifically: judging that the image comes from The authenticity of the photographing equipment.

The embodiments of this specification provide an image authenticity judgment system, which is applied to a client, and the image authenticity judgment system may include a second acquisition module, a generation module, a transmission module, a third acquisition module, and an upload module.

The second acquiring module may be configured to acquire the shooting parameter sequence generated and delivered by the server.

A generating module, configured to generate the original image based on the shooting parameter sequence. In some embodiments, the shooting parameter sequence is randomly generated by the server. In some embodiments, the shooting parameter sequence is randomly generated by the server based on shooting parameters of the shooting device; the shooting parameters correspond to identification information of the shooting device. In some embodiments, the sequence of shooting parameters includes a sequence of color temperature parameters.

The sending module can be used to send the original image to the server.

The third obtaining module may be configured to obtain the information sent by the server and including the judgment result of the authenticity of the original image.

The uploading module may be configured to upload the identification information of the photographing device of the client to the server.

It should be understood that the above-described system and its modules can be implemented in various ways. For example, in some embodiments, the system and its modules may be implemented in hardware, software, or a combination of software and hardware. Wherein, the hardware part can be realized by using dedicated logic; the software part can be stored in a memory and executed by a suitable instruction execution system, such as a microprocessor or specially designed hardware. Those skilled in the art will appreciate that the methods and systems described above may be implemented using computer-executable instructions and/or embodied in processor control code, for example on a carrier medium such as a disk, CD or DVD-ROM, such as a read-only memory (firmware) ) or a data carrier such as an optical or electronic signal carrier. The system and its modules of this specification can be implemented not only by hardware circuits such as very large scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc. , can also be implemented by, for example, software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software (eg, firmware).

It should be noted that, the above description of the image authenticity judgment system and its modules is only for convenience of description, and does not limit the description to the scope of the illustrated embodiments. It can be understood that for those skilled in the art, after understanding the principle of the system, various modules may be combined arbitrarily, or a subsystem may be formed to connect with other modules without departing from the principle. Such deformations are all within the protection scope of this specification.

FIG. 2 is an exemplary flowchart of a method for judging the authenticity of an image according to some embodiments of the present specification, and the method is applied to the server side. In some embodiments, the process 200 may be implemented by the first computing system 140 shown in FIG. 1 . As shown in FIG. 2, the process 200 may include the following steps:

Step 210, obtain the original image from the client. In some embodiments, this step 210 may be performed by the first acquisition module.

In some embodiments, the client can be any terminal that includes a camera device. Such as mobile phones, tablets or laptops, etc. In some embodiments, the original image may be an image of the object to be detected. For example, the image of the object to be detected obtained by the application platform. In some embodiments, the object to be detected may be any object that needs to be judged whether it is collected by the photographing device on-site, that is, it is judged whether it is a real object collected by the photographing device or a false object forged in advance before the photographing device collects. For example, the object to be detected may be a face, a palm print, a fingerprint, etc. to be detected, or a certificate to be detected, such as an ID card, a driver's license, and the like.

In some embodiments, the original image may be an image included in a video of the object to be detected recorded by the photographing device. The video of the object to be detected recorded by the photographing device may be a video of a preset duration. For example, a 3-second or 5-second video. Correspondingly, the original image may be an image included in a 3-second or 5-second video.

In some embodiments, the original image may be one or more images of the object to be detected captured by the capturing device. The multiple images may be multiple images generated by the photographing device continuously photographing the object to be detected, or may be multiple images generated by the photographing device photographing the object to be detected at preset time intervals. The embodiments of this specification do not specifically limit how the photographing device obtains multiple images of the object to be detected.

In this specification, the original image may include still one or more photographs, may also include video, or a mixture thereof.

In some embodiments, the first acquisition module may acquire the original image in various ways. In some embodiments, the acquisition module may acquire the original image from the storage device. Among them, the original image is pre-generated and stored in the storage device. For example, the photographing device of the client 110 collects the video and/or image of the object to be detected and sends it to the storage device for storage. At this time, the acquisition module can directly obtain the original image from the storage device. In some embodiments, the acquisition module may acquire in real time the video and/or image of the object to be detected acquired by the photographing device of the client 110 .

In some embodiments, the original image may be generated by the client based on a sequence of shooting parameters. Specifically, acquiring the original image from the client may include: generating a shooting parameter sequence, and sending the shooting parameter sequence to the client; acquiring an original image from the client, where the original image is generated by the client based on the shooting parameter sequence.

In some embodiments, the sequence of shooting parameters may be a sequence of multiple shooting parameters. In some embodiments, the shooting parameters may refer to parameters used by the shooting device when recording or shooting. In some embodiments, the shooting parameters may include a color temperature parameter, a sharpening degree parameter, a color saturation parameter, a brightness parameter, a contrast parameter, a shutter parameter, an aperture parameter, and the like. Correspondingly, when the shooting parameter is a color temperature parameter, the shooting parameter sequence may be a color temperature parameter sequence, that is, a sequence composed of multiple color temperature parameters. In some embodiments, the above capture parameters may be mixed or combined to generate a sequence of capture parameters. For the purpose of simplifying the description, the embodiments of this specification take the shooting parameter sequence as the color temperature parameter sequence as an example for description. It should be known that the shooting parameter sequence is not limited to the color temperature parameter sequence, for example, it can also be a color saturation parameter sequence or an aperture. A sequence of parameters, or other parameters, or a combination of various parameters. The embodiments of the present specification do not limit this.

