Screen copying identification method, device and equipment

Abstract

The invention discloses a screen copying identification method, device and equipment, and the method comprises the steps: carrying out the LBP feature extraction of a target image through a plurality of preset pixel radiuses, and obtaining a plurality of LBP feature data sets; scaling the target image by using a scaling factor to obtain a target scaled image under the scaling factor; for each target scaled image, executing the steps S131-132: S131, performing edge detection and frequency domain analysis on the target scaled image to obtain the moire feature and moire intensity of the target scaled image; s132, inputting a plurality of LBP feature data sets and the moire features and moire intensity of the target scaled image into a preset multi-feature input convolutional neural network to obtain a copying probability and a non-copying probability of the target scaled image; and determining whether the target image is copied or not according to the copying probabilities and the non-copying probabilities of the plurality of target scaled images. The invention belongs to the field of image copying recognition. According to the invention, the accuracy of image copying identification can be improved.

Description

Technical Field

[0001] The present invention relates to the field of image copy recognition, and in particular to a screen copy recognition method, device and equipment. Background Art

[0002] Image re-take recognition is to detect whether an image is produced by taking another existing image (such as a screen, paper printout, or other display device), which is usually called "re-take" or "secondary shot". Re-taken images often have specific characteristics, such as moiré, geometric distortion, color distortion, blur, etc.

[0003] Existing methods for detecting duplicate images (screens) are mainly divided into two categories, including grayscale histogram similarity comparison and convolutional neural network (CNN) detection. Grayscale histogram similarity comparison relies on the existence of the original image. However, in practical applications, the detected photos are often only allowed to be taken on site. When detecting partial screen shots, the convolutional neural network (CNN) may not have obvious moiré patterns, resulting in detection failure. In addition, the pixel spatial frequency of the photosensitive element of a normal photo is close to the spatial frequency of the stripes in the image, resulting in the formation of colored and irregularly shaped stripes during the superposition process, which can also lead to misjudgment. Therefore, how to improve the accuracy of duplicate recognition is an urgent problem to be solved. Summary of the invention

[0004] The present invention solves the technical problem of inaccurate image copy recognition in the prior art by providing a screen copy recognition method, device and equipment, and achieves the technical effect of improving the accuracy of image copy recognition.

[0005] In a first aspect, the present invention provides a screen copy recognition method, the method comprising:

[0006] LBP feature extraction is performed on the target image at several preset pixel radii to obtain several LBP feature data sets, where each preset pixel radius corresponds to an LBP feature data set;

[0007] Set a number of scaling factors, and scale the target image with the scaling factors to obtain the target scaled image under the scaling factors;

[0008] For each target zoomed image, steps S131-132 are executed, including: step S131, edge detection and frequency domain analysis are performed on the target zoomed image to obtain moiré features and moiré intensity of the target zoomed image;

[0009] Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image;

[0010] Whether a target image is copied is determined according to the copy probabilities and non-copy probabilities of several target zoomed images.

[0011] Further, determining whether the target image is a copy according to the copy probability and non-copy probability of the plurality of target scaled images includes:

[0012] Determine a duplication determination result of each target zoomed image according to a duplication probability or a non-duplication probability of the target zoomed image, the duplication determination result being a duplication or a non-duplication;

[0013] According to the duplication determination results of the plurality of target zoomed images, it is determined whether the target image is duplicated.

[0014] Further, according to the probability of copying or non-copying of each target zoomed image, determining the copying determination result of the target zoomed image includes:

[0015] If the copy probability of the target zoomed image is greater than the copy probability threshold, the copy determination result of the target zoomed image is copy; or,

[0016] If the copy probability of the target zoomed image is less than or equal to the copy probability threshold, the copy determination result of the target zoomed image is not a copy; or,

[0017] If the non-copy probability of the target zoomed image is greater than the non-copy probability threshold, the copy determination result of the target zoomed image is non-copy; or,

[0018] If the non-copy probability of the target zoomed image is less than or equal to the non-copy probability threshold, the copy determination result of the target zoomed image is copy.

