Image sequence identification method, device, equipment and readable storage medium

Abstract

The application discloses a kind of identification method, device and equipment of image sequence and readable storage medium, belong to image processing technical field.Method includes: obtaining image sequence, image sequence includes at least two images;First spatial feature of image sequence is extracted, first spatial feature is used to characterize the counterfeit area in each image in at least two images;First time sequence feature of image sequence is extracted, first time sequence feature is used to characterize the difference area between adjacent images in at least two images;According to first spatial feature and first time sequence feature, the authenticity of image sequence is identified, authenticity is used to characterize whether image sequence is counterfeit image sequence.The application comprehensively image sequence in each image in counterfeit area and the difference area between adjacent images, identify the authenticity of image sequence, improve the accuracy of identification result.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of image processing, in particular to a method and device for identifying an image sequence, and a readable storage medium. BACKGROUND

[0002] With the development of image processing technology, face editing software, image synthesis software and the like have gradually entered the public view. While promoting the development of the entertainment industry, they have also brought great challenges to the authenticity identification of images.

[0003] In the related art, for an image sequence such as a video, when identifying the authenticity of the image sequence, a single image is first selected from the image sequence, and then a spatial feature of the image is extracted. The spatial feature is used to represent a forged region in the image. The authenticity of the image is identified based on the spatial feature, and the authenticity identification result of the image is taken as the authenticity identification result of the image sequence.

[0004] Since the selection of the image from the image sequence is random, the authenticity identification result of the single image is less accurate when taken as the authenticity identification result of the image sequence. SUMMARY

[0005] Embodiments of the present application provide a method and device for identifying an image sequence, and a readable storage medium, which can be used to solve the problem of low accuracy of the authenticity identification result of the image sequence. The technical solution includes the following contents.

[0006] In one aspect, the present application provides a method for identifying an image sequence, which includes:

[0007] obtaining an image sequence, the image sequence including at least two images;

[0008] extracting a first spatial feature of the image sequence, the first spatial feature being used to represent a forged region in each image of the at least two images;

[0009] extracting a first time sequence feature of the image sequence, the first time sequence feature being used to represent a difference region between adjacent images in the at least two images;

[0010] identifying the authenticity of the image sequence according to the first spatial feature and the first time sequence feature, the authenticity being used to represent whether the image sequence is a forged image sequence.

[0011] In another aspect, the present application provides a device for identifying an image sequence, which includes:

[0012] an obtaining module configured to obtain an image sequence, the image sequence including at least two images;

[0013] extract a first spatial feature of the image sequence, the first spatial feature being used to represent a forged region in each of the at least two images;

[0014] extract a first temporal feature of the image sequence, the first temporal feature being used to represent a difference region between adjacent images in the at least two images;

[0015] discriminate a reality of the image sequence according to the first spatial feature and the first temporal feature, the reality being used to represent whether the image sequence is a forged image sequence.

[0016] In a possible implementation, the apparatus further includes:

[0017] The acquisition module is further configured to acquire a sequence feature of the image sequence, the sequence feature being used to represent image information of each of the at least two images;

[0018] The segmentation module is configured to segment the sequence feature based on channel information to obtain a first channel feature and a second channel feature.

[0019] The extraction module is configured to acquire the first spatial feature of the image sequence based on the first channel feature.

[0020] The extraction module is configured to acquire the first temporal feature of the image sequence based on the second channel feature.

[0021] In a possible implementation, the extraction module is configured to acquire the first temporal feature of the image sequence based on the second channel feature and the first spatial feature.

[0022] In a possible implementation, the extraction module is configured to acquire a first sub-feature of the image sequence based on the second channel feature, acquire a second sub-feature of the image sequence based on the first spatial feature, and fuse the first sub-feature and the second sub-feature to obtain the first temporal feature of the image sequence.

[0023] In a possible implementation, the extraction module is configured to perform feature processing on the second channel feature in a first direction to obtain a first direction feature, perform feature processing on the second channel feature in a second direction to obtain a second direction feature, and fuse the first direction feature, the second direction feature, and the second channel feature to obtain the first sub-feature of the image sequence.

[0024] In a possible implementation, the discrimination module is configured to: fuse the first spatial feature and the first temporal feature to obtain a first fused feature; determine the authenticity probability of the image sequence based on the first fused feature; and determine the authenticity of the image sequence based on the authenticity probability of the image sequence.

[0025] In a possible implementation, the first spatial feature and the first temporal feature are obtained based on a sequence feature of the image sequence, and the discrimination module is configured to fuse the first spatial feature, the first temporal feature, and the sequence feature to obtain a first fused feature.

[0026] In a possible implementation, the discrimination module is configured to: split the first fused feature based on channel information to obtain a third channel feature and a fourth channel feature; obtain a second spatial feature of the image sequence based on the third channel feature; obtain a second temporal feature of the image sequence based on the fourth channel feature; fuse the second spatial feature and the second temporal feature to obtain a second fused feature; and determine the authenticity probability of the image sequence based on the second fused feature.

[0027] In a possible implementation, the first spatial feature and the first temporal feature are extracted based on a space-time model, and the apparatus further includes:

[0028] The obtaining module is configured to obtain a sample image sequence and an authenticity label result of the sample image sequence, the sample image sequence including at least two sample images.

[0029] The extraction module is configured to extract a sample spatial feature of the sample image sequence, the sample spatial feature being used to represent a forged region in each sample image in the at least two sample images.

[0030] The extraction module is configured to extract a sample temporal feature of the sample image sequence, the sample temporal feature being used to represent a difference region between adjacent sample images in the at least two sample images.

[0031] The determination module is configured to determine an authenticity probability of the sample image sequence based on the sample spatial feature and the sample temporal feature.

[0032] The obtaining module is configured to obtain the space-time model based on the authenticity probability of the sample image sequence and the authenticity label result.

[0033] In a possible implementation, the image sequence is obtained based on a video, and the apparatus further includes:

[0034] recommend the video corresponding to the image sequence in response to the authenticity of the image sequence being non-forged;

[0035] filter out the video corresponding to the image sequence in response to the authenticity of the image sequence being forged.

[0036] In a possible implementation, the image sequence is an image sequence containing a face, and the acquisition module is configured to acquire the image sequence in response to a triggered target request, the target request being a payment request or an identity verification request.

[0037] The apparatus further includes:

[0038] The acquisition module is further configured to acquire a target face in response to the authenticity of the image sequence being non-forged, and perform an operation corresponding to the target request based on a face contained in the image sequence and the target face.

[0039] The generation module is configured to generate a prompt information in response to the authenticity of the image sequence being forged, the prompt information being used to prompt a failure of an operation based on the target request.

[0040] In a possible implementation, the acquisition module is configured to extract at least two target frame images from a video, the target frame images being images containing a target object, and acquire the image sequence based on the at least two target frame images.

[0041] In a possible implementation, the acquisition module is configured to determine a target region in each of the at least two target frame images, the target region being a region containing the target object, and cut the target region in each of the at least two target frame images to obtain the image sequence.

[0042] In a possible implementation, the acquisition module is configured to perform image recognition on each of the at least two target frame images to obtain region information of the target object in each of the at least two target frame images, and expand the region information of the target object in each of the at least two target frame images according to an expansion ratio to obtain a target region in each of the at least two target frame images.

[0043] In another aspect, an embodiment of the present application provides an electronic device, including a processor and a memory, the memory storing at least one program code, the at least one program code being loaded and executed by the processor to enable the electronic device to implement the image sequence identification method described above.

[0044] In another aspect, a computer readable storage medium is provided, the computer readable storage medium storing at least one program code, the at least one program code being loaded and executed by a processor to enable a computer to implement any of the above-described image sequence identification methods.

[0045] In another aspect, a computer program or computer program product is provided, the computer program or computer program product storing at least one computer instruction, the at least one computer instruction being loaded and executed by a processor to enable a computer to implement any of the above-described image sequence identification methods.

[0046] The technical solutions provided by the embodiments of the present application at least bring the following beneficial effects:

[0047] The technical solutions provided by the embodiments of the present application identify the authenticity of the image sequence according to the first spatial feature and the first time sequence feature, the image sequence includes at least two images, the first spatial feature is used to represent the forged area in each image, and the first time sequence feature is used to represent the difference area between adjacent images in the at least two images, thereby comprehensively considering the forged area in each image and the difference area between adjacent images in the image sequence, identifying the authenticity of the image sequence, and improving the accuracy of the identification result. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Figure 1 is an implementation environment schematic diagram of an image sequence identification method provided by the embodiments of the present application;

[0050] Figure 2 is a flowchart of an image sequence identification method provided by the embodiments of the present application;

[0051] Figure 3 is a structure schematic diagram of a spatial sub-model provided by the embodiments of the present application;

[0052] Figure 4 is an extraction schematic diagram of a first time sequence feature provided by the embodiments of the present application;

[0053] Figure 5 is a structure schematic diagram of an information supplement sub-model provided by the embodiments of the present application;

[0054] Figure 6 is a structure schematic diagram of a first direction feature extraction part provided by the embodiments of the present application;

[0055] Figure 7 is a structural schematic diagram of a second direction feature extraction part provided by an embodiment of the present application;

[0056] Figure 8 is a structural schematic diagram of a spatio-temporal model provided by an embodiment of the present application;

[0057] Figure 9 is a schematic diagram of an image region represented by a fused feature provided by an embodiment of the present application;

[0058] Figure 10 is a structural schematic diagram of an image sequence identification device provided by an embodiment of the present application;

[0059] Figure 11 is a structural schematic diagram of a terminal device provided by an embodiment of the present application;

[0060] Figure 12 is a structural schematic diagram of a server provided by an embodiment of the present application. DETAILED DESCRIPTION

[0061] To make the objectives, technical solutions and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0062] Figure 1 is an implementation environment schematic diagram of an image sequence identification method provided by an embodiment of the present application, as shown in the implementation environment includes an electronic device 11, the image sequence identification method in the embodiment of the present application can be executed by the electronic device 11. Exemplarily, the electronic device 11 can include at least one of a terminal device or a server. Figure 1

[0063] The terminal device can be at least one of a smart phone, a game console, a desktop computer, a tablet computer, an electronic book reader, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player and a laptop computer.