In some embodiments, the shooting parameter sequence may further include time information and/or image number information. In some embodiments, the time information is used to reflect the information of the corresponding time period when the photographing device uses the corresponding photographing parameters to record the video of the object to be detected. For example, taking the above-mentioned original image as an image included in a 3s video and the shooting parameter sequence as a color temperature parameter sequence as an example, if the 3s is evenly divided into 3 time periods, namely 0-1s, 1-2s and 2-3s; Then the time information can reflect the information of the videos recorded in the time periods of 0-1s, 1-2s and 2-3s respectively using the corresponding color temperature parameters. In some embodiments, the information on the number of images is used to reflect the information on the number of corresponding images of the object to be detected that the photographing device uses with the corresponding photographing parameters. As an example, take the original image as 15 images continuously shot by the shooting device, and the shooting parameter sequence as the color temperature parameter sequence. 10 images and the 10th to 15th images; the number of images can reflect the information of the 1st to 5th images, the 5th to 10th images, and the 10th to 15th images, respectively, using the corresponding color temperature parameters.

In some embodiments, the server side may generate the shooting parameter sequence based on the shooting parameter set contained in the client terminal. Specifically, generating a shooting parameter sequence may include: determining identification information of the shooting device; determining a shooting parameter set of the shooting device based on the identification information; generating the shooting parameter sequence based on the shooting parameter set.

In some embodiments, the identification information may include the model or performance parameters of the photographing device. In some embodiments, the first obtaining module may determine the identification information of the photographing device based on an operating system included in the client. For example, if the client is a mobile phone, the model or performance parameter corresponding to the shooting device may be determined based on the IOS or Android operating system included in the mobile phone.

In some embodiments, the set of shooting parameters may be all or part of the settable shooting parameters contained in the shooting device. In some embodiments, different models of photographing devices may have different sets of photographing parameters. Still taking the above example as an example, the IOS mobile phone and the Android mobile phone have different shooting parameter sets.

In some embodiments, the first computing system 140 (i.e., the server side) may randomly select a preset number of shooting parameters from the shooting parameter set, and generate a shooting parameter sequence based on the shooting parameters. For example, still taking the color temperature parameter sequence as an example, if the selectable color temperature parameter ranges of the photographing device include 400-420nm, 460-470nm, 568-572nm, 6000-6500k, 10000-12000k, and 601-606nm, then the first computing system 140 You can select 5 color temperature parameter points from the above 6 color temperature parameter points, such as 460-470nm (hereinafter referred to as color temperature parameter 1), 568-572nm (hereinafter referred to as color temperature parameter 2), 6000-6500k (hereinafter referred to as color temperature parameter 3) , 10000-12000k (hereinafter referred to as color temperature parameter 4) and 601-606 nm (hereinafter referred to as color temperature parameter 5) to generate a color temperature parameter sequence, for example, the color temperature parameter sequence can be: color temperature parameter 3-color temperature parameter 1-color temperature parameter 5.

As mentioned above, the shooting parameter sequence may include time information and/or image number information, therefore, the color temperature parameter sequence may include time information and/or image number information. For example, still taking the above-mentioned original image as an image included in a 3s video and the color temperature parameter sequence including time information as an example, the color temperature parameter sequence can be s={1, 2, 3}, which can reflect the 0 in the 3s video. -1s is recorded with color temperature parameter 1, 1-2s is recorded with color temperature parameter 2, and 2-3s is recorded with color temperature parameter 3.

In some embodiments, the first computing system 140 (ie, the server) may deliver the sequence of shooting parameters to the client through the network.

As mentioned earlier, the raw image can be generated by the client based on the sequence of shooting parameters. In some embodiments, the original image may be generated by the client's photographing device by photographing or recording a corresponding image or video based on the photographing parameter sequence. Still taking the above example as an example, the client can use the color temperature parameter sequence s={1, 2, 3} to record 0-1s of the video with color temperature parameter 1, and record 1-2s of the video with color temperature parameter 2, and use the color temperature parameter 3 Record 2-3s of the video, record the complete 3s video, and then use the image included in the 3s video as the original image.

Step 220: Extract multiple images or image parts in the original image as multiple extracted images according to a preset extraction rule. In some embodiments, this step 220 may be performed by an extraction module.

In some embodiments, the plurality of extracted images may be a plurality of images or portions of images in the original image. In some embodiments, the plurality of extracted images may be a video of the object to be detected recorded by the photographing device or a plurality of images included in a plurality of captured images of the object to be detected. In some embodiments, the plurality of extracted images may be image portions of each of the plurality of images. For example, an image of a certain area is extracted from each image as an image part.

In some embodiments, the preset extraction rules may be specifically set according to actual requirements. In some embodiments, the preset extraction rules may match how the original images were acquired. In some embodiments, preset extraction rules may be matched to sequences of shooting parameters. In some embodiments, the preset extraction rule may be matched with time information and/or image number information reflected by the shooting parameter sequence.

Still taking the above example as an example, the original image is 0-1s of video recorded with color temperature parameter point 1, 1-2s of video recorded with color temperature parameter point 2, and 3s video of 2-3s recorded with color temperature parameter point 3, then The preset extraction rule may be to extract any image from the video segments of 0-1s, 1-2s, and 2-3s, respectively, to generate multiple extracted images. Referring to FIG. 3 , FIG. 3 is an exemplary schematic diagram illustrating extracting a plurality of extracted images according to this example. As shown in Figure 3, the original image 310 is a video with a duration of 3s, then the last frame of image (shown in gray) can be extracted from the video segments of 0-1s, 1-2s, and 2-3s respectively, as multiple Image 320 is extracted.

