Video image intelligent synthesis method based on artificial intelligence

Abstract

The invention relates to the field of video image synthesis, and particularly discloses a video image intelligent synthesis method based on artificial intelligence, which comprises the following steps: acquiring video image data to be synthesized and preprocessing; according to the data size and the diversification degree of the to-be-synthesized video image data, the target and the demand of the video image synthesis task, and the attribute of each artificial intelligence model of the video image synthesis, the adaptive artificial intelligence model of the video image synthesis is screened, and the synthesis effect of the video image is improved; basic parameters of video image synthesis are further set, indexes such as a peak signal-to-noise ratio, a structural similarity index, a feature similarity index and information entropy of video image synthesis are detected, the effect of video image synthesis is evaluated, whether video image synthesis needs to be optimized or not is judged, corresponding processing is carried out, and then intelligent synthesis of video images is achieved. The efficiency of video image synthesis is improved, and the quality of video image synthesis is guaranteed.

Description

Technical Field

[0001] The invention relates to the field of video image synthesis, and in particular to an intelligent video image synthesis method based on artificial intelligence. Background Art

[0002] With the advancement of artificial intelligence technology, the maturity of computer vision technology and the improvement of hardware performance, strong technical support has been provided for intelligent synthesis of video images. Video image intelligent synthesis technology has developed rapidly and has been widely used in the film and television entertainment industry, advertising and marketing industry, social media, education and training industry, etc. It not only greatly improves the creative efficiency and quality, but also reduces production costs, which is conducive to promoting the development of related industries.

[0003] However, the existing video image intelligent synthesis technology still has some limitations and shortcomings in practical applications.

[0004] For example, the existing Chinese patent with publication number CN114071155A discloses an artificial intelligence processing method and system, including a video uploading module, a video decoding module, an image extraction module, a character image extraction module, a definition enhancement module, an image beautification module, an image synthesis module, a background material selection module, and a video generation module, wherein the video uploading module is connected to the video decoding module, the video decoding module is connected to the image extraction module, the image extraction module is connected to the character image extraction module, the character image extraction module is connected to the definition enhancement module, the definition enhancement module is connected to the image beautification module, the image beautification module is connected to the image synthesis module, and the image synthesis module is respectively connected to the background material selection module and the video generation module. The invention can facilitate the processing of character images and backgrounds, and can adjust and beautify the definition of images, which saves time and effort compared to manual image processing.

[0005] However, the above patent has the following problems: 1. Different types of artificial intelligence models have different advantages and limitations in synthesizing video images. Selecting a suitable artificial intelligence model is crucial to the synthesis effect of video images. When the above patent uses artificial intelligence to synthesize video images, the artificial intelligence model for video image synthesis is not screened, which may make the artificial intelligence model for video image synthesis not highly adaptable, thereby affecting the effect of video image synthesis.

[0006] 2. The above patent does not detect and evaluate the effect of video image synthesis and further obtain an optimization solution for video image synthesis after the video image synthesis, thereby failing to better guarantee the quality of video image synthesis. Summary of the invention

[0007] In response to the above problems, the present invention proposes an intelligent video image synthesis method based on artificial intelligence. The specific technical solution is as follows: An intelligent video image synthesis method based on artificial intelligence comprises the following steps: Step 1, video image data acquisition and preprocessing: Collect the video image data to be synthesized and perform preprocessing.

[0008] Step 2. Selection of artificial intelligence models for video image synthesis: Obtain the data volume and diversity of the video image data to be synthesized, as well as the objectives and requirement information of the video image synthesis task, where the objectives and requirement information include video style, content type, synthesis complexity and synthesis accuracy, and obtain the attribute information of each artificial intelligence model for video image synthesis, where the attribute information includes the computing resource requirement level and inference speed, analyze the matching index of each artificial intelligence model for video image synthesis, and screen the adapted artificial intelligence model for video image synthesis.

[0009] Step 3. Analysis and setting of video image synthesis parameters: According to the objectives and requirement information of the video image synthesis task and the adapted artificial intelligence model of video image synthesis, analyze the appropriate basic parameters for video image synthesis, including resolution, frame rate, color space, contrast, brightness, saturation, and transparency, and further set the basic parameters for video image synthesis.

[0010] Step 4: Video image synthesis effect detection and evaluation: Detect the effect indicators of video image synthesis, where the effect indicators include peak signal-to-noise ratio, structural similarity index, feature similarity index, information entropy, and analyze the effect evaluation index of video image synthesis.