[0064] ​The server can be one server, or a server cluster composed of multiple servers, or any one of a cloud computing platform and a virtualization center, and the embodiments of the present application do not limit this. The server can be in communication connection with the terminal device through a wired network or a wireless network. The server can have functions of data processing, data storage, and data transceiving, and the embodiments of the present application do not limit this.

[0065] The image sequence identification method of the embodiments of the present application can be implemented based on artificial intelligence technology. Artificial intelligence (AI) is the theory, method, technology and application system for using a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology of computer science, which attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines, so that the machine has the functions of perception, reasoning and decision-making.

[0066] Artificial intelligence technology is a comprehensive discipline, involving a wide range of fields, both hardware and software technologies. Artificial intelligence basic technologies generally include technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0067] Based on the above implementation environment, the embodiments of the present application provide an image sequence identification method, which can be implemented by the electronic device 11 in the system 10 shown in Figure 2 The flowchart of the image sequence identification method provided by the embodiments of the present application is taken as an example, which can be executed by the electronic device 11 in the system 10 shown in Figure 1 As shown in Figure 2 The method includes steps 21-24.

[0068] Step 21, obtaining an image sequence, the image sequence including at least two images.

[0069] In the embodiments of the present application, each image in the image sequence corresponds to a unique sequence number, and the sequence number corresponding to the image is used to represent the position of the image in the image sequence.

[0070] For example, the image sequence includes image A, image B and image C, wherein the sequence number corresponding to image A is 2, the sequence number corresponding to image B is 1, and the sequence number corresponding to image C is 3. Therefore, the order of the images in the image sequence is image B, image A and image C.

[0071] In a possible implementation, the image sequence is acquired by: extracting at least two target frame images from the video, the target frame image being an image containing a target object; and acquiring the image sequence based on the at least two target frame images.

[0072] The target object includes, but is not limited to, a person, a face, a vehicle, a building, and the like. Image recognition is performed on each frame image in the video to obtain an object contained in each frame image, and a target frame image containing the target object is determined based on the object contained in each frame image, thereby achieving extraction of the at least two target frame images from the video. In this embodiment of the present application, the image recognition manner is not limited, and for example, a neural network model, an image processing detection method (Oriented FAST And Rotated BRIEF, ORB), a scale-invariant feature transform (Scale-invariant Feature Transform, SIFT), or the like is used to perform image recognition on each frame image in the video.

[0073] When the at least two target frame images are extracted from the video, a plurality of frame images to be screened can be first sampled at equal intervals from the video, and then image recognition is performed on each frame image to be screened, and a target frame image containing the target object is determined based on an object contained in each frame image to be screened.

[0074] For example, starting from the first frame of the video, one frame image is extracted every 10 frames as a frame image to be screened, thereby obtaining 150 frame images to be screened, image recognition is performed on the 150 frame images to be screened, and a target frame image containing a face is determined based on an object contained in each of the 150 frame images to be screened.

[0075] In a possible implementation, an order number of each target frame image is determined, and the image sequence is determined based on the order number of each target frame image. The order number of each target frame image can be determined based on a position of each target frame image in the video.

[0076] For example, the third frame image in the video is a target frame image A, the fourth frame image in the video is a target frame image B, and the sixth frame image in the video is a target frame image C, and the order numbers of the target frame images A-C are 1-3 in sequence, that is, the order of each image in the image sequence is the target frame image A, the target frame image B, and the target frame image C.

[0077] In another possible implementation, the image sequence is acquired based on the at least two target frame images, including: determining a target region in each target frame image of the at least two target frame images, the target region being a region containing the target object; and cutting the target region in each target frame image to obtain the image sequence.

[0078] In the embodiments of the present application, after identifying each target frame image through image recognition technology, for each target frame image, the target region in the target frame image is determined. The target region includes but is not limited to the region of the target object, and the size and shape of the target region are not limited. For example, the target region includes the region of the target object and the background region, and the target region is a regular closed shape such as a rectangle or a circle, or an irregular closed shape.

[0079] In the embodiments of the present application, after identifying each target frame image through image recognition technology, for each target frame image, the target region in the target frame image is determined. The target region includes but is not limited to the region of the target object, and the size and shape of the target region are not limited. For example, the target region includes the region of the target object and the background region, and the target region is a regular closed shape such as a rectangle or a circle, or an irregular closed shape.

[0080] The region information is the position information of the region where the target object is located. For example, if the region where the target object is located is a rectangle, the region information is the coordinates of three vertices, or the coordinates of four vertices, or the coordinates of one vertex and the lengths of two edges. If the target object is a circle, the region information is the center coordinates and the radius, or the center coordinates and the diameter.

[0081] In a possible implementation, each target frame image is identified through a multi-task cascaded convolutional network (MTCNN) to obtain the region information of the target object in each target frame image. The region information is the position information of a rectangle. The region information of the target object in each target frame image is expanded according to an expansion ratio to obtain the target region in each target frame image. The target region includes the region where the target object is located and the background region. In the embodiments of the present application, the size of the expansion ratio is not limited. For example, the expansion ratio is 1.2 times. That is, the region information of the target object in each target frame image is expanded by 1.2 times to obtain the target region in each target frame image.

[0082] In another possible implementation, after identifying the region information of the target object in each target frame image, the region information of the target object is taken as the target region, that is, the target region in each target frame image is directly identified.

[0083] After obtaining the target region in each target frame image, each target region is cut from the target region in each target frame image to obtain each target region. The sorting number of each target region is determined, and the image sequence is determined based on the sorting number of each target region. The sorting number of the target region is determined based on the position of the target frame image corresponding to the target region in the video.

[0084] For example, the third frame image in the video is a target frame image A, the target frame image A includes a target region a, the fourth frame image in the video is a target frame image B, the target frame image B includes a target region b, and the sixth frame image in the video is a target frame image C, the target frame image C includes a target region c. The respective ordering numbers of the target regions a-c are 1-3 in turn, that is, the order of the images in the image sequence is target region a, target region b, and target region c.

[0085] It should be noted that for any target frame image, the target frame image includes at least one target object, and when the target frame image includes at least two target objects, region information in which each target object is located is identified. For each target object, region information in which the target object is located in at least two target frame images is used to obtain an image sequence corresponding to the target object. The process of obtaining the image sequence corresponding to the target object is described in the related description of “extracting at least two target frame images from a video and obtaining an image sequence based on the at least two target frame images”, which is not repeated here.

[0086] In the embodiment of the present application, after obtaining the image sequence, the method further includes: obtaining sequence features of the image sequence, the sequence features being used to represent image information of each image; and performing segmentation on the sequence features based on channel information to obtain first channel features and second channel features.

[0087] After obtaining the image sequence, sequence features of the image sequence are extracted, and the sequence features are represented as [c, t, h, w]. The sequence features are segmented based on channel information to obtain first channel features and second channel features. For example, the sequence features are segmented along a channel direction to obtain the first channel features and the second channel features. The first channel features are represented as X1, and the second channel features are represented as X2. X1 and X2 are both [c / 2, t, h, w]. Wherein, c represents a channel number of the image sequence, t represents an image number of the image sequence, h represents a height of an image in the image sequence, and w represents a width of the image in the image sequence.

[0088] In the embodiment of the present application, the first channel features are used to extract first spatial features of the image sequence, and the second channel features are used to extract first time sequence features of the image sequence. Details are described below in relation to steps 22 and 23.

[0089] Step 22, first spatial features of the image sequence are extracted, and the first spatial features are used to represent a forged region in each image of the at least two images.

[0090] The embodiments of the present application do not limit the extraction manner of the spatial features (such as the first spatial feature, the second spatial feature mentioned below, the sample spatial feature, etc.). For example, the first spatial feature of the image sequence is extracted based on a spatial sub-model, which is also called a spatial inconsistency module (SIM), and is used to extract the spatial feature representing the forged region in each image in the image sequence through the attention mechanism in the space.

[0091] In a possible implementation, the first spatial feature of the image sequence is extracted, including: obtaining the first spatial feature of the image sequence based on the first channel feature.