For another example, different regions of an image may be extracted according to preset rules, and each region may be used as an extracted image, or multiple regions may be extracted from each of multiple images as multiple extracted images.

Step 230: Based on the multiple extracted images, determine the degree of matching between the multiple extracted images and the preset sequence by using the trained machine learning model. In some embodiments, this step 230 may be performed by a determination module.

In some embodiments, the machine learning model may be a pre-trained model. For the training process of the machine learning model, reference may be made to FIG. 5 and its related description, which will not be repeated here. A trained machine learning model can determine how well multiple extracted images match a preset sequence.

In some embodiments, a convolutional neural network model can be used to determine the features of each extracted image, and the output features can be compared to a preset sequence. The convolutional neural network model used can be obtained by training. During training, images can be obtained by using the shooting parameters corresponding to the preset sequence, and further extracted as training data, with the corresponding preset sequence as the label, through the iterative method of optimizing the loss function. to train.

In some embodiments, a combination of a convolutional neural network unit and a sequence-to-sequence (Seq2Seq) unit may be used as a machine learning model. For details, see FIG. 4 and related descriptions, which will not be repeated here.

A preset sequence is a conditional value used for feature recognition of an image. It can be a sequence composed of multiple values, multiple vectors, or other data, or a value or other representation can be uniformly referred to as a sequence.

In some embodiments, the preset sequence corresponds to the client's camera device. In some embodiments, the preset sequence may be a sequence corresponding to changes in parameters in the sequence of shooting parameters. Illustratively, still taking the above example as an example, if the shooting parameter sequence is the color temperature parameter sequence s={1, 2, 3}, the preset sequence may be the change of color temperature parameters 1 to 2 and the change of color temperature parameters 2 to 3. A sequence of changes made up of changes.

In some embodiments, the changed values of the parameters in the preset sequence may be determined by preset encoding information. For example, the preset coding information includes that the change of color temperature parameter 1 to 2 is represented by the code character a, and the change of the color temperature parameter 2 to 3 is represented by the code character b. Then the preset sequence can be s'={a,b}.

As another example, the preset sequence may be a code value used to distinguish photographing equipment. The preset sequence may have various representation forms, and its functions are not substantially different, which is not limited in this specification.

In some embodiments, the matching degree may reflect the similarity between the plurality of extracted images and the preset sequence. In some embodiments, the matching degree may reflect the similarity between the sequence of changes of the predicted shooting parameters of the multiple extracted images and the preset sequence, that is, the difference between the sequence of changes of the predicted shooting parameters of the multiple extracted images and the parameters in the sequence of shooting parameters Similarity between sequences of changes. It can be understood that the greater the value of the matching degree, the greater the similarity between the two, and the greater the possibility that the original image is real.

For the acquisition of the matching degree, various transformation methods can be used, such as inputting the preset sequence into the machine learning model at the same time to directly output the matching degree, etc. There is no essential difference between these methods, which is not limited in this specification.

As can be seen from the above description, in some embodiments, by verifying the association relationship between the image and the photographing device, an attacker can effectively prevent an attacker from bypassing the photographing device and directly uploading a prefabricated image. In particular, by analyzing different parts of the image extraction based on machine learning models, rather than simply verifying based on the parameters attached to the image, even if the attacker hijacks the shooting device and obtains the verification parameters, it is impossible to simply generate data for verification. , greatly improving the defense strength.

Further, in some implementations, by generating and delivering the shooting parameter sequence to the client, the shooting device of the client generates the corresponding original image. Therefore, the shooting parameters of the original image actually shot by the shooting device must be the same as the shooting parameter sequence. The shooting parameters are the same. Since the pre-prepared fake images cannot contain exactly the same shooting parameters, the attacker can prevent the attacker from hijacking the shooting equipment and complete the identity authentication through the fake images. Since the shooting parameter sequence in some embodiments of this specification is randomly issued and ready to use, an attacker cannot make false images through the shooting device in advance, which greatly improves the reliability of on-site verification.

On the other hand, in the embodiment of the present specification, the change sequence of the predicted shooting parameters is compared with the change sequence of the parameters in the shooting parameter sequence to determine the matching degree, that is, the matching degree is determined by comparing the changes of the parameters, which can eliminate the influence of the environment on the specific parameter value. For example, taking the color temperature parameter as an example, the environment (such as light) will affect the determination of the color temperature parameter, which improves the accuracy of subsequent determination of authenticity through matching, thereby improving the accuracy of identity authentication.

Step 240, based on the matching degree, determine the authenticity of the original image. This step 240 may be performed by a judgment module.

In some embodiments, judging the authenticity of the original image is specifically: judging the authenticity of the image from a photographing device. According to the description of the above step 210, it can be known that the original image from the photographing device is a real image, and on the contrary, it is a fake image.

In some embodiments, the judgment module may judge the authenticity of the original image based on the matching degree. For example, when the matching degree is greater than a preset threshold, the original image is a real image.

In some embodiments, when the determination module determines that the original image is a fake image, the first computing system 140 may send a relevant instruction to the client 110 to terminate further operations of the client 110 (such as registering the application platform).

This step may be performed by other executive bodies, or performed in other transforming manners, and has no substantial impact on the technical solutions of this specification.

FIG. 4 is a schematic diagram of an exemplary structure of a machine learning model according to some embodiments of the present specification.