[0011] Step 5: Determine and process the optimization needs of video image synthesis: Determine whether the video image synthesis needs to be optimized based on the effect evaluation index of the video image synthesis, and perform corresponding processing.

[0012] Compared with the prior art, the artificial intelligence-based intelligent video image synthesis method described in the present invention has the following beneficial effects: 1. The present invention obtains the data amount and diversity of the video images to be synthesized, the goals and demand information of the video image synthesis task, and the attribute information of each artificial intelligence model for video image synthesis, and then screens the adaptive artificial intelligence model for video image synthesis, which can improve the matching degree of the artificial intelligence model for video image synthesis, thereby improving the synthesis effect of the video image.

[0013] 2. The present invention evaluates the effect of video image synthesis by detecting indicators such as the peak signal-to-noise ratio, structural similarity index, feature similarity index, information entropy, etc. of video image synthesis, determines whether the video image synthesis needs to be optimized, and performs corresponding processing, thereby better ensuring the quality of video image synthesis. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0015] Figure 1 It is a schematic diagram of the method flow of the present invention.

[0016] Figure 2 This is a structural block diagram of the artificial intelligence model for selecting video image synthesis of the present invention.

[0017] Figure 3 It is a structural block diagram of the present invention for evaluating the video image synthesis effect. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1 As shown, the present invention provides an artificial intelligence-based video image intelligent synthesis method, which includes the following steps: Step 1, video image data acquisition and preprocessing: acquiring the video image data to be synthesized and performing preprocessing.

[0020] In a specific embodiment, the video image data to be synthesized includes images of different scenes, different themes, and different lighting conditions.

[0021] It should be noted that the preprocessing of the synthesized video image data includes operations such as image cropping, scaling, and normalization to ensure data consistency and availability. At the same time, the synthesized video image data can also be processed such as denoising and enhancement to improve image quality.

[0022] Step 2. Selection of artificial intelligence models for video image synthesis: Obtain the data volume and diversity of the video image data to be synthesized, as well as the objectives and requirement information of the video image synthesis task, where the objectives and requirement information include video style, content type, synthesis complexity and synthesis accuracy, and obtain the attribute information of each artificial intelligence model for video image synthesis, where the attribute information includes the computing resource requirement level and inference speed, analyze the matching index of each artificial intelligence model for video image synthesis, and screen the adapted artificial intelligence model for video image synthesis.

[0023] As a preferred option, see Figure 2 As shown, the specific analysis process of step 2 includes: obtaining the accumulated data volume of the video image data to be synthesized.

[0024] It should be noted that the cumulative data volume of the video image data to be synthesized can be measured by indicators such as bit rate, file size and number of pixels.

[0025] The number of scene changes, the number of lighting condition changes, and the number of shooting angle changes of the video image data to be synthesized are obtained, and they are substituted into the relationship function between the preset number of scene changes, the number of lighting condition changes, the number of shooting angle changes, and the degree of diversity of the video image data to obtain the degree of diversity of the video image data to be synthesized.

[0026] It should be noted that the more times the scene changes, the more times the lighting conditions change, and the more times the shooting angle changes, the higher the degree of diversity of the video image data.

[0027] Extract the data volume range and diversity range of video image data acceptable to each artificial intelligence model for video image synthesis stored in the database.

[0028] It should be noted that the data volume range and diversity range of video image data acceptable to each artificial intelligence model for video image synthesis stored in the database are obtained based on the summary and analysis of information on historically synthesized video images by each artificial intelligence model for video image synthesis.

[0029] It should be noted that when selecting an artificial intelligence model for video image synthesis, it is necessary to consider the artificial intelligence model's ability to adapt to the diversity of video image data. Some models with strong generalization capabilities, such as models based on the Transformer architecture, can better handle diverse video image data.

[0030] If the cumulative data volume of the video image data to be synthesized falls within the data volume range of the video image data acceptable to a certain artificial intelligence model, then the data volume of the video image data acceptable to the artificial intelligence model meets the requirements; otherwise, the data volume of the video image data acceptable to the artificial intelligence model does not meet the requirements. The compliance coefficients of the video image data volumes acceptable to the artificial intelligence model corresponding to the data volumes of the video image data acceptable to the artificial intelligence model that meet the requirements and do not meet the requirements are set, and the compliance coefficients of the video image data volumes acceptable to each artificial intelligence model for video image synthesis are obtained by screening, and recorded as δ i , i represents the number of the i-th artificial intelligence model of video image synthesis, i = 1, 2, ..., n.