[0019] Further, determining whether the target image is a copy according to the copy determination results of the plurality of target zoomed images includes:

[0020] If the ratio of the number of duplications in the plurality of duplication determination results to the number of duplication determination results exceeds a preset ratio, the target image is a duplication.

[0021] Furthermore, after obtaining a plurality of LBP feature data sets, the method further includes:

[0022] Each LBP feature data set is processed by histogram equalization.

[0023] Further, including:

[0024] The preset pixel radius are 1, 2, 3, 4 and 5.

[0025] Further, including:

[0026] Scaling factors include 0.1, 0.2, 0.5, 0.7, and 1.0.

[0027] Further, determining the moiré intensity includes:

[0028] The moiré intensity of the target zoomed image is determined according to the sum of the moiré features of the target zoomed image.

[0029] In a second aspect, the present invention provides a screen copy recognition device, the device comprising:

[0030] The first feature extraction module is used to perform LBP feature extraction on the target image at a plurality of preset pixel radii to obtain a plurality of LBP feature data sets, wherein each preset pixel radius corresponds to an LBP feature data set;

[0031] A scaling factor setting module is used to set a number of scaling factors and scale the target image with the scaling factors to obtain a target scaled image under the scaling factors;

[0032] The probability output module is used to execute steps S131-132 for each target zoomed image, including: step S131, edge detection and frequency domain analysis are performed on the target zoomed image to obtain the moiré feature and moiré intensity of the target zoomed image;

[0033] Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image;

[0034] The copy determination module is used to determine whether a target image is copied according to the copy probability and non-copy probability of a plurality of target zoomed images.

[0035] In a third aspect, the present invention provides an electronic device, comprising:

[0036] processor;

[0037] a memory for storing processor-executable instructions;

[0038] Among them, the processor is configured to execute to implement a screen copy recognition method as provided in the first aspect.

[0039] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0040] The present invention provides a screen copying recognition method, which includes: performing LBP feature extraction on a target image with a plurality of preset pixel radii respectively to obtain a plurality of LBP feature data sets, wherein each preset pixel radius corresponds to an LBP feature data set; setting a plurality of scaling factors, and scaling the target image with the scaling factors to obtain the target scaled image under the scaling factors; executing steps S131-132 for each target scaled image, including: step S131, performing edge detection and frequency domain analysis on the target scaled image to obtain moiré features and moiré intensity of the target scaled image; step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target scaled image into a preset multi-feature input convolutional neural network to obtain a copying probability and a non-copying probability of the target scaled image; and determining whether the target image is a copy according to the copying probabilities and non-copying probabilities of the plurality of target scaled images.

[0041] The present invention performs multiple LBP feature extractions with a smaller pixel radius to capture subtle texture features in an image, such as small edges, local details, and rapidly changing tones, and is helpful for identifying fine local information, such as small objects or local edges.

[0042] The present invention can effectively enhance the local contrast of the image through CLAHE processing, so that the performance of the LBP feature is more consistent under different lighting conditions, which can significantly improve the adaptability and accuracy of the LBP feature in detection and avoid misjudgment caused by lighting changes.

[0043] The present invention proposes a method combining edge detection with frequency domain analysis to extract moiré features, which can solve the technical problem that the detection of moiré features relies on direct analysis in the spatial domain and is easily affected by the complexity of image content and background noise, resulting in low accuracy of moiré detection.

[0044] In order to ensure that the moiré features can be effectively captured under different shooting conditions (such as distance and angle changes), the present invention uses multi-factor scaling and performs moiré detection on each scaled image. This can capture the scale change problem of the moiré features caused by changes in shooting distance and angle, thereby improving the robustness of detection.

[0045] Under the conditions of multiple multi-factor scaling values, reasoning is performed using preset models respectively. After obtaining the reasoning results, a multi-factor voting mechanism is used to obtain the final result. Compared with the single feature detection in the existing technology, the recognition accuracy and robustness of the reproduced images are significantly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0047] Figure 1 A schematic diagram of a flow chart of a screen copy recognition method provided by the present invention;

[0048] Figure 2 A schematic diagram of the structure of a screen copy recognition device provided by the present invention;

[0049] Figure 3 The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0050] The embodiment of the present invention solves the technical problem of inaccurate image re-shooting recognition in the prior art by providing a screen re-shooting recognition method.