As shown in FIG. 4 , the machine learning model 400 may include at least a plurality of convolutional neural network units 410 and a sequence-to-sequence unit 420 . Each of the plurality of convolutional neural network units 410 may be configured to process each of the plurality of extracted images to obtain an image representation vector corresponding to each of the extracted images. 4, each of the plurality of convolutional neural network units 410 may process each of the plurality of extracted images (eg, extracted image 1 to extracted image n) extracted in the above step 240 to obtain the extracted image. The image representation vector of the image. In some embodiments, the convolutional neural network unit 410 may employ a conventional convolutional neural network including a base convolutional layer 4101 and a fully connected layer 4102, eg, LeNet, AlexNet, GoogLeNet, and the like.

The sequence-to-sequence unit 420 may process the image representation vector to obtain a sequence of changes of predicted shooting parameters of a plurality of extracted images. Specifically, the sequence-to-sequence unit 420 may process the image representation vector output by each fully connected layer 4102 in the multiple convolutional neural network units 410 to obtain a sequence of changes of predicted shooting parameters of multiple extracted images.

In some embodiments, the sequence of changes in the predicted shooting parameters of the plurality of extracted images may refer to a sequence formed by changes of the predicted shooting parameters between each of the multiple extracted images. In some embodiments, the predicted shot parameters match shot parameters included in the sequence of shot parameters. For example, if the shooting parameter sequence is a color temperature parameter sequence, the predicted shooting parameter is a color temperature parameter.

Illustratively, still take the above-mentioned multiple extracted images as the last frame image extracted from the video segments of 0-1s, 1-2s and 2-3s respectively, if the last frame of image is extracted image 1, extracted Image 2 and extracted image 3, the shooting parameter sequence is a color temperature parameter sequence, then the predicted shooting parameter change sequence can be the predicted color temperature parameter change between the extracted image 1 and the extracted image 2, and the predicted change between the extracted image 2 and the extracted image 3. A sequence of changes in the color temperature parameter. For example, the change sequence of the predicted shooting parameters output by the sequence-to-sequence unit 420 may be H={a, b}. It can be seen from the above example that a represents the change of the predicted color temperature parameter 1 to 2, and b represents the change of the predicted color temperature parameter 2 to 3.

Since the extracted image input by the convolutional neural network unit 410 may be extracted from the original image collected by the photographing device, the environmental factors (such as light and shade) of the location where the photographing device is collected may affect the color distribution of the original image, due to some Shooting parameters (such as color temperature parameters) reflect the color distribution of the image and therefore may reduce the prediction accuracy of the machine learning model.

In some embodiments, in order to solve the above problems, the parameters of the machine learning model can be adjusted by using the comparison image frames, so that the machine learning model can accurately obtain the change sequence of the predicted shooting parameters of multiple extracted images, thereby reducing the environmental factors affecting the machine learning model. 400 impact. Specifically, the comparison image frame can be a frame of image captured under the specified shooting parameters, and the image representation vector is obtained by inputting the frame image into the convolutional neural network unit 410, and then the image representation vector is obtained based on the relationship between the image representation vector and the vector of the specified shooting parameters. The difference between the parameters of the convolutional neural network unit 410 is adjusted until the obtained image representation vector is the same as the specified shooting parameter value.

FIG. 5 is an exemplary flowchart of training a machine learning model according to some embodiments of the present specification. As previously mentioned, the machine learning model 400 may be a model constructed by the base convolutional layer 4101 , the fully connected layer 4102 , and the sequence-to-sequence unit 420 . In some embodiments, the process 500 may include the following steps:

Step 510: Acquire a plurality of training samples carrying labels, where the training samples include a plurality of sample image frames obtained based on the sample shooting parameters, and the labels include a change relationship of the sample shooting parameters among the plurality of sample image frames.

In some embodiments, the training samples may be data input into the initial machine learning model for training the machine learning model. In some embodiments, the training samples may include a plurality of sample image frames acquired based on sample capture parameters. For example, still taking the sample shooting parameters as the color temperature parameters, and the color temperature parameters including color temperature parameters 1 to 6 as an example, one of the training samples may be the sample image frame 1 obtained based on the color temperature parameter 1, and the sample image frame obtained based on the color temperature parameter 2. 2. The sample image frame 3 obtained based on the color temperature parameter 5.

In some embodiments, the label may include the variation relationship of the sample capture parameters among the plurality of sample image frames. In some embodiments, the label may be a sequence of changes in sample capture parameters between multiple sample image frames. Still taking the above example as an example, the label can be c={c, d}, where c represents the change between the sample image frame 1 and the sample image frame 2 by the color temperature parameters 1 to 2, and d represents the sample image frame 2 and the sample image frame 2. Variation between sample image frames 3 by color temperature parameters 2 to 5.

Step 520: Train an initial machine learning model based on the plurality of labeled training samples to obtain the machine learning model.

According to the related description in FIG. 4 , some shooting parameters (such as color temperature parameters) reflect the color distribution of the image. Its impact on the image is global, for example, the color distribution of the same extracted image under different receptive fields should be similar or the same. However, the convolutional neural network unit 410 in the machine learning model 400 shown in FIG. 4 will focus on the contour information in the image. Therefore, in some embodiments, the machine learning model 400 shown in FIG. 4 may be improved, so that It identifies global features.