[0031] In a specific embodiment, the compliance coefficient of the amount of video image data acceptable to the artificial intelligence model that meets the requirements is 1, and the compliance coefficient of the amount of video image data acceptable to the artificial intelligence model that does not meet the requirements is 0.

[0032] If the diversity degree of the video image data to be synthesized falls within the range of the diversity degree of video image data acceptable to a certain artificial intelligence model for video image synthesis, then the diversity degree of video image data acceptable to the artificial intelligence model meets the requirements; otherwise, the diversity degree of video image data acceptable to the artificial intelligence model does not meet the requirements. The compliance coefficient of the diversity degree of video images acceptable to the artificial intelligence model corresponding to the diversity degree of video image data acceptable to the artificial intelligence model that meets the requirements and does not meet the requirements is set, and the compliance coefficient of the diversity degree of video images acceptable to each artificial intelligence model for video image synthesis is screened and recorded as ε i .

[0033] In a specific embodiment, if the degree of diversity of video image data acceptable to the artificial intelligence model meets the requirements, the corresponding compliance coefficient of the degree of diversity of video images acceptable to the artificial intelligence model is 1; if the degree of diversity of video image data acceptable to the artificial intelligence model does not meet the requirements, the corresponding compliance coefficient of the degree of diversity of video images acceptable to the artificial intelligence model is 0.

[0034] By analyzing the formula Get the first matching index of each artificial intelligence model for video image synthesis Wherein γ1 represents a correction factor of a preset first matching index, β1 and β2 represent weights of a preset acceptable video image data volume conformity coefficient and an acceptable video image diversity conformity coefficient, respectively, and β1+β2=1.

[0035] It should be noted that the setting of the weights of the acceptable video image data volume compliance coefficient and the acceptable video image diversity degree compliance coefficient is based on the importance of the acceptable video image data volume and the acceptable video image diversity degree compliance coefficient when selecting the artificial intelligence model for video image synthesis. In a specific embodiment, the weights of the acceptable video image data volume compliance coefficient and the acceptable video image diversity degree compliance coefficient are 0.6 and 0.4, respectively.

[0036] As a preferred solution, the specific analysis process of step 2 further includes: obtaining the video style, content type, synthesis complexity and synthesis accuracy of the video image synthesis task.

[0037] It should be noted that the video style, content type, synthesis complexity and synthesis accuracy of the video image synthesis task are obtained through user input information.

[0038] In another specific embodiment, the video image to be synthesized is identified and compared with feature elements corresponding to each video style and feature elements corresponding to each content type stored in a database to obtain the video style and content type of the video image synthesis task.

[0039] Extract the matching factors of each artificial intelligence model for video image synthesis and various video styles and the matching factors of each artificial intelligence model and each content type stored in the database, screen and obtain the matching factors of each artificial intelligence model for video image synthesis and the video style and content type of the video image synthesis task, and record them as

[0040] In a specific embodiment, if you want to synthesize a video with a specific artistic style, such as oil painting style, watercolor style, or cartoon style, you can consider using a style transfer model. For example, a style transfer algorithm based on deep learning can apply an artistic style to the original video image to achieve a unique visual effect.

[0041] In a specific embodiment, if the synthesis task has very high requirements on details and precision, such as the synthesis of medical images, the processing of satellite images, etc., an artificial intelligence model with high-resolution processing capabilities and precise modeling capabilities can be selected, such as a deep model based on a convolutional neural network, which can effectively extract image features and accurately model the details of the image.

[0042] The acceptable complexity threshold and achievable accuracy threshold of each artificial intelligence model for video image synthesis stored in the extraction database are extracted.

[0043] If the synthesis complexity of the video image synthesis task is less than or equal to the acceptable complexity threshold of a certain artificial intelligence model for video image synthesis, then the acceptable complexity of the artificial intelligence model meets the requirements; otherwise, the acceptable complexity of the artificial intelligence model does not meet the requirements. The acceptable complexity compliance coefficients corresponding to the acceptable complexity of the artificial intelligence model that meets the requirements and does not meet the requirements are set, and the acceptable complexity compliance coefficients of each artificial intelligence model for video image synthesis are screened and recorded as

[0044] In a specific embodiment, the acceptable complexity compliance coefficients of the artificial intelligence model corresponding to the acceptable complexity meeting the requirements and not meeting the requirements are 1 and 0 respectively.