[0092] As shown in Figure 3 , Figure 3 is a structural schematic diagram of a spatial sub-model provided by an embodiment of the present application. The first channel feature is input into the spatial sub-model. On the one hand, the first channel feature is subjected to 3x3 convolution processing to obtain the first channel feature after the convolution processing. On the other hand, the first channel feature is sequentially subjected to down-sampling processing, 1x3 second direction convolution processing, 3x1 first direction convolution processing and up-sampling processing to obtain the processed first channel feature. After the processed first channel feature and the first channel feature without processing are subjected to an addition operation (also called skip connection processing), normalization processing is performed to obtain the first channel feature after the normalization processing. After the first channel feature after the convolution processing and the first channel feature after the normalization processing are subjected to point multiplication, 3x3 convolution processing is performed to obtain the first spatial feature.

[0093] In the embodiments of the present application and the following embodiments, the down-sampling processing is to expand the receptive field, the 1x3 second direction convolution processing is to extract the information in the second direction, the 3x1 first direction convolution processing is to extract the information in the first direction, the skip connection processing is to make up for the information loss caused by the up-sampling processing, and the normalization processing can be implemented based on the Sigmoid function and is used to assign weights to each spatial information.

[0094] In step 23, the first temporal feature of the image sequence is extracted, and the first temporal feature is used to represent the difference region between adjacent images in the at least two images.

[0095] The embodiments of the present application do not limit the extraction manner of the temporal features (such as the first temporal feature, the second temporal feature mentioned below, the sample temporal feature, etc.). For example, the temporal feature of the image sequence is extracted based on a temporal sub-model, which is also called a temporal inconsistency module (TIM), and is used to extract the temporal feature representing the difference region between adjacent images in the image sequence through the attention mechanism in the time sequence.

[0096] The first temporal feature of the image sequence is extracted based on the second channel feature.

[0097] In the embodiments of the present application, the first temporal feature of the image sequence is extracted based on the second channel feature and the temporal sub-model. In a possible implementation, the first temporal feature of the image sequence is obtained based on the second channel feature and the first spatial feature, including: obtaining a first sub-feature of the image sequence based on the second channel feature; obtaining a second sub-feature of the image sequence based on the first spatial feature; and fusing the first sub-feature and the second sub-feature to obtain the first temporal feature of the image sequence.

[0098] In the embodiments of the present application, the first temporal feature of the image sequence is obtained based on the second channel feature and the temporal sub-model, and the first spatial feature and the information supplement sub-model. The information supplement sub-model, also referred to as an information supplement module (ISM), can selectively extract useful temporal features from spatial features.

[0099] In a possible implementation, the first temporal feature of the image sequence is obtained based on the second channel feature and the first spatial feature, including: obtaining a first sub-feature of the image sequence based on the second channel feature; obtaining a second sub-feature of the image sequence based on the first spatial feature; and fusing the first sub-feature and the second sub-feature to obtain the first temporal feature of the image sequence.

[0100] In the embodiments of the present application, the second channel feature is input into the temporal sub-model, and the first sub-feature of the image sequence is output by the temporal sub-model. The first spatial feature is input into the information supplement sub-model, and the second sub-feature of the image sequence is output by the information supplement sub-model. Then, the first sub-feature and the second sub-feature are fused to obtain the first temporal feature of the image sequence.

[0101] As Figure 4 described, Figure 4 is a schematic diagram of the extraction of the first temporal feature provided by the embodiments of the present application. The second channel feature is input into the temporal sub-model, and the temporal sub-model includes a first direction feature extraction part and a second direction feature extraction part. On the one hand, the second channel feature passes through the first direction feature extraction part to obtain a first direction feature, and on the other hand, the second channel feature passes through the second direction feature extraction part to obtain a second direction feature. The first direction feature and the second direction feature are added, and the added features are multiplied by the second channel feature. The multiplied features are subjected to a 3x1x1 convolution process to obtain a first sub-feature, and the first sub-feature is output by the temporal sub-model. The first spatial feature is input into the information supplement sub-model, and the second sub-feature is output by the information supplement sub-model. The first sub-feature and the second sub-feature are added, and after the addition, the added features are subjected to a 3x3 convolution process to obtain the first temporal feature.

[0102] wherein, asFigure 5 As shown, Figure 5 This is a schematic diagram of the structure of an information supplementation sub-model provided in an embodiment of this application. After the first spatial feature is input into the information supplementation sub-model, the information supplementation sub-model first performs average aggregation processing on the first spatial feature to obtain a global representation. This global representation is then processed by a one-dimensional convolution in a single channel direction, followed by normalization processing and weight allocation. After that, it is multiplied by the first spatial feature. The feature after the multiplication is then processed by a 3×1×1 convolution to obtain the second sub-feature, which is then output by the information supplementation sub-model.

[0103] It should be noted that, in this embodiment, adding the first directional feature and the second directional feature and then multiplying them by the second channel feature is to fuse low-to-mid-level features with high-level features, thereby improving feature accuracy. The first directional feature and the second directional feature are perpendicular in direction; for example, the first directional feature is a vertical feature, and the second directional feature is a horizontal feature.

[0104] In one possible implementation, obtaining the first sub-feature of the image sequence based on the second channel features includes: performing feature processing on the second channel features in a first direction to obtain first-direction features; performing feature processing on the second channel features in a second direction to obtain second-direction features; and fusing the first-direction features, the second-direction features, and the second channel features to obtain the first sub-feature of the image sequence.

[0105] In this embodiment, the second channel features are processed by a first direction feature extraction section, which performs first direction feature processing on the second channel features to obtain first direction features. For example, the first direction is the vertical direction.

[0106] like Figure 6 As shown, Figure 6 This is a schematic diagram of the structure of a first-direction feature extraction part provided in an embodiment of this application. When the second channel feature passes through the first-direction feature extraction part, it can first undergo 1×1 convolution processing to reduce the channel dimension of the second channel feature, thereby reducing the number of parameters and improving computational efficiency. Then, the second channel feature is reshaped, or the second channel feature can be reshaped directly. As mentioned in the description of "segmenting sequence features based on channel information to obtain the first channel feature and the second channel feature", the second channel feature is [c / 2, t, h, w], and the second channel feature after the reshaping operation is [w, c / 2r, h, t].

[0107] Next, the second channel features after the reshaping operation are subjected to forward differential processing along the t direction to obtain differential features, which are [w, c / 2r, h, t-1]. The differential features at time t are defined as 0. At this time, the differential features are represented as [w, c / 2r, h, t]. These differential features represent the changes in the image sequence in the first direction.

[0108] To more accurately extract the changes in the image sequence in the first direction, the difference feature is processed by a 3×1 first-direction convolution to obtain the difference feature after first-direction convolution. Then, the difference feature is processed by downsampling, followed by a 3×1 first-direction convolution and upsampling to obtain the upsampled difference feature. This difference feature, the difference feature after first-direction convolution, and the difference feature after upsampling are then added together (i.e., skip-connection processing) to fuse low- and mid-level features with high-level features, resulting in the added difference feature. Finally, the added difference feature is processed by a 1×1 convolution and normalization to obtain the first-direction feature.

[0109] Where c represents the number of channels in the image sequence, t represents the number of images in the image sequence, h represents the height of the images in the image sequence, w represents the width of the images in the image sequence, and r represents the channel compression ratio, i.e., the factor by which the channel dimension is reduced. The difference features are processed by a 3×1 first-direction convolution, followed by downsampling, a 3×1 first-direction convolution, and upsampling to extract multi-scale information. The summed difference features are then processed by a 1×1 convolution to increase the channel dimension to c / 2.

[0110] In this embodiment of the application, the second channel feature is processed by the second direction feature extraction part, and the second direction feature extraction part performs second direction feature processing on the second channel feature to obtain the second direction feature.

[0111] like Figure 7 As shown, Figure 7 This is a schematic diagram of the structure of a second-direction feature extraction part provided in an embodiment of this application. When the second channel features pass through the second-direction feature extraction part, they can first undergo 1×1 convolution processing to reduce the channel dimension of the second channel features, thereby reducing the number of parameters and improving computational efficiency. Then, the second channel features are reshaped, or the reshaping operation can be performed directly. Next, the reshaped second channel features are subjected to forward differencing along the t-direction to obtain difference features, which represent the changes in the image sequence in the second direction.

[0112] In order to more accurately extract the change of the image sequence in the second direction, on the one hand, the differential feature is subjected to 1x3 second direction convolution processing to obtain a differential feature after second direction convolution processing. On the other hand, the differential feature is subjected to down-sampling processing, 1x3 second direction convolution processing and up-sampling processing in sequence to obtain a differential feature after up-sampling processing. The differential feature, the differential feature after second direction convolution processing and the differential feature after up-sampling processing are added (i.e. skip connection processing) to fuse the middle-low layer feature and the high layer feature to obtain an added differential feature. The added differential feature is subjected to 1x1 convolution processing and normalization processing in sequence to obtain a second direction feature.

[0113] The principle of the second direction feature extraction part is similar to the principle of the first direction feature extraction part, and can be seen in the related description of the first direction feature extraction part, which will not be repeated here.