As shown in FIG. 6 , on the basis of the machine learning model 400 illustrated in FIG. 4 , the constructed machine learning model 600 may further include a first sampling convolution layer 4103 and a second sampling convolution layer 4104. In some embodiments, a first sampled convolutional layer 4103 and a second sampled convolutional layer 4104 may be added to the machine learning model 400 during training. Specifically, the basic convolution layer 4101 of the convolutional neural network unit 410 is connected to the first sampling convolution layer 4103 and the second sampling convolution layer 4104 respectively.

Through the first sampling convolution layer 4103 and the second sampling convolution layer 4104, the parameters of the machine learning model 600 can be adjusted during the training process of the machine learning model 600, thereby ensuring that the convolutional neural network unit 410 for the same extracted image in different The color distribution under the receptive field is similar or the same. Avoid the convolutional neural network unit 410 focusing on extracting contour information in the image (for example, focusing on extracting the outline of objects in the image), strengthen the convolutional neural network unit 410's ability to recognize global features, and then improve the convolutional neural network unit 410. Recognition ability to extract images.

In some embodiments, the first sampled convolutional layer 4103 and the second sampled convolutional layer 4104 may be atrous convolutional layers. In some embodiments, the convolution kernels of the first sampled convolutional layer 4103 and the second sampled convolutional layer 4104 have the same size. For example, the size of the convolution kernel of the first sampling convolution layer 4103 and the second sampling convolution layer 4104 is both 3*3. In some embodiments, the convolution sampling points of the convolution kernels of the first sampling convolution layer 4103 and the second sampling convolution layer 4104 have different spacings. For example, the sampling interval of the first sampling convolution layer 4103 is 0, and the sampling interval of the second sampling convolution layer 4104 is 2. For another example, the sampling interval of the first sampling convolution layer 4103 is 2, and the sampling interval of the second sampling convolution layer 4104 is 0.

When the sampling intervals of the first sampling convolution layer 4103 and the second sampling convolution layer 4104 are different, correspondingly, the first sampling convolution layer 4103 and the second sampling convolution layer 4104 have different receptive fields for the same extracted image . By way of example, still taking the sampling interval of the first sampling convolutional layer 4103 as 0 and the sampling interval of the second sampling convolutional layer 4104 as 2 as an example, since the sampling interval of the second sampling convolutional layer 4104 is larger, the sampling interval of the second sampling convolutional layer 4104 is larger. The receptive field of the two-sampling convolutional layer 4103 is larger than that of the second sampling convolutional layer 4104, and a wider receptive field is obtained through the second sampling convolutional layer 4104, which makes better use of the global features of the image.

In some embodiments, an initial machine learning model may be trained end-to-end based on a plurality of labeled training samples to obtain a trained machine learning model (eg, machine learning models 400 and 600 ). Specifically, the parameters of the initial machine learning model can be continuously adjusted to reduce the loss function value corresponding to each training sample, so that the loss function value satisfies the preset condition. For example, the loss function value converges, or the loss function value is smaller than a preset value. When the loss function satisfies the preset conditions, the model training is completed, and the trained machine learning model is obtained.

In some embodiments, the loss function value corresponding to each training sample may be determined through the following process: processing a plurality of sample image frames through an initial machine learning model, obtaining the change relationship of the predicted sample shooting parameters among the plurality of sample image frames, The difference between the change relationship of the sample shooting parameters and the change relationship of the sample shooting parameters in the label determines the loss function value corresponding to the training sample.

As previously mentioned, the machine learning model 600 may be a model constructed by a base convolutional layer 4101 , a fully connected layer 4102 , a first sampled convolutional layer 4103 , a second sampled convolutional layer 4104 , and a sequence-to-sequence unit 420 . In addition to determining the loss function value corresponding to the training sample in the above-mentioned manner, the model 600 can also determine the color distribution difference of each extracted image under different receptive fields based on the first sampling convolution layer 4103 and the second sampling convolution layer 4104, The loss function corresponding to the training sample is determined based on the difference.

In some embodiments, the KL divergence of the first sampling convolutional layer 4103 and the second sampling convolutional layer 4104 can be calculated based on the feature vectors output by the two to determine the color distribution difference of each extracted image under different receptive fields, It is taken as the constraint parameter of the convolutional neural network unit 410, and then the convolutional neural network unit 410 can ensure that the color distribution of the same extracted image under different receptive fields is approximate by adjusting the constraint parameter. For example, adjust the constraint parameter to a minimum value of 0.

Specifically, the constraint parameter D _KL of the convolutional neural network unit 410 can be determined by the following formula (1):

D _KL =argminKL(P _conv1 (x)||P _conv2 (x)) (1)

Among them, KL(P _conv1 (x)||P _conv2 (x)) represents the KL divergence calculation for P _conv1 (x) and P _conv2 (x), and P _conv1 (x) represents the output of the first sampling convolutional layer. Feature vector, P _conv2 (x) represents the feature vector output by the second sampling convolution layer, and argmin represents the minimum value of the KL divergence calculation. Optimally, the constraint parameter is 0. At this time, the difference between the feature vectors output by the first sampling convolution layer 4103 and the second sampling convolution layer 4104 is the smallest, and the feature distribution is the most similar. Therefore, the same extracted image has different receptive fields. The color distribution below is the same.

It can be understood that when the above-mentioned loss function constructed based on labels satisfies the preset conditions, the model training is completed, and the trained machine learning model 400 can be obtained, or when the loss functions constructed based on labels and constraint parameters satisfy the preset conditions, the model After the training is completed, the trained machine learning model 600 is obtained.