[0051] The technical solution of the present invention is to solve the above technical problems, and the overall idea is as follows:

[0052] A screen copying recognition method, the method comprising: performing LBP feature extraction on a target image with a plurality of preset pixel radii respectively to obtain a plurality of LBP feature data sets, each preset pixel radius corresponding to an LBP feature data set; setting a plurality of scaling factors, and scaling the target image with the scaling factors to obtain the target scaled image under the scaling factors; executing steps S131-132 for each target scaled image, comprising: step S131, performing edge detection and frequency domain analysis on the target scaled image to obtain moiré features and moiré intensity of the target scaled image; step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target scaled image into a preset multi-feature input convolutional neural network to obtain the copying probability and non-copying probability of the target scaled image; and determining whether the target image is a copy according to the copying probability and non-copying probability of the plurality of target scaled images.

[0053] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0054] First of all, the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0055] The present invention provides Figure 1 A screen copy recognition method shown includes steps S11-S14:

[0056] Regarding step S11, LBP feature extraction is performed on the target image at several preset pixel radii to obtain several LBP feature data sets, and each preset pixel radius corresponds to an LBP feature data set.

[0057] LBP (Local Binary Pattern) is a simple and effective operator used to describe the local texture features of an image. It generates a binary pattern by comparing the central pixel of an image with the pixels in its neighborhood to represent the texture information of the local area.

[0058] When performing LBP feature extraction on the target image with several preset pixel radii, it means that multiple LBP feature extractions can be performed using different neighborhood sizes (i.e., different pixel radii). Each preset pixel radius defines a circular neighborhood around the center pixel, and the pixels in the neighborhood will be used to calculate the LBP value.

[0059] For the target image, different LBP feature maps will be obtained by using different pixel radii. Each LBP feature map represents the texture information at a specific scale. If the image is extracted at n different pixel radii, n LBP feature data sets will be obtained, each of which corresponds to a specific pixel radius. The preset pixel radii are 1, 2, 3, 4 and 5. In other words, LBP feature extraction can be performed once with pixel radii of 1, 2, 3, 4 and 5, for a total of 5 extractions.

[0060] At a pixel radius of 1-5, subtle texture features in the image, such as small edges, local details, and rapidly changing tones, can be captured. This scale helps identify fine local information, such as small objects or local edges.

[0061] After obtaining a plurality of LBP feature data sets, the method further includes: performing histogram equalization processing on each LBP feature data set.

[0062] In order to further improve the stability of the extracted LBP features under different lighting conditions, the LBP features can be processed based on histogram equalization. Through CLAHE processing, the local contrast of the image can be effectively enhanced, making the performance of the LBP features more consistent under different lighting conditions.

[0063] Histogram equalization processing can significantly improve the adaptability and accuracy of LBP features in detection and avoid misjudgment caused by changes in illumination.

[0064] Regarding step S12, several scaling factors are set, and the target image is scaled by the scaling factors to obtain the target scaled image under the scaling factors. The scaling factors may include 0.1, 0.2, 0.5, 0.7 and 1.0. In addition, the scaling factors may also be set according to actual conditions.

[0065] The scaling factor is recorded as a scaling factor. For example, when the scaling factor is 0.1, the target image is scaled at a scaling factor of 0.1 to obtain a target scaled image at a scaling factor of 0.1.

[0066] Regarding step S13, for each target zoomed image, steps S131-132 are performed, including:

[0067] Step S131 , edge detection and frequency domain analysis are performed on the target zoomed image to obtain moiré features and moiré intensity of the target zoomed image.

[0068] Specifically, for each target zoomed image, the edge information in the image is first extracted using the Canny or Sobel edge detection algorithm.

[0069] The extracted edge information can help locate the important feature areas in each target zoomed image and reduce the interference of background noise on subsequent frequency domain analysis. Based on edge detection, a two-dimensional Fourier transform is performed on the target zoomed image to convert the image information of the target zoomed image from the spatial domain to the frequency domain. Moiré patterns usually appear as high-frequency components of the inherent frequency in the frequency domain. By analyzing the high-frequency components, the moiré features of each target zoomed image can be obtained.