[0114] In the embodiment of the application, the first direction feature, the second direction feature and the second channel feature are fused to obtain the first sub-feature of the image sequence. That is, in the description of "obtaining the first sub-feature of the image sequence based on the second channel feature", the first direction feature and the second direction feature are added, the added features are multiplied with the second channel feature, and the multiplied features are subjected to 3x1x1 convolution processing to obtain the first sub-feature.

[0115] After obtaining the first sub-feature, the first sub-feature and the second sub-feature are fused, that is, the first sub-feature and the second sub-feature are added, and after the addition, the added features are subjected to 3x3 convolution processing to obtain the first time sequence feature of the image sequence.

[0116] In step 24, the authenticity of the image sequence is identified according to the first spatial feature and the first time sequence feature. The authenticity is used to represent whether the image sequence is a fake image sequence.

[0117] In the embodiment of the application, the manner of identifying the authenticity of the image sequence according to the first spatial feature and the first time sequence feature is not limited. In a possible implementation manner, the authenticity of the image sequence is identified according to the first spatial feature and the first time sequence feature, including: fusing the first spatial feature and the first time sequence feature to obtain a first fused feature; determining the authenticity probability of the image sequence based on the first fused feature; and determining the authenticity of the image sequence based on the authenticity probability of the image sequence.

[0118] The first spatial feature and the first temporal feature are spliced according to channel information such as a channel direction, so as to realize fusion of the first spatial feature and the first temporal feature, and obtain the first fusion feature. In a possible implementation manner, the first spatial feature and the first temporal feature are obtained based on sequence features of the image sequence, and the fusion of the first spatial feature and the first temporal feature obtains the first fusion feature, including: fusing the first spatial feature, the first temporal feature, and the sequence features to obtain the first fusion feature.

[0119] In the embodiment of the application, the first spatial feature and the first temporal feature are spliced according to channel information such as a channel direction, and the spliced features and the sequence features are added, so as to realize fusion of the first spatial feature, the first temporal feature, and the sequence features, and obtain the first fusion feature.

[0120] Then, the image sequence is classified based on the first fusion feature, and a reality probability of the image sequence is obtained. The reality probability is used to represent a probability that the image sequence is a fake image sequence, and the reality probability is greater than or equal to 0 and less than or equal to 1. The manner of determining the reality probability of the image sequence based on the first fusion feature is not limited.

[0121] In a possible implementation manner, the first fusion feature is input into a full connection layer, and the full connection layer outputs the reality probability of the image sequence.

[0122] After that, when the reality probability of the image sequence is greater than or equal to a target probability, it is determined that the reality of the image sequence is a fake image sequence, and when the reality probability of the image sequence is less than the target probability, it is determined that the reality of the image sequence is a non-fake image sequence. The value of the target probability is not limited and is flexibly adjusted according to an actual application scene. For example, the target probability is 0.5.

[0123] It can be understood that when the reality probability is used to represent a probability that the image sequence is a non-fake image sequence, when the reality probability of the image sequence is greater than or equal to the target probability, it is determined that the reality of the image sequence is a non-fake image sequence, and when the reality probability of the image sequence is less than the target probability, it is determined that the reality of the image sequence is a fake image sequence.

[0124] In a possible implementation manner, the image sequence is classified according to the first spatial feature, and a first reality probability of the image sequence is obtained. The image sequence is classified according to the first temporal feature, and a second reality probability of the image sequence is obtained. The reality of the image sequence is determined based on the first reality probability and the second reality probability.

[0125] The first authenticity probability and the second authenticity probability are different authenticity probabilities. The authenticity of the image sequence can be determined based on the first authenticity probability, and the authenticity of the image sequence can be determined based on the second authenticity probability. Details of how to determine the authenticity of the image sequence based on the authenticity probability are described above and will not be repeated here.

[0126] When the authenticity of the image sequence is determined to be fake based on the first authenticity probability and / or the authenticity of the image sequence is determined to be fake based on the second authenticity probability, the authenticity of the image sequence is fake. When the authenticity of the image sequence is determined to be non-fake based on the first authenticity probability and the authenticity of the image sequence is determined to be non-fake based on the second authenticity probability, the authenticity of the image sequence is non-fake.

[0127] In another possible implementation, determining the authenticity probability of the image sequence based on the first fusion feature includes: splitting the first fusion feature based on channel information to obtain third channel features and fourth channel features; obtaining second spatial features of the image sequence based on the third channel features; obtaining second temporal features of the image sequence based on the fourth channel features; fusing the second spatial features and the second temporal features to obtain second fusion features; and determining the authenticity probability of the image sequence based on the second fusion features.

[0128] In the embodiments of the present application, the first fusion feature is split based on channel information, for example, the first fusion feature is split along the channel direction to obtain third channel features and fourth channel features. Details of splitting the sequence features to obtain the first channel features and the second channel features are described above. The second spatial features of the image sequence are obtained based on the third channel features. Details of obtaining the first spatial features of the image sequence based on the first channel features are described above. The second temporal features of the image sequence are obtained based on the fourth channel features. Details of obtaining the first temporal features of the image sequence based on the second channel features are described above. The second fusion features are obtained by fusing the second spatial features and the second temporal features. Details of fusing the first spatial features and the first temporal features to obtain the first fusion features are described above. The authenticity probability of the image sequence is determined based on the second fusion features. Details of determining the authenticity probability of the image sequence based on the first fusion features are described above. The implementation principles of the above contents are similar and will not be repeated here.

[0129] By splitting, processing, and fusing the first fusion feature, the obtained second fusion feature is more accurate, so as to improve the identification accuracy of the image sequence. In order to further improve the identification accuracy of the image sequence, the second fusion feature can be split, processed, and fused multiple times, which is not limited in the embodiments of the present application.

[0130] In a possible implementation, the first spatial feature and the first temporal feature are extracted based on the spatio-temporal model, and before the first spatial feature of the image sequence is extracted, the method further includes: obtaining a sample image sequence and a truth annotation result of the sample image sequence, the sample image sequence including at least two sample images; extracting a sample spatial feature of the sample image sequence, the sample spatial feature being used to represent a forged region in each sample image in the at least two sample images; extracting a sample temporal feature of the sample image sequence, the sample temporal feature being used to represent a difference region between adjacent sample images in the at least two sample images; determining a truth probability of the sample image sequence based on the sample spatial feature and the sample temporal feature; and obtaining the spatio-temporal model based on the truth probability of the sample image sequence and the truth annotation result.

[0131] In the embodiments of the present application, the spatial features (including the sample spatial feature and the first spatial feature, the second spatial feature, etc.) and the temporal features (including the sample temporal feature and the first temporal feature, the second temporal feature, etc.) are extracted by using the spatio-temporal model. When the spatio-temporal model is trained, a sample image sequence is obtained from a sample video, and a truth annotation result of the sample image sequence is obtained. The manner of obtaining the sample image sequence is described in step 21, and the implementation principles are the same, which will not be described here. The truth annotation result is either forged or non-forged. For example, 0 represents forged, and 1 represents non-forged.

[0132] For example, when the sample image sequence is constructed, 150 frames of images are sampled at equal intervals from a sample face video, target frame images containing faces are selected from the 150 frames of images, image recognition is performed on each target frame image to obtain a region where a face is located in each target frame image, the region is enlarged by 1.2 times to obtain a target region, the target region is cut so that the target region contains the entire face and part of the surrounding background, and in a manner that each 8 target regions construct a sample image sequence, a plurality of sample image sequences are obtained.

[0133] The manner of extracting the sample spatial feature of the sample image sequence is described in relation to the extraction of the first spatial feature of the image sequence. The manner of extracting the sample temporal feature of the sample image sequence is described in relation to the extraction of the first temporal feature of the image sequence. The manner of determining the truth probability of the sample image sequence based on the sample spatial feature and the sample temporal feature is described in relation to the fusion of the first spatial feature and the first temporal feature to obtain the first fusion feature and the determination of the truth probability of the image sequence based on the first fusion feature. The implementation principles of the above contents are similar, which will not be described here.

[0134] In the embodiments of the present application, the loss value of the model is calculated based on the authenticity probability of the sample image sequence and the authenticity annotation result, and the model parameters are optimized using the loss value to obtain the spatio-temporal model. The loss function is not limited. For example, the loss function is an adaptive moment estimation (Adam) algorithm.

[0135] In an example provided by the embodiments of the present application, the authenticity of the sample image sequence is manually annotated when training the spatio-temporal model. The sample image sequence includes 8 images, the channel compression rate is set to 16, the size of the images in the sample image sequence is 224*224, the Adam algorithm is used as the loss function to optimize the network parameters, the training frequency is 30 times, the batch size is equal to 16, the initial learning rate is 0.0002, and the learning rate is reduced by one-tenth every ten cycles. The number of sample image sequences is expanded based on the way of flipping images in several directions. The several directions are at least one. For example, the number of sample image sequences is expanded based on the way of flipping images in at least one of the horizontal direction or the vertical direction.

[0136] In a possible implementation, the image sequence is obtained based on a video. After identifying the authenticity of the image sequence based on the first spatial feature and the first time sequence feature, the method further includes: in response to the authenticity of the image sequence being non-fake, recommending the video corresponding to the image sequence to a user; and in response to the authenticity of the image sequence being fake, filtering out the video corresponding to the image sequence.