[0045] If the synthesis accuracy of the video image synthesis task is less than or equal to the accuracy threshold of a certain artificial intelligence model for video image synthesis, then the accuracy of the artificial intelligence model meets the requirements; otherwise, the accuracy of the artificial intelligence model does not meet the requirements. The accuracy compliance coefficients corresponding to the accuracy of the artificial intelligence model that meets the requirements and does not meet the requirements are set, and the accuracy compliance coefficients of each artificial intelligence model for video image synthesis are screened and recorded as

[0046] In a specific embodiment, the accuracy compliance coefficients of the artificial intelligence model that meet the requirements and do not meet the requirements are 1 and 0 respectively.

[0047] By analyzing the formula Get the second matching index of each artificial intelligence model for video image synthesis Wherein γ2 represents the correction factor of the preset second matching index, e represents the natural constant, η1, η2, η3, η4 represent the weights of the preset video style matching factor, content type matching factor, acceptable complexity compliance coefficient and precision compliance coefficient respectively, and η1+η2+η3+η4=1.

[0048] It should be noted that the weights of the video style matching factor, content type matching factor, acceptable complexity compliance coefficient and precision compliance coefficient are set based on the matching degree of the video style and content type between the artificial intelligence model and the video image synthesis task when selecting the artificial intelligence model for video image synthesis, as well as the importance of the acceptable complexity and precision of the artificial intelligence model. In a specific embodiment, the weights of the video style matching factor, content type matching factor, acceptable complexity compliance coefficient and precision compliance coefficient are 0.2, 0.2, 0.3 and 0.3 respectively.

[0049] As a preferred solution, the specific analysis process of step two also includes: extracting the attribute information of each artificial intelligence model for video image synthesis stored in the database, and obtaining the computing resource requirement level and inference speed of each artificial intelligence model for video image synthesis.

[0050] The computing resource level that the video image synthesis device can provide is obtained. If the computing resource requirement level of a certain artificial intelligence model of video image synthesis is equal to or lower than the computing resource level that the video image synthesis device can provide, then the computing resource requirement of the artificial intelligence model can be met. Otherwise, the computing resource requirement of the artificial intelligence model cannot be met. The computing resource requirement compliance coefficient of the artificial intelligence model corresponding to the computing resource requirement that can be met and cannot be met is set, and the computing resource requirement compliance coefficient of each artificial intelligence model of video image synthesis is screened and recorded as κ i .

[0051] In a specific embodiment, the computing resource requirements of the artificial intelligence model can be satisfied and the computing resource requirements of the corresponding artificial intelligence model are satisfied, and the compliance coefficients are 1 and 0 respectively.

[0052] It should be noted that different video image synthesis artificial intelligence models have different requirements for computing resources during training and inference. Some complex video image synthesis artificial intelligence models, such as deep neural network models, require a lot of computing power and memory resources.

[0053] The inference speed of each artificial intelligence model for video image synthesis is denoted as v i .

[0054] It should be noted that for some application scenarios that require real-time synthesis, the inference time of the artificial intelligence model for video image synthesis is very important. If the inference speed of the artificial intelligence model for video image synthesis is too slow, it will cause delays in the synthesized video and affect the user experience.

[0055] By analyzing the formula Get the third matching index of each artificial intelligence model for video image synthesis Wherein γ3 represents the correction factor of the preset third matching index, λ1 and λ2 represent the weights of the preset computing resource requirement compliance coefficient and inference speed respectively, λ1+λ2=1, and n represents the number of artificial intelligence models for video image synthesis.

[0056] It should be noted that the setting of the computing resource requirement compliance coefficient and the weights of the inference speed is based on whether the computing resource requirements of the artificial intelligence model can be met when selecting the artificial intelligence model for video image synthesis and the importance of the inference speed of the artificial intelligence model. In a specific embodiment, the weights of the computing resource requirement compliance coefficient and the inference speed are 0.7 and 0.3 respectively.