In some embodiments, the convolutional neural network unit 410 and the machine learning model 600 may perform joint training to optimize the parameters of the basic convolutional layer 4101, so that the feature vector generated by the basic convolutional layer 4101 can better reflect the in-image and the shooting parameter-related overall features, thereby improving the recognition effect of the machine learning model 600 .

FIG. 7 is another exemplary flowchart of a method for judging the authenticity of an image according to some embodiments of the present specification, and the method is applied to a client. In some embodiments, the process 700 may be implemented by the client 110 shown in FIG. 1 . As shown in FIG. 7, the process 700 may include the following steps:

Step 710: Obtain the shooting parameter sequence generated and delivered by the server.

In some embodiments, this step 710 may be performed by a second acquisition module.

In some embodiments, the shooting parameter sequence may be randomly generated on the server side. In some embodiments, the shooting parameter sequence may be randomly generated by the server based on a shooting parameter set of a shooting device; the shooting parameter set corresponds to the identification information of the shooting device. In some embodiments, the sequence of shooting parameters may include a sequence of color temperature parameters. For the specific details of step 710, reference may be made to the above-mentioned step 210 and its related description.

Step 720: Generate the original image based on the shooting parameter sequence.

In some embodiments, this step 720 may be performed by a generation module.

For the specific details of step 720, reference may be made to the above-mentioned step 210 and its related description.

Step 730: Send the original image to the server.

In some embodiments, this step 730 may be performed by a sending module.

In some embodiments, the client 110 may send the original image to the server (eg, the first computing system 140 ) through the network. For the specific details of the original image, reference may be made to the foregoing step 210, and details are not repeated here.

Step 740: Obtain the information sent by the server that includes the result of the judgment on the authenticity of the original image.

In some embodiments, the client 110 may obtain the information sent by the server (eg, the first computing system 140 ) through the network, and including the judgment result of the authenticity of the original image. In some embodiments, the information of the judging result of the authenticity of the original image may include the judging result of whether the original image is authentic.

In some embodiments, the client may also obtain verification information of the original image sent based on the judgment result. For example, whether the face or documents meet the requirements. In some embodiments, the client may also obtain relevant instructions sent by the server based on the judgment result. For example, when it is determined that the original image is a fake image, the server may send a termination instruction to terminate further operations of the client (for example, registering the application platform).

In the above, the embodiments of this specification describe the method for judging the authenticity of an image from the perspectives of the server side and the client side. Hereinafter, the embodiments of the present specification describe the method for judging the authenticity of an image from the overall perspective of the server side and the client side.

FIG. 8 is a schematic diagram of interaction between a server and a client according to some embodiments of the present specification.

As shown in FIG. 8 , the interaction between the server and the client in the schematic interaction diagram 800 includes but is not limited to: the server obtains the device model from the client, and then the server can determine the shooting parameter set included in the shooting device based on the device model of the client, A shooting parameter sequence is generated based on the shooting parameter set. The server sends the shooting parameter sequence to the client, and the shooting device of the client generates the original image based on the shooting parameter sequence. The server side obtains the original image from the client side, judges the authenticity of the original image through the methods of steps 220 to 240 above, and sends information including the judgment result to the client side.

An embodiment of the present specification further provides an apparatus for judging the authenticity of an image, the apparatus includes a processor and a memory; the memory is used for storing instructions, and the processor is used for executing the instructions, so as to achieve the above-mentioned items The operation corresponding to the method of judging the authenticity of the image.

Embodiments of this specification also provide a computer-readable storage medium, where the storage medium stores computer instructions, and after the computer reads the computer instructions in the storage medium, the computer executes the method corresponding to the method for judging the authenticity of an image described in any preceding item. operate.

The possible beneficial effects of the embodiments of this specification include, but are not limited to: (1) By generating and delivering a shooting parameter sequence to the client, the shooting device of the client generates a corresponding original image. Therefore, the original image actually shot by the shooting device The shooting parameters of the image must be consistent with the shooting parameters in the shooting parameter sequence. Since the pre-prepared fake images cannot contain exactly the same shooting parameters, the attacker can prevent the attacker from hijacking the shooting equipment and complete the identity authentication through the fake image; (2) The shooting parameter sequence is randomly issued, and it can be used immediately, and the attacker can also It is impossible to make false images through the shooting equipment in advance, which greatly improves the reliability of on-site verification; (3) The recognition ability of the convolutional neural network unit to the global features of the image is strengthened, and the recognition of the extracted image by the convolutional neural network unit is improved. ability to improve the prediction accuracy of machine learning models. It should be noted that different embodiments may have different beneficial effects, and in different embodiments, the possible beneficial effects may be any one or a combination of the above, or any other possible beneficial effects.

The basic concepts have been described above. Obviously, for those skilled in the art, the above detailed disclosure is merely an example, and does not constitute a limitation of the present specification. Although not explicitly described herein, various modifications, improvements, and corrections to this specification may occur to those skilled in the art. Such modifications, improvements, and corrections are suggested in this specification, so such modifications, improvements, and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

Meanwhile, the present specification uses specific words to describe the embodiments of the present specification. Such as "one embodiment," "an embodiment," and/or "some embodiments" means a certain feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that two or more references to "an embodiment" or "one embodiment" or "an alternative embodiment" in various places in this specification are not necessarily referring to the same embodiment . Furthermore, certain features, structures or characteristics of the one or more embodiments of this specification may be combined as appropriate.