[0137] The image sequence identification method provided by the embodiments of the present application can be applied to a video recommendation scene. In the video recommendation scene, for each video to be recommended, an image sequence is obtained based on the video, and the authenticity of the image sequence is identified in the manner of steps 21-24.

[0138] When the authenticity of the image sequence is non-fake, the video corresponding to the image sequence is recommended to a user, and the video is displayed on a recommendation page. When the authenticity of the image sequence is fake, the video corresponding to the image sequence is filtered out, and the video is prevented from being displayed on the recommendation page.

[0139] For example, the image sequence a is obtained based on a video A, the image sequence b is obtained based on a video B, and the image sequence c is obtained based on a video C. The authenticity of the image sequence a and the image sequence c is non-fake, and the authenticity of the image sequence b is fake. Then, the video A and the video C are recommended to a user, that is, the video A and the video C are displayed on a recommendation page. At the same time, the video B is filtered out to avoid being displayed on the recommendation page.

[0140] In a possible implementation, the image sequence is an image sequence containing a face, and the obtaining of the image sequence comprises: obtaining the image sequence in response to a triggered target request, the target request being a payment request or an identity verification request; and after the authenticity of the image sequence is identified according to the first spatial feature and the first time sequence feature, the method further comprises: in response to the authenticity of the image sequence being non-fake, obtaining a target face, and performing an operation corresponding to the target request based on the face contained in the image sequence and the target face; and in response to the authenticity of the image sequence being fake, generating prompt information, the prompt information being used to prompt that the operation based on the target request fails.

[0141] The image sequence identification method provided by the embodiments of the present application can be applied to payment scenarios, identity verification scenarios, and the like.

[0142] In a payment scenario, when a triggered payment request is obtained, a face video is obtained in response to the payment request, an image sequence containing a face is obtained based on the face video, and the authenticity of the image sequence is identified in a manner of steps 21-24.

[0143] When the authenticity of the image sequence is non-fake, a target face is obtained, the target face being a face read from a certain storage area, which can be a storage area configured by the electronic device itself, a server-side storage area, or an external storage area such as an identity card. Then, an operation corresponding to the payment request is performed based on the face contained in the image sequence and the target face. When the face contained in the image sequence is consistent with the target face, the payment operation is performed, and when the face contained in the image sequence is inconsistent with the target face, prompt information of a payment operation failure is generated.

[0144] When the authenticity of the image sequence is fake, prompt information of a payment operation failure is generated.

[0145] In an identity verification scenario, when a triggered identity verification request is obtained, a face video is obtained in response to the identity verification request, an image sequence containing a face is obtained based on the face video, and the authenticity of the image sequence is identified in a manner of steps 21-24.

[0146] When the authenticity of the image sequence is non-fake, a target face is obtained, the target face being a face read from a certain storage area, which can be a storage area configured by the electronic device itself, a server-side storage area, or an external storage area such as an identity card. Then, an operation corresponding to the identity verification request is performed based on the face contained in the image sequence and the target face. When the face contained in the image sequence is consistent with the target face, it indicates that the identity verification is successful, and prompt information of identity verification success is generated. When the face contained in the image sequence is inconsistent with the target face, it indicates that the identity verification fails, and prompt information of identity verification failure is generated.

[0147] When the authenticity of the image sequence is fake, a prompt information of identity verification failure is generated.

[0148] The technical scheme provided by the embodiment of the present application discriminates the authenticity of the image sequence according to the first spatial feature and the first time sequence feature of the image sequence, the image sequence includes at least two images, the first spatial feature is used to represent the fake area in each image, and the first time sequence feature is used to represent the difference area between adjacent images in the at least two images, thereby comprehensively discriminating the fake area in each image and the difference area between adjacent images in the image sequence, discriminating the authenticity of the image sequence, and improving the accuracy of the discrimination result.

[0149] The above describes the image sequence discrimination method of the embodiment of the present application from the perspective of method steps, which will be described in detail in combination with a payment scenario. In the payment scenario, when a triggered payment request is acquired, in response to the payment request, a face video is acquired, the face video is equally divided into 16 segments, the middle frame image in each segment is sampled as a target frame image containing a face, image recognition is performed on each target frame image to obtain the area where the face in each target frame image is located, the area is enlarged by 1.2 times to obtain a target area, the target area is cut so that the target area contains the entire face and part of the surrounding background, thereby obtaining an image sequence, which includes 16 images.

[0150] When discriminating the authenticity of the image sequence, a neural network model is used for discrimination. A deep residual network (Resnet) 50 is used as a skeleton network, a spatial-temporal model is used to replace the original block of the Resnet 50, thereby obtaining the neural network model used in the embodiment of the present application. The spatial-temporal model is also called spatial-temporal inconsistency learning (STIL), which is used to extract the first spatial feature and the first time sequence feature of the image sequence.

[0151] As shown in FIG. 1, the neural network model includes a spatial-temporal model, a first feature extraction network, and a second feature extraction network. Figure 8 As shown in FIG. 1, the neural network model includes a spatial-temporal model, a first feature extraction network, and a second feature extraction network. Figure 8 is a structural schematic diagram of a spatial-temporal model provided by the embodiment of the present application, the spatial-temporal model includes a spatial sub-model, a time sequence sub-model, and an information supplementing sub-model, wherein the spatial-temporal model is used to cut the features input into the spatial-temporal model to obtain the features input into the spatial sub-model and the features input into the time sequence sub-model.

[0152] The spatial sub-model is used to extract a first spatial feature of the image sequence. In one aspect, the feature of the input spatial sub-model is processed by convolution to obtain a convolution-processed feature. In another aspect, the feature of the input spatial sub-model is processed by convolution twice, and the features processed by the two times of convolution are added to the feature of the input spatial sub-model to obtain an added feature. Then, the added feature and the convolution-processed feature are dot-multiplied, and then processed by convolution to obtain the first spatial feature. For details, refer to the description of step 22 above, which will not be repeated here.

[0153] The temporal sub-model is used to extract a first sub-feature of the image sequence. In one aspect, the feature of the input temporal sub-model is processed by reshaping and forward difference to obtain a difference feature, which is processed by two kinds of convolution, and the features processed by the two kinds of convolution are added to the difference feature. After convolution and normalization, a first direction feature is obtained. In another aspect, the feature of the input temporal sub-model is processed by reshaping and forward difference to obtain a difference feature, which is processed by two kinds of convolution, and the features processed by the two kinds of convolution are added to the difference feature. After convolution and normalization, a second direction feature is obtained. The first direction feature and the second direction feature are added, and the added feature is multiplied by the feature of the input temporal sub-model, and then processed by convolution to obtain the first sub-feature. For details, refer to the description of “obtaining the first sub-feature of the image sequence based on the second channel feature” above, which will not be repeated here.

[0154] The information supplement sub-model is used to extract a second sub-feature of the image sequence. After the first spatial feature is input into the information supplement sub-model, the information supplement sub-model processes the first spatial feature by one-dimensional convolution and normalization, and then dot-multiplies the processed spatial feature with the first spatial feature and processes by convolution to obtain the second sub-feature. For details, refer to the description of “obtaining the second sub-feature of the image sequence based on the first spatial feature” above, which will not be repeated here.

[0155] In the embodiments of the present application, the space-time model is also used to add the first sub-feature and the second sub-feature, and then processes the added feature by convolution to obtain a first temporal feature of the image sequence. Then, the first spatial feature, the first temporal feature, and the feature of the input space-time model are added to fuse the first spatial feature, the first temporal feature, and the feature of the input space-time model to obtain a first fused feature, which is output by the space-time model.

[0156] In the embodiment of the present application, a convolutional layer is used to extract sequence features of the image sequence, and the sequence features are obtained as the features of the spatio-temporal model and input into the spatio-temporal model. The spatio-temporal model divides the sequence features into two parts, one part of the features is processed by the spatial sub-model to obtain the first spatial feature, and the other part of the features is processed by the temporal sub-model to obtain the first sub-feature, wherein the first spatial feature is processed by the information supplement sub-model to obtain the second sub-feature. The spatio-temporal model obtains the first fusion feature based on the first sub-feature, the second sub-feature and the first spatial feature and outputs the first fusion feature. Then, the first fusion feature is input into the spatio-temporal model as the feature of the spatio-temporal model, and the second fusion feature is obtained by the spatio-temporal model according to the above feature processing mode of the spatio-temporal model and output. Then, the second fusion feature is input into the spatio-temporal model as the feature of the spatio-temporal model, and the third fusion feature is obtained by the spatio-temporal model according to the above feature processing mode of the spatio-temporal model and output, and so on, until the target number such as 10 times is reached. The obtained fusion feature is input into the fully connected layer, and the authenticity probability of the image sequence is output by the fully connected layer, and the authenticity of the image sequence is obtained based on the authenticity probability of the image sequence.