[0070] Determining the moiré intensity includes: determining the moiré intensity of the target zoomed image according to the sum of the moiré features of the target zoomed image.

[0071] The following is a description of the above process, specifically:

[0072] Converting the target scaled image to the frequency domain can be accomplished by performing a two-dimensional discrete Fourier transform (2DDFT). The Fourier transform converts the image from the spatial domain to the frequency domain, and the information of different frequency components in the target scaled image can be determined;

[0073] In the frequency domain, the spectrum of the target zoomed image can be obtained. The spectrum shows the energy distribution of each frequency component in the target zoomed image. For an image containing moiré, obvious peaks or stripes will appear in the inherent frequency region, which are the frequency components corresponding to the moiré.

[0074] The main frequency components of the moiré pattern are retained while removing high-frequency noise and other irrelevant details. The main frequencies related to the moiré pattern are determined by applying a low-pass filter that allows signals below a pre-set cutoff frequency to pass through while blocking signals above this frequency.

[0075] After low-pass filtering, the remaining main frequency components are moiré patterns. The intensity of moiré patterns is determined by calculating the total energy of the moiré-related parts of the filtered spectrum. The higher the total energy, the more obvious the moiré patterns are or the greater the intensity.

[0076] Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image.

[0077] Taking the above-mentioned embodiment of the present invention as an example, for example, when the scaling factor is 0.1, the LBP feature data sets corresponding to pixel radii 1, 2, 3, 4 and 5, the moiré intensity of the target scaled image when the scaling factor is 0.1, and the moiré features are input into the preset multi-feature input convolutional neural network to obtain the probability of a copy and a non-copy of the target scaled image.

[0078] The above steps are performed on the target scaled image under each scale factor to obtain the copy probability and non-copy probability of the target scaled image under each scale factor.

[0079] Step S14, determining whether the target image is a copy according to the copy probabilities and non-copy probabilities of the plurality of target zoomed images.

[0080] Specifically, it includes: determining the duplication determination result of each target zoomed image according to the duplication probability or non-duplication probability of the target zoomed image, the duplication determination result being duplication or non-duplication; determining whether the target image is duplication according to the duplication determination results of several target zoomed images.

[0081] According to the probability of copying or non-copying of each target zoomed image, the copying determination result of the target zoomed image is determined, which specifically includes the following methods [1]-[4]: 【1】

[0083] If the copy probability of the target zoomed image is greater than the copy probability threshold, the copy determination result of the target zoomed image is copy; or, 【2】

[0085] If the copy probability of the target zoomed image is less than or equal to the copy probability threshold, the copy determination result of the target zoomed image is not a copy; or, 【3】

[0087] If the non-copy probability of the target zoomed image is greater than the non-copy probability threshold, the copy determination result of the target zoomed image is non-copy; or, 【4】

[0089] If the non-copy probability of the target zoomed image is less than or equal to the non-copy probability threshold, the copy determination result of the target zoomed image is copy.

[0090] It should be noted that any one of the methods [1]-[4] can be selected. It is understandable that the sum of the non-copy probability and the copy probability of the target zoomed image is 100%. In other words, when the non-copy probability and the copy probability of the target zoomed image are known, it is only necessary to set either the copy probability threshold or the non-copy probability threshold to determine the copy determination result of the target zoomed image. The non-copy probability threshold or the copy probability threshold can be set according to the actual situation. Specifically, the non-copy probability threshold or the copy probability threshold can be set to 50%.

[0091] Determining whether the target image is a copy according to the copy determination results of the plurality of target zoomed images includes: if the ratio of the number of copies in the plurality of copy determination results to the number of copy determination results exceeds a preset ratio, the target image is a copy.

[0092] For example, the number of copy determination results is 5, the number of copies in the copy determination results is 3, the preset ratio is 50%, 3 / 5>50%, then the target image is a copy, and the preset ratio can usually be determined according to actual conditions, such as 50%. In addition, the present invention also provides Table 1 for a clearer explanation.