Furthermore, those skilled in the art will appreciate that aspects of this specification may be illustrated and described in several patentable categories or situations, including any new and useful process, machine, product, or combination of matter, or combinations of them. of any new and useful improvements. Accordingly, various aspects of this specification may be performed entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software. The above hardware or software may be referred to as a "data block", "module", "engine", "unit", "component" or "system". Furthermore, aspects of this specification may be embodied as a computer product comprising computer readable program code embodied in one or more computer readable media.

A computer storage medium may contain a propagated data signal with the computer program code embodied therein, for example, on baseband or as part of a carrier wave. The propagating signal may take a variety of manifestations, including electromagnetic, optical, etc., or a suitable combination. Computer storage media can be any computer-readable media other than computer-readable storage media that can communicate, propagate, or transmit a program for use by coupling to an instruction execution system, apparatus, or device. Program code on a computer storage medium may be transmitted over any suitable medium, including radio, cable, fiber optic cable, RF, or the like, or a combination of any of the foregoing.

The computer program coding required for the operation of the various parts of this manual may be written in any one or more programming languages, including object-oriented programming languages such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python etc., conventional procedural programming languages such as C language, Visual Basic, Fortran2003, Perl, COBOL2002, PHP, ABAP, dynamic programming languages such as Python, Ruby and Groovy, or other programming languages. The program code may run entirely on the user's computer, or as a stand-alone software package on the user's computer, or partly on the user's computer and partly on a remote computer, or entirely on the remote computer or processing device. In the latter case, the remote computer can be connected to the user's computer through any network, such as a local area network (LAN) or wide area network (WAN), or to an external computer (eg, through the Internet), or in a cloud computing environment, or as a service Use eg software as a service (SaaS).

Furthermore, unless explicitly stated in the claims, the order of processing elements and sequences described in this specification, the use of alphanumerics, or the use of other names is not intended to limit the order of the processes and methods of this specification. While the foregoing disclosure discusses by way of various examples some embodiments of the invention that are presently believed to be useful, it is to be understood that such details are for purposes of illustration only and that the appended claims are not limited to the disclosed embodiments, but rather The requirements are intended to cover all modifications and equivalent combinations falling within the spirit and scope of the embodiments of this specification. For example, although the system components described above may be implemented by hardware devices, they may also be implemented by software-only solutions, such as installing the described systems on existing processing devices or mobile devices.

Similarly, it should be noted that, in order to simplify the expressions disclosed in this specification and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this specification, various features may sometimes be combined into one embodiment, in the drawings or descriptions thereof. However, this method of disclosure does not imply that the subject matter of the description requires more features than are recited in the claims. Indeed, there are fewer features of an embodiment than all of the features of a single embodiment disclosed above.

Some examples use numbers to describe quantities of ingredients and attributes, it should be understood that such numbers used to describe the examples, in some examples, use the modifiers "about", "approximately" or "substantially" to retouch. Unless stated otherwise, "about", "approximately" or "substantially" means that a variation of ±20% is allowed for the stated number. Accordingly, in some embodiments, the numerical parameters set forth in the specification and claims are approximations that can vary depending upon the desired characteristics of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and use a general digit reservation method. Notwithstanding that the numerical fields and parameters used in some embodiments of this specification to confirm the breadth of their ranges are approximations, in specific embodiments such numerical values are set as precisely as practicable.

For each patent, patent application, patent application publication, and other material, such as article, book, specification, publication, document, etc., cited in this specification, the entire contents of which are hereby incorporated by reference into this specification are hereby incorporated by reference. Application history documents that are inconsistent with or conflict with the contents of this specification are excluded, as are documents (currently or hereafter appended to this specification) limiting the broadest scope of the claims of this specification. It should be noted that, if there is any inconsistency or conflict between the descriptions, definitions and/or use of terms in the accompanying materials of this specification and the contents of this specification, the descriptions, definitions and/or use of terms in this specification shall prevail .

Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations are also possible within the scope of this specification. Accordingly, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to those expressly introduced and described in this specification.

Claims (26)

1. A method for judging the authenticity of an image, applied to the server side, wherein the method includes:

   Get the original image from the client;

   Extract multiple images or image parts in the original image as multiple extracted images according to preset extraction rules;

   Based on the multiple extracted images, determining the degree of matching between the multiple extracted images and the preset sequence by using the trained machine learning model;

   Based on the matching degree, the authenticity of the original image is judged; the preset sequence corresponds to the shooting device of the client, and judging the authenticity of the original image is specifically: judging that the image comes from the shooting device. authenticity.
2. The method of claim 1, wherein the obtaining the original image from the client comprises:

   Generate a sequence of shooting parameters;

   sending the shooting parameter sequence to the client;

   The original image is acquired from the client, and the original image is generated by the client based on the sequence of shooting parameters.
3. The method of claim 2, wherein the generating a sequence of shooting parameters comprises:

   determining the identification information of the photographing device;

   determining a shooting parameter set of the shooting device based on the identification information;

   Based on the shooting parameter set, the shooting parameter sequence is generated.
4. The method of claim 3, wherein the generating the shooting parameter sequence based on the shooting parameter set comprises:

   A preset number of shooting parameters are randomly selected from the shooting parameter set, and the shooting parameter sequence is generated based on the shooting parameters.
5. The method of claim 4, wherein the sequence of shooting parameters includes a sequence of color temperature parameters.
6. The method of claim 1, wherein the machine learning model includes at least a plurality of convolutional neural network units and a sequence-to-sequence unit;