[0157] In the embodiment of the present application, the fusion feature input into the fully connected layer is obtained after the feature processing by the plurality of spatio-temporal models, and the fusion feature represents the forged regions of each image in the image sequence and the difference regions between adjacent images, as shown in Figure 9 Figure 9 is a schematic diagram of an image region represented by a fusion feature provided by the embodiment of the present application. The image sequence is composed of images (1)-(4), and the image sequence is input into the neural network model, and the fusion feature is output by the last spatio-temporal model of the neural network model. The region represented by the reference numeral 91 in the image (1) is the forged region of the image (1) contained in the fusion feature, the region represented by the reference numeral 92 in the image (2) is the forged region of the image (2) and the difference region between the image (2) and the image (1) contained in the fusion feature, the region represented by the reference numeral 93 in the image (3) is the forged region of the image (3) and the difference region between the image (3) and the image (2) contained in the fusion feature, and the region represented by the reference numeral 94 in the image (4) is the forged region of the image (4) and the difference region between the image (4) and the image (3) contained in the fusion feature.

[0158] When the authenticity of the image sequence is a non-forged image sequence, the target face is acquired, and a payment operation is performed based on the target face and the face contained in the image sequence. When the authenticity of the image sequence is a forged image sequence, a prompt information of a failed payment operation is generated.

[0159] ​The related information about performing a payment operation based on the target face and the face included in the image sequence is described in detail in the above description of "in response to the authenticity of the image sequence being non-forged, obtaining the target face, and performing an operation corresponding to the target request based on the face included in the image sequence and the target face; and in response to the authenticity of the image sequence being forged, generating a prompt information, the prompt information being used to prompt that the operation based on the target request fails". Details are not described herein again.

[0160] Figure 10 Fig. 1 shows a structural schematic diagram of an image sequence identification device provided by an embodiment of the present application, and Fig. 2 shows a structural schematic diagram of an image sequence identification device provided by another embodiment of the present application. Figure 10 The device includes:

[0161] The obtaining module 1001 is configured to obtain an image sequence, the image sequence including at least two images.

[0162] The extraction module 1002 is configured to extract a first spatial feature of the image sequence, the first spatial feature being used to represent a forged region in each image of the at least two images.

[0163] The extraction module 1002 is further configured to extract a first time sequence feature of the image sequence, the first time sequence feature being used to represent a difference region between adjacent images of the at least two images.

[0164] The identification module 1003 is configured to identify authenticity of the image sequence according to the first spatial feature and the first time sequence feature, the authenticity being used to represent whether the image sequence is a forged image sequence.

[0165] In a possible implementation, the device further includes:

[0166] The obtaining module 1001 is further configured to obtain a sequence feature of the image sequence, the sequence feature being used to represent image information of each image of the at least two images.

[0167] The segmentation module is configured to segment the sequence feature based on the channel information to obtain a first channel feature and a second channel feature.

[0168] The extraction module 1002 is configured to obtain the first spatial feature of the image sequence based on the first channel feature.

[0169] The extraction module 1002 is configured to obtain the first time sequence feature of the image sequence based on the second channel feature.

[0170] In a possible implementation, the extraction module 1002 is configured to obtain the first time sequence feature of the image sequence based on the second channel feature and the first spatial feature.

[0171] In a possible implementation, the extraction module 1002 is configured to acquire a first sub-feature of the image sequence based on the second channel feature; acquire a second sub-feature of the image sequence based on the first spatial feature; and fuse the first sub-feature and the second sub-feature to obtain the first temporal feature of the image sequence.

[0172] In a possible implementation, the extraction module 1002 is configured to perform feature processing on the second channel feature in a first direction to obtain a first direction feature; perform feature processing on the second channel feature in a second direction to obtain a second direction feature; and fuse the first direction feature, the second direction feature, and the second channel feature to obtain the first sub-feature of the image sequence.

[0173] In a possible implementation, the discrimination module 1003 is configured to fuse the first spatial feature and the first temporal feature to obtain a first fused feature; determine a probability of authenticity of the image sequence based on the first fused feature; and determine authenticity of the image sequence based on the probability of authenticity of the image sequence.

[0174] In a possible implementation, the first spatial feature and the first temporal feature are obtained based on a sequence feature of the image sequence, and the discrimination module 1003 is configured to fuse the first spatial feature, the first temporal feature, and the sequence feature to obtain the first fused feature.

[0175] In a possible implementation, the discrimination module 1003 is configured to split the first fused feature based on channel information to obtain a third channel feature and a fourth channel feature; acquire a second spatial feature of the image sequence based on the third channel feature; acquire a second temporal feature of the image sequence based on the fourth channel feature; fuse the second spatial feature and the second temporal feature to obtain a second fused feature; and determine the probability of authenticity of the image sequence based on the second fused feature.

[0176] In a possible implementation, the first spatial feature and the first temporal feature are extracted based on a space-time model, and the apparatus further includes:

[0177] The acquisition module 1001 is configured to acquire a sample image sequence and a ground truth of authenticity of the sample image sequence, the sample image sequence including at least two sample images.

[0178] The extraction module 1002 is configured to extract a sample spatial feature of the sample image sequence, the sample spatial feature being used to represent a forged region in each sample image in the at least two sample images.

[0179] The extraction module 1002 is configured to extract a sample temporal feature of the sample image sequence, the sample temporal feature being used to represent a difference region between adjacent sample images in the at least two sample images.

[0180] determine a probability of authenticity of the sample image sequence based on the sample space feature and the sample time sequence feature;

[0181] The acquisition module 1001 is configured to acquire the spatio-temporal model based on the probability of authenticity of the sample image sequence and the authenticity annotation result.

[0182] In a possible implementation, the image sequence is acquired based on a video, and the device further includes:

[0183] The recommendation module is configured to, in response to the authenticity of the image sequence being non-fake, recommend a video corresponding to the image sequence.

[0184] The filtering module is configured to, in response to the authenticity of the image sequence being fake, filter out a video corresponding to the image sequence.

[0185] In a possible implementation, the image sequence is an image sequence containing a face, and the acquisition module 1001 is configured to, in response to a triggered target request, acquire the image sequence, the target request being a payment request or an identity verification request.

[0186] The device further includes:

[0187] The acquisition module 1001 is further configured to, in response to the authenticity of the image sequence being non-fake, acquire a target face, and perform an operation corresponding to the target request based on the face contained in the image sequence and the target face.

[0188] The generation module is configured to, in response to the authenticity of the image sequence being fake, generate a prompt information, the prompt information being used to prompt a failure of the operation performed based on the target request.

[0189] In a possible implementation, the acquisition module 1001 is configured to extract at least two target frame images from the video, the target frame image being an image containing a target object; and acquire the image sequence based on the at least two target frame images.

[0190] In a possible implementation, the acquisition module 1001 is configured to determine a target region in each of the at least two target frame images, the target region being a region containing the target object; and cut the target region in each of the at least two target frame images to obtain the image sequence.

[0191] In a possible implementation, the acquisition module 1001 is configured to perform image recognition on each of the at least two target frame images to obtain region information of the target object in each of the at least two target frame images; and expand the region information of the target object in each of the at least two target frame images according to an expansion ratio to obtain the target region in each of the at least two target frame images.

[0192] The device discriminates the authenticity of the image sequence according to the first spatial feature and the first time sequence feature. The image sequence includes at least two images. The first spatial feature is used to represent the forged area in each image. The first time sequence feature is used to represent the difference area between adjacent images in the at least two images. The authenticity of the image sequence is discriminated by comprehensively considering the forged area in each image and the difference area between adjacent images, and the accuracy of the discrimination result is improved.

[0193] It should be understood that the above Figure 10 The device provided in the application is only exemplified by the above division of the functional modules when realizing the functions. In actual application, the above functions can be completed by different functional modules according to the needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is described in the method embodiments, which will not be described here.

[0194] Figure 11 A structure block diagram of a terminal device 1100 provided by an example embodiment of the application is shown. The terminal device 1100 can be a portable mobile terminal, such as a smart phone, a tablet computer, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a notebook computer or a desktop computer. The terminal device 1100 can also be referred to as a user equipment, a portable terminal, a laptop terminal, a desktop terminal, and other names.

[0195] Generally, the terminal device 1100 includes a processor 1101 and a memory 1102.

[0196] ​The processor 1101 can include one or more processing cores, such as a 4-core processor, an 8-core processor, and the like. The processor 1101 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), a PLA (Programmable Logic Array). The processor 1101 can also include a main processor and a coprocessor, the main processor being a processor for processing data in an awake state, also known as a CPU (Central Processing Unit), and the coprocessor being a low-power processor for processing data in a standby state. In some embodiments, the processor 1101 can be integrated with a GPU (Graphics Processing Unit) for rendering and drawing content required to be displayed by the display screen. In some embodiments, the processor 1101 can also include an AI (Artificial Intelligence) processor for processing machine learning-related computing operations.

[0197] The memory 1102 can include one or more computer-readable storage media that can be non-transitory. The memory 1102 can also include a high-speed random access memory, and a nonvolatile memory such as one or more disk storage devices, flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 1102 is used to store at least one instruction for being executed by the processor 1101 to implement the image sequence identification method provided by the method embodiments in the present application.

[0198] In some embodiments, the terminal device 1100 can also optionally include a peripheral device interface 1103 and at least one peripheral device. The processor 1101, the memory 1102, and the peripheral device interface 1103 can be connected through a bus or a signal line. Each peripheral device can be connected to the peripheral device interface 1103 through a bus, a signal line, or a circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 1104, a display screen 1105, a camera assembly 1106, an audio circuit 1107, and a power supply 1109.