[0093]

[0094]

[0095] In summary, the present invention provides a screen copying recognition method, which includes: performing LBP feature extraction on a target image with several preset pixel radii respectively to obtain several LBP feature data sets, each preset pixel radius corresponds to an LBP feature data set; setting several scaling factors, and scaling the target image with the scaling factors to obtain the target scaled image under the scaling factors; for each target scaled image, executing steps S131-132, including: step S131, performing edge detection and frequency domain analysis on the target scaled image to obtain the moiré feature and moiré intensity of the target scaled image; step S132, inputting several LBP feature data sets, the moiré feature and the moiré intensity of the target scaled image into a preset multi-feature input convolutional neural network to obtain the copying probability and non-copying probability of the target scaled image; determining whether the target image is copied according to the copying probability and non-copying probability of the several target scaled images.

[0096] The present invention performs multiple LBP feature extractions with a smaller pixel radius to capture subtle texture features in the image, such as small edges, local details, and rapidly changing tones, which helps to identify fine local information, such as small objects or local edges.

[0097] The present invention can effectively enhance the local contrast of the image through CLAHE processing, so that the performance of the LBP feature is more consistent under different lighting conditions, which can significantly improve the adaptability and accuracy of the LBP feature in detection and avoid misjudgment caused by lighting changes.

[0098] The present invention proposes a method of combining edge detection with frequency domain analysis to extract moiré features, which can solve the technical problem that the detection of moiré features often relies on direct analysis in the spatial domain and is easily affected by the complexity of image content and background noise, resulting in low accuracy of moiré detection.

[0099] In order to ensure that the moiré features can be effectively captured under different shooting conditions (such as distance and angle changes), the present invention uses multi-factor scaling and performs moiré detection on each scaled image. This can capture the scale change problem of the moiré features caused by changes in shooting distance and angle, thereby improving the robustness of detection.

[0100] Under the conditions of multiple multi-factor scaling values, the models are preset for reasoning respectively. After obtaining the reasoning results, the multi-factor voting mechanism is used to get the final result. Compared with the single feature detection in the existing technology, the recognition accuracy and robustness of the reproduced images are significantly improved.

[0101] Compared with the traditional grayscale histogram comparison method, the detection accuracy of the present invention is improved by about 20%; in normal pictures containing moiré patterns, the misjudgment rate is reduced by 20%, effectively reducing the misjudgment phenomenon of moiré pattern detection caused by environmental factors; through multi-factor scaling and voting mechanism, the detection effect of the detection method at different shooting distances and angles is more stable, and the misjudgment rate is reduced by 5%; after introducing data enhancement and network optimization measures, the performance of the model on different data sets is more consistent, and the accuracy on the verification set is compared with the training set, the decrease is controlled within 5%, and the response time is optimized by 50%; the present invention can complete the detection without relying on the original image, is suitable for a wider range of practical application scenarios, and improves the applicability of the system.

[0102] Based on the same inventive concept, the present invention provides Figure 2 A screen copy recognition device is shown, the device comprising:

[0103] The first feature extraction module 21 is used to perform LBP feature extraction on the target image at a plurality of preset pixel radii to obtain a plurality of LBP feature data sets, where each preset pixel radius corresponds to an LBP feature data set;

[0104] A scaling factor setting module 22 is used to set a number of scaling factors and scale the target image with the scaling factors to obtain a target scaled image under the scaling factors;

[0105] The probability output module 23 is used to execute steps S131-132 for each target zoomed image, including: step S131, edge detection and frequency domain analysis are performed on the target zoomed image to obtain the moiré feature and moiré intensity of the target zoomed image;

[0106] Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image;

[0107] The copy determination module 24 is used to determine whether a target image is copied according to the copy probability and non-copy probability of a plurality of target zoomed images.

[0108] Based on the same inventive concept, the present invention also provides Figure 3 An electronic device as shown includes:

[0109] Processor 31;

[0110] A memory 32 for storing instructions executable by the processor 31;

[0111] The processor 31 is configured to execute to implement a screen copy recognition method as provided above.