   The determining the degree of matching between the multiple extracted images and the preset sequence based on the multiple extracted images through the trained machine learning model, including:

   Processing each of the plurality of extracted images by each of the plurality of convolutional neural network units, respectively, to obtain an image representation vector corresponding to each of the extracted images;

   The image representation vector is processed by the sequence-to-sequence unit to obtain a change sequence of the predicted shooting parameters of the multiple extracted images;

   The matching degree is determined based on the change sequence of the predicted shooting parameters and the preset sequence.
7. The method of claim 1, wherein the machine learning model is obtained by training as follows:

   Acquiring a plurality of training samples carrying labels, the training samples comprising a plurality of sample image frames obtained based on the sample shooting parameters, and the labels comprising the variation relationship of the sample shooting parameters between the plurality of sample image frames;

   An initial machine learning model is trained based on the plurality of labeled training samples to obtain the machine learning model.
8. A system for judging the authenticity of an image, applied to a server side, wherein the system includes:

   The first acquisition module is used to acquire the original image from the client;

   an extraction module, configured to extract multiple images or image parts in the original image as multiple extracted images according to a preset extraction rule;

   A determination module, configured to determine the degree of matching between the multiple extracted images and the preset sequence by using the trained machine learning model based on the multiple extracted images;

   a judging module for judging the authenticity of the original image based on the matching degree; the preset sequence corresponds to the shooting device of the client, and judging the authenticity of the original image is specifically: judging that the image comes from The authenticity of the photographing equipment.
9. The system of claim 8, wherein the first obtaining module is further configured to:

   Generate a sequence of shooting parameters;

   sending the shooting parameter sequence to the client;

   The original image is acquired from the client, and the original image is generated by the client based on the sequence of shooting parameters.
10. The system of claim 8, wherein the first obtaining module is further configured to:

    determining the identification information of the photographing device;

    determining a shooting parameter set of the shooting device based on the identification information;

    Based on the shooting parameter set, the shooting parameter sequence is generated.
11. The system of claim 8, wherein the first obtaining module is further configured to:

    A preset number of shooting parameters are randomly selected from the shooting parameter set, and the shooting parameter sequence is generated based on the shooting parameters.
12. The system of claim 11, wherein the sequence of shooting parameters includes a sequence of color temperature parameters.
13. The system of claim 8, wherein the machine learning model includes at least a plurality of convolutional neural network units and a sequence-to-sequence unit; the determining module is further configured to:

    Processing each of the plurality of extracted images by each of the plurality of convolutional neural network units, respectively, to obtain an image representation vector corresponding to each of the extracted images;

    The image representation vector is processed by the sequence-to-sequence unit to obtain a change sequence of the predicted shooting parameters of the multiple extracted images;

    The matching degree is determined based on the change sequence of the predicted shooting parameters and the preset sequence.
14. The system of claim 8, wherein the machine learning model is trained by the following methods:

    Acquiring a plurality of training samples carrying labels, the training samples comprising a plurality of sample image frames obtained based on the sample shooting parameters, and the labels comprising the variation relationship of the sample shooting parameters between the plurality of sample image frames;

    An initial machine learning model is trained based on the plurality of labeled training samples to obtain the machine learning model.
15. A method for judging the authenticity of an image, wherein, applied to a client, the method includes:

    Obtain the shooting parameter sequence generated and delivered by the server;

    generating the original image based on the sequence of shooting parameters;

    sending the original image to the server;

    Obtain the information sent by the server and including the result of the judgment on the authenticity of the original image.
16. The method according to claim 15, wherein, before the acquiring the shooting parameter sequence generated and delivered by the server, the method further comprises:

    The identification information of the client's photographing device is uploaded to the server.
17. The method of claim 15, wherein the shooting parameter sequence is randomly generated by the server.
18. The method of claim 17, wherein the shooting parameter sequence is randomly generated by the server based on a shooting parameter set of the shooting device; the shooting parameter set corresponds to the identification information of the shooting device.
19. 19. The method of claim 18, wherein the sequence of shooting parameters includes a sequence of color temperature parameters.
20. A system for judging the authenticity of an image, applied to a client, wherein the system includes:

    The second acquisition module is used to acquire the shooting parameter sequence generated and issued by the server;

    a generating module, configured to generate the original image based on the shooting parameter sequence;

    a sending module, configured to send the original image to the server;

    The third obtaining module is configured to obtain the information sent by the server and including the judgment result of the authenticity of the original image.
21. The system according to claim 20, wherein the system further comprises an uploading module, the uploading module is configured to upload the identification information of the photographing device of the client to the server.
22. The system of claim 20, wherein the shooting parameter sequence is randomly generated by the server.
23. The system of claim 22, wherein the shooting parameter sequence is randomly generated by the server based on a shooting parameter set of the shooting device; the shooting parameter set corresponds to the identification information of the shooting device.
24. 24. The system of claim 23, wherein the sequence of shooting parameters includes a sequence of color temperature parameters.
25. An apparatus for judging the authenticity of an image, the apparatus includes a processor and a memory; the memory is used for storing instructions, wherein the processor is used for executing the instructions, so as to realize the claims 1 to 7 or claim 15 Operations corresponding to the method for judging the authenticity of an image described in any one of to 19.
26. A computer-readable storage medium, the storage medium stores computer instructions, and after the computer reads the computer instructions in the storage medium, the computer executes the judgment image according to any one of claims 1 to 7 or claims 15 to 19 The method of authenticity corresponds to the operation.

Images