[0199] The peripheral interface 1103 can be used to connect at least one I / O (Input / Output) related peripheral device to the processor 1101 and the memory 1102. In some embodiments, the processor 1101, the memory 1102 and the peripheral interface 1103 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 1101, the memory 1102 and the peripheral interface 1103 can be implemented on a separate chip or circuit board, and the present embodiments are not limited in this regard.

[0200] The radio frequency circuit 1104 is used to receive and send RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 1104 communicates with a communication network and other communication devices through electromagnetic signals. The radio frequency circuit 1104 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 1104 includes an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and the like. The radio frequency circuit 1104 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to the World Wide Web, a metropolitan area network, an intranet, various generations of mobile communication networks (2G, 3G, 4G and 5G), a wireless local area network and / or a WiFi (Wireless Fidelity) network. In some embodiments, the radio frequency circuit 1104 can also include NFC (Near Field Communication) related circuitry, and the present application is not limited in this regard.

[0201] The display screen 1105 is configured to display a UI (User Interface). The UI can include graphics, text, icons, video, and any combination thereof. When the display screen 1105 is a touch display screen, the display screen 1105 is further configured to capture touch signals on or above the surface of the display screen 1105. The touch signals can be input to the processor 1101 as control signals for processing. In this case, the display screen 1105 can also be configured to provide virtual buttons and / or virtual keyboard, also known as soft buttons and / or soft keyboard. In some embodiments, the display screen 1105 can be one, disposed on the front panel of the terminal device 1100; in other embodiments, the display screen 1105 can be at least two, respectively disposed on different surfaces of the terminal device 1100 or in a folding design; in other embodiments, the display screen 1105 can be a flexible display screen, disposed on a curved surface or a folding surface of the terminal device 1100. Even, the display screen 1105 can also be disposed in an irregular shape, i.e. a special-shaped screen. The display screen 1105 can be made of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), etc.

[0202] The camera assembly 1106 is configured to capture images or videos. Optionally, the camera assembly 1106 includes a front camera and a rear camera. Typically, the front camera is disposed on the front panel of the terminal, and the rear camera is disposed on the back of the terminal. In some embodiments, the rear camera is at least two, which is any one of a main camera, a depth-of-field camera, a wide-angle camera, and a long-focus camera, to realize the background blur function of the main camera and the depth-of-field camera, the panoramic shooting and VR (Virtual Reality) shooting function of the main camera and the wide-angle camera, or other fusion shooting functions. In some embodiments, the camera assembly 1106 can further include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. The dual-color temperature flash refers to the combination of a warm light flash and a cold light flash, which can be used for light compensation under different color temperatures.

[0203] The audio circuit 1107 can include a microphone and a speaker. The microphone is used to collect sound waves of the user and the environment, and convert the sound waves into an electrical signal input to the processor 1101 for processing, or input to the radio frequency circuit 1104 to realize voice communication. For the purpose of stereo sound collection or noise reduction, the microphone can be multiple, respectively arranged at different parts of the terminal device 1100. The microphone can also be an array microphone or an omnidirectional collection type microphone. The speaker is used to convert the electrical signal from the processor 1101 or the radio frequency circuit 1104 into sound waves. The speaker can be a traditional diaphragm speaker, or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, not only can it convert electrical signals into sound waves that humans can hear, but it can also convert electrical signals into sound waves that humans cannot hear for ranging purposes. In some embodiments, the audio circuit 1107 can also include a headphone jack.

[0204] The power supply 1109 is used to supply power to each component in the terminal device 1100. The power supply 1109 can be alternating current, direct current, disposable battery or rechargeable battery. When the power supply 1109 includes a rechargeable battery, the rechargeable battery can be a wired charging battery or a wireless charging battery. The wired charging battery is a battery charged through a wired line, and the wireless charging battery is a battery charged through a wireless coil. The rechargeable battery can also be used to support fast charging technology.

[0205] In some embodiments, the terminal device 1100 further includes one or more sensors 1110. The one or more sensors 1110 include, but are not limited to, an acceleration sensor 1111, a gyroscope sensor 1112, a pressure sensor 1113, an optical sensor 1115, and a proximity sensor 1116.

[0206] The acceleration sensor 1111 can detect the acceleration magnitude in three coordinate axes of the coordinate system established by the terminal device 1100. For example, the acceleration sensor 1111 can be used to detect the components of the gravitational acceleration in three coordinate axes. The processor 1101 can control the display screen 1105 to display the user interface in a landscape view or a portrait view according to the gravitational acceleration signal collected by the acceleration sensor 1111. The acceleration sensor 1111 can also be used for game or user motion data collection.

[0207] The gyroscope sensor 1112 can detect the body orientation and rotation angle of the terminal device 1100, and the gyroscope sensor 1112 can collect 3D actions of the user on the terminal device 1100 in cooperation with the acceleration sensor 1111. The processor 1101 can realize the following functions according to the data collected by the gyroscope sensor 1112: motion sensing (such as changing the UI according to the user's tilt operation), image stabilization when shooting, game control, and inertial navigation.

[0208] The pressure sensor 1113 can be disposed at the side frame of the terminal device 1100 and / or the lower layer of the display screen 1105. When the pressure sensor 1113 is disposed at the side frame of the terminal device 1100, the holding signal of the user to the terminal device 1100 can be detected, and the left-hand or right-hand recognition or the shortcut operation can be performed by the processor 1101 according to the holding signal collected by the pressure sensor 1113. When the pressure sensor 1113 is disposed at the lower layer of the display screen 1105, the controllability control on the UI interface can be realized by the processor 1101 according to the pressure operation of the user to the display screen 1105. The controllability control includes at least one of the button control, the scroll bar control, the icon control and the menu control.

[0209] The optical sensor 1115 is used to collect the ambient light intensity. In an embodiment, the processor 1101 can control the display brightness of the display screen 1105 according to the ambient light intensity collected by the optical sensor 1115. Specifically, when the ambient light intensity is high, the display brightness of the display screen 1105 is increased; when the ambient light intensity is low, the display brightness of the display screen 1105 is decreased. In another embodiment, the processor 1101 can also dynamically adjust the shooting parameter of the camera assembly 1106 according to the ambient light intensity collected by the optical sensor 1115.

[0210] The proximity sensor 1116, also called the distance sensor, is usually disposed at the front panel of the terminal device 1100. The proximity sensor 1116 is used to collect the distance between the user and the front of the terminal device 1100. In an embodiment, when the proximity sensor 1116 detects that the distance between the user and the front of the terminal device 1100 gradually decreases, the display screen 1105 is switched from the bright screen state to the off-screen state by the processor 1101; when the proximity sensor 1116 detects that the distance between the user and the front of the terminal device 1100 gradually increases, the display screen 1105 is switched from the off-screen state to the bright screen state by the processor 1101.

[0211] Those skilled in the art can understand that the structures shown in the above embodiments do not constitute a limitation on the terminal device 1100, and the terminal device 1100 can include more or fewer components than those shown in the figures, or combine certain components, or adopt different component arrangements. Figure 11

[0212] Figure 12 ​A structural diagram of a server provided in the embodiments of the present application is shown in FIG. 12. The server 1200 can have great differences due to different configurations or performances, and can include one or more processors (Central Processing Units, CPUs) 1201 and one or more memories 1202. The one or more memories 1202 store at least one piece of program code, which is loaded and executed by the one or more processors 1201 to implement the image sequence identification method provided in any of the above method embodiments. Of course, the server 1200 can also have a wired or wireless network interface, a keyboard, an input and output interface, and other components for implementing device functions, and the like, so as to perform input and output. The server 1200 can also include other components for implementing device functions, which are not described herein.

[0213] In an example embodiment, a computer-readable storage medium is also provided, which stores at least one piece of program code. The at least one piece of program code is loaded and executed by a processor to enable an electronic device to implement any of the above image sequence identification methods.

[0214] Optionally, the above computer-readable storage medium can be a Read-Only Memory (ROM), a Random Access Memory (RAM), a Compact Disc Read-Only Memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0215] In an example embodiment, a computer program or computer program product is also provided, which stores at least one computer instruction. The at least one computer instruction is loaded and executed by a processor to enable a computer to implement any of the above image sequence identification methods.

[0216] It should be understood that "multiple" referred to herein refers to two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. The character " / " generally represents that the associated objects before and after it are in an "or" relationship.

[0217] The above sequence numbers of the embodiments of the present application are only for description, and do not represent the advantages or disadvantages of the embodiments.

[0218] The above merely describes example embodiments of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, and the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method of authentication of a sequence of images, characterized in that, The method includes: Obtain an image sequence, wherein the image sequence includes at least two images; Obtain the sequence features of the image sequence, the sequence features being used to characterize the image information of each of the at least two images; The sequence features are segmented based on channel information to obtain first channel features and second channel features; The first spatial feature of the image sequence is obtained based on the first channel feature, and the first spatial feature is used to characterize the fake region in each of the at least two images; The first temporal feature of the image sequence is obtained based on the second channel feature, and the first temporal feature is used to characterize the difference region between adjacent images in the at least two images; The authenticity of the image sequence is identified based on the first spatial feature and the first temporal feature, wherein the authenticity is used to characterize whether the image sequence is a forged image sequence.