[0112] Since the electronic device introduced in this embodiment is an electronic device used to implement the information processing method in the embodiment of the present invention, based on the information processing method introduced in the embodiment of the present invention, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present invention is not described in detail here. As long as the electronic device used by a person skilled in the art to implement the information processing method in the embodiment of the present invention, it belongs to the scope of protection of the present invention.

[0113] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0114] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0115] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0116] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0117] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0118] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A screen copy recognition method, characterized in that: The method comprises: LBP feature extraction is performed on the target image at several preset pixel radii to obtain several LBP feature data sets, where each preset pixel radius corresponds to an LBP feature data set; Setting a number of scaling factors, and scaling the target image with the scaling factors to obtain a target scaled image under the scaling factors; For each target zoomed image, steps S131-132 are executed, including: step S131, edge detection and frequency domain analysis are performed on the target zoomed image to obtain moiré features and moiré intensity of the target zoomed image; Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image; According to the copy-probabilities and non-copy-probabilities of a plurality of target zoomed images, it is determined whether the target image is copied.

2. A screen copy recognition method as claimed in claim 1, characterized in that: The step of determining whether the target image is a copy according to the copy probability and non-copy probability of the plurality of target scaled images comprises: Determine a copy determination result of each target zoomed image according to the copy probability or non-copy probability of the target zoomed image, wherein the copy determination result is copy or non-copy; According to the copy-shooting determination results of the plurality of target zoomed images, it is determined whether the target image is copied.

3. A screen copy recognition method as claimed in claim 2, characterized in that: The step of determining the copying determination result of each target zoomed image according to the copying probability or non-copying probability of the target zoomed image comprises: If the copy probability of the target zoomed image is greater than the copy probability threshold, the copy determination result of the target zoomed image is copy; or, If the copy probability of the target zoomed image is less than or equal to the copy probability threshold, the copy determination result of the target zoomed image is not a copy; or, If the non-copy probability of the target zoomed image is greater than the non-copy probability threshold, the copy determination result of the target zoomed image is non-copy; or, If the non-copy probability of the target zoomed image is less than or equal to the non-copy probability threshold, the copy determination result of the target zoomed image is copy.

4. A screen copy recognition method as claimed in claim 3, characterized in that: The step of determining whether the target image is a copy according to the copy determination results of the plurality of target zoomed images comprises: If the ratio of the number of copies among the plurality of copy determination results to the number of copy determination results exceeds a preset ratio, the target image is a copy.

5. A screen copy recognition method as claimed in claim 1, characterized in that: After obtaining a number of LBP feature data sets, the method further includes: Each LBP feature data set is processed by histogram equalization.

6. A screen copy recognition method as claimed in claim 1, characterized in that: include: The preset pixel radius are 1, 2, 3, 4 and 5.

7. A screen copy recognition method as claimed in claim 1, characterized in that: include: Scaling factors include 0.1, 0.2, 0.5, 0.7, and 1.

0.

8. A screen copy recognition method as claimed in claim 1, characterized in that: Determining the intensity of moiré includes: The moiré intensity of the target zoomed image is determined according to the sum of the moiré features of the target zoomed image.

9. A screen copy recognition device, characterized in that: The device comprises: The first feature extraction module is used to perform LBP feature extraction on the target image at a plurality of preset pixel radii to obtain a plurality of LBP feature data sets, wherein each preset pixel radius corresponds to an LBP feature data set; A scaling factor setting module, used to set a number of scaling factors, and scale the target image with the scaling factors to obtain a target scaled image under the scaling factors; The probability output module is used to execute steps S131-132 for each target zoomed image, including: Step S131, performing edge detection and frequency domain analysis on the target zoomed image to obtain moiré features and moiré intensity of the target zoomed image; Step S132, inputting a plurality of LBP feature data sets, the moiré features and the moiré intensity of the target zoomed image into a preset multi-feature input convolutional neural network to obtain a copy probability and a non-copy probability of the target zoomed image; The copy determination module is used to determine whether the target image is copied according to the copy probability and non-copy probability of a plurality of target zoomed images.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute to implement a screen copy recognition method as described in any one of claims 1 to 8.

Images