2. The method of claim 1, wherein, The step of obtaining the first temporal feature of the image sequence based on the second channel feature includes: The first temporal feature of the image sequence is obtained based on the second channel feature and the first spatial feature.

3. The method of claim 2, wherein, The step of obtaining the first temporal feature of the image sequence based on the second channel feature and the first spatial feature includes: Based on the second channel features, the first sub-feature of the image sequence is obtained; Based on the first spatial feature, the second sub-feature of the image sequence is obtained; The first sub-feature and the second sub-feature are fused to obtain the first temporal feature of the image sequence.

4. The method of claim 3, wherein, The step of obtaining the first sub-feature of the image sequence based on the second channel feature includes: The second channel features are processed in the first direction to obtain the first direction features; The second channel features are processed in the second direction to obtain the second direction features; The first directional feature, the second directional feature, and the second channel feature are fused to obtain the first sub-feature of the image sequence.

5. The method of claim 1, wherein, The step of identifying the authenticity of the image sequence based on the first spatial feature and the first temporal feature includes: The first spatial feature and the first temporal feature are fused to obtain the first fused feature; The authenticity probability of the image sequence is determined based on the first fusion feature; The authenticity of the image sequence is determined based on the probability of its authenticity.

6. The method of claim 5, wherein, The first spatial feature and the first temporal feature are obtained based on the sequence features of the image sequence. The step of fusing the first spatial feature and the first temporal feature to obtain the first fused feature includes: The first spatial feature, the first temporal feature, and the sequence feature are fused to obtain the first fused feature.

7. The method of claim 5, wherein, Determining the authenticity probability of the image sequence based on the first fusion feature includes: Based on the channel information, the first fused feature is segmented to obtain the third channel feature and the fourth channel feature; The second spatial features of the image sequence are obtained based on the third channel features; The second temporal feature of the image sequence is obtained based on the fourth channel feature; The second spatial feature and the second time sequence feature are fused to obtain a second fusion feature; The authenticity probability of the image sequence is determined based on the second fusion feature.

8. The method according to any one of claims 1 to 7, characterized in that, The first spatial feature and the first time sequence feature are extracted based on a space-time model, and before the first spatial feature of the image sequence is obtained based on the first channel feature, the method further includes: obtaining a sample image sequence and authenticity annotation results of the sample image sequence, the sample image sequence including at least two sample images; extracting sample spatial features of the sample image sequence, the sample spatial features being used to represent forged areas in each sample image in the at least two sample images; extracting sample time sequence features of the sample image sequence, the sample time sequence features being used to represent difference areas between adjacent sample images in the at least two sample images; determining authenticity probabilities of the sample image sequence based on the sample spatial features and the sample time sequence features; obtaining the space-time model based on the authenticity probabilities of the sample image sequence and the authenticity annotation results.

9. The method according to any one of claims 1 to 7, characterized in that, The image sequence is obtained based on a video, and after the authenticity of the image sequence is identified according to the first spatial feature and the first time sequence feature, the method further includes: in response to the authenticity of the image sequence being non-forged, recommending a video corresponding to the image sequence; in response to the authenticity of the image sequence being forged, filtering out the video corresponding to the image sequence.

10. The method according to any one of claims 1 to 7, characterized in that, The image sequence is a face-containing image sequence, and the obtaining of the image sequence includes: in response to a triggered target request, obtaining the image sequence, the target request being a payment request or an identity verification request; after the authenticity of the image sequence is identified according to the first spatial feature and the first time sequence feature, the method further includes: in response to the authenticity of the image sequence being non-forged, obtaining a target face and performing an operation corresponding to the target request based on a face contained in the image sequence and the target face; in response to the authenticity of the image sequence being forged, generating a prompt information, the prompt information being used to prompt a failure of the operation performed based on the target request.

11. The method according to any one of claims 1 to 7, characterized in that, The obtaining of the image sequence includes: extracting at least two target frame images from a video, the target frame images being images containing a target object; determining target areas in each target frame image in the at least two target frame images, the target areas being areas containing the target object; cutting the target areas in the each target frame image to obtain the image sequence.

12. An apparatus for authenticating a sequence of images, characterized in that The device includes: an obtaining module configured to obtain an image sequence, the image sequence including at least two images; the obtaining module is further configured to obtain sequence features of the image sequence, the sequence features being used to represent image information of each image in the at least two images; a cutting module configured to cut the sequence features based on channel information to obtain first channel features and second channel features; The extraction module is configured to acquire first spatial features of the image sequence based on the first channel features, the first spatial features being used to represent the forged areas in each of the at least two images. The extraction module is further configured to acquire first temporal features of the image sequence based on the second channel features, the first temporal features being used to represent the difference areas between adjacent images in the at least two images. The discrimination module is configured to discriminate the authenticity of the image sequence according to the first spatial features and the first temporal features, the authenticity being used to represent whether the image sequence is a forged image sequence.

13. The apparatus of claim 12, wherein, The extraction module is configured to acquire first temporal features of the image sequence based on the second channel features and the first spatial features.

14. The apparatus of claim 13, wherein, The extraction module is configured to acquire first sub-features of the image sequence based on the second channel features, acquire second sub-features of the image sequence based on the first spatial features, and fuse the first sub-features and the second sub-features to obtain the first temporal features of the image sequence.

15. The apparatus of claim 14, wherein, The extraction module is configured to perform feature processing in a first direction on the second channel features to obtain first direction features, perform feature processing in a second direction on the second channel features to obtain second direction features, and fuse the first direction features, the second direction features, and the second channel features to obtain the first sub-features of the image sequence.

16. The apparatus of claim 12, wherein, The discrimination module is configured to fuse the first spatial features and the first temporal features to obtain first fused features, determine the authenticity probability of the image sequence based on the first fused features, and determine the authenticity of the image sequence based on the authenticity probability of the image sequence.

17. The apparatus of claim 16, wherein, The first spatial features and the first temporal features are obtained based on sequence features of the image sequence, and the discrimination module is configured to fuse the first spatial features, the first temporal features, and the sequence features to obtain first fused features.

18. The apparatus of claim 16, wherein, The discrimination module is configured to split the first fused features based on channel information to obtain third channel features and fourth channel features. The third channel features are used to acquire second spatial features of the image sequence. The fourth channel features are used to acquire second temporal features of the image sequence, and the second spatial features and the second temporal features are fused to obtain second fused features. The second fused features are used to determine the authenticity probability of the image sequence.

19. The apparatus of any of claims 12-18, wherein, The first spatial features and the first temporal features are extracted based on a space-time model, and the device further includes: The acquisition module is configured to acquire a sample image sequence and authenticity annotation results of the sample image sequence, the sample image sequence including at least two sample images. The extraction module is configured to extract sample spatial features of the sample image sequence, the sample spatial features being used to represent forged areas in each of the at least two sample images. The extraction module is configured to extract sample spatial features of the sample image sequence, the sample spatial features being used to represent forged areas in each of the at least two sample images. The extraction module is configured to extract sample temporal features of the sample image sequence, the sample temporal features being used to represent difference regions between adjacent sample images in the at least two sample images. The determination module is configured to determine a probability of authenticity of the sample image sequence based on the sample spatial features and the sample temporal features. The acquisition module is configured to acquire the spatio-temporal model based on the probability of authenticity of the sample image sequence and a result of authenticity labeling.

20. The apparatus of any of claims 12-18, wherein, The image sequence is acquired based on a video, and the device further includes: The recommendation module is configured to recommend a video corresponding to the image sequence in response to the authenticity of the image sequence being non-fake. The filtering module is configured to filter out the video corresponding to the image sequence in response to the authenticity of the image sequence being fake.

21. The apparatus of any of claims 12-18, wherein, The image sequence is a face-containing image sequence, and the acquisition module is configured to acquire the image sequence in response to a triggered target request, the target request being a payment request or an identity verification request. The device further includes: The acquisition module is further configured to acquire a target face in response to the authenticity of the image sequence being non-fake, and perform an operation corresponding to the target request based on a face contained in the image sequence and the target face. The generation module is configured to generate a prompt information in response to the authenticity of the image sequence being fake, the prompt information being used to prompt a failure of the operation based on the target request.

22. The apparatus of any of claims 12-18, wherein The acquisition module is configured to extract at least two target frame images from a video, the target frame images being images containing a target object. A target region in each of the at least two target frame images is determined, the target region being a region containing the target object; and a target region in each of the at least two target frame images is cut to obtain the image sequence.

23. An electronic device, comprising: The electronic device includes a processor and a memory, the memory storing at least one program code, the at least one program code being loaded and executed by the processor to enable the electronic device to implement the image sequence identification method according to any one of claims 1 to 11.

24. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one program code, the at least one program code being loaded and executed by the processor to enable the computer to implement the image sequence identification method according to any one of claims 1 to 11.

25. A computer program product, characterised in that, The computer program product stores at least one computer instruction, the at least one computer instruction being loaded and executed by the processor to enable the computer to implement the image sequence identification method according to any one of claims 1 to 11.

Images