[0057] As a preferred solution, the specific analysis process of step 2 further includes: the first matching index of each artificial intelligence model synthesized by the video image Second matching index The third matching index Substitute into the analysis formula Get the matching index of each artificial intelligence model for video image synthesis Wherein, μ1, μ2, and μ3 represent the weights of the preset first matching index, second matching index, and third matching index respectively, and μ1+μ2+μ3=1.

[0058] In a specific embodiment, the weights of the first matching index, the second matching index, and the third matching index are 0.3, 0.35, and 0.35, respectively.

[0059] As a preferred solution, the specific analysis process of step 2 also includes: comparing the matching indexes of the various artificial intelligence models for video image synthesis, and taking the artificial intelligence model corresponding to the maximum matching index as the adapted artificial intelligence model for video image synthesis.

[0060] It should be noted that different types of artificial intelligence models have different advantages and limitations in synthesizing video images. For example, generative adversarial networks perform well in generating realistic images, but there may be problems with unstable training; variational autoencoders are relatively stable, but the synthesized images may not be clear enough. Therefore, choosing a suitable artificial intelligence model for video image synthesis is crucial to the synthesis effect of video images.

[0061] It should be noted that the various artificial intelligence models for video image synthesis include but are not limited to: generative adversarial networks, diffusion models, variational autoencoders, Transformer models, style transfer models based on deep learning, etc.

[0062] In this embodiment, the present invention obtains the data volume and diversity of the video images to be synthesized, the objectives and requirement information of the video image synthesis task, and the attribute information of each artificial intelligence model for video image synthesis, and then screens the adapted artificial intelligence model for video image synthesis, thereby improving the matching degree of the artificial intelligence model for video image synthesis and thus improving the synthesis effect of the video image.

[0063] Step 3. Analysis and setting of video image synthesis parameters: According to the objectives and requirement information of the video image synthesis task and the adapted artificial intelligence model of video image synthesis, analyze the appropriate basic parameters for video image synthesis, including resolution, frame rate, color space, contrast, brightness, saturation, and transparency, and further set the basic parameters for video image synthesis.

[0064] As a preferred solution, the specific analysis process of step three is: extracting the case information of historical video image synthesis stored in the database, obtaining the goals and demand information of each historical video image synthesis case, the selected artificial intelligence model and the basic parameters of video image synthesis, using a machine learning algorithm to train a basic parameter analysis model for video image synthesis, substituting the goals and demand information of the video image synthesis task and the adapted artificial intelligence model of video image synthesis into the basic parameter analysis model for video image synthesis to obtain appropriate basic parameters for video image synthesis, wherein the basic parameters include resolution, frame rate, color space, contrast, brightness, saturation, and transparency, and further setting the basic parameters for video image synthesis.

[0065] Step 4: Video image synthesis effect detection and evaluation: Detect the effect indicators of video image synthesis, where the effect indicators include peak signal-to-noise ratio, structural similarity index, feature similarity index, information entropy, and analyze the effect evaluation index of video image synthesis.

[0066] As a preferred option, see Figure 3 As shown, the specific analysis process of step 4 includes: detecting the synthesized video image, obtaining the peak signal-to-noise ratio, structural similarity index, feature similarity index, and information entropy of the synthesized video image, which are respectively denoted as σ P , σ S , σ F , σ I .

[0067] It should be noted that the methods for obtaining the peak signal-to-noise ratio, the structural similarity index, the feature similarity index, and the information entropy are relatively mature existing technologies and will not be elaborated here.

[0068] It should be noted that the peak signal-to-noise ratio is an image quality evaluation indicator based on the error between corresponding pixels, that is, based on error sensitivity. Generally speaking, the higher the peak signal-to-noise ratio, the smaller the difference between the synthesized image and the original image, and the better the synthesis effect.

[0069] It should be noted that the structural similarity index is composed of three parts: brightness contrast, contrast contrast, and structural contrast. A value between 0 and 1 is obtained by comparing the similarities between the synthesized image and the original image in these three aspects. The closer the value of the structural similarity index is to 1, the higher the structural similarity between the synthesized image and the original image, and the better the synthesis effect.

[0070] It should be noted that the feature similarity index measures the similarity between the synthesized image and the original image based on features such as image phase consistency and gradient amplitude. The higher the value of the feature similarity index, the better the synthesis effect.

[0071] It should be noted that information entropy is used to measure the richness of information in an image. The larger the information entropy, the richer the information contained in the image and the better the synthesis effect.

[0072] As a preferred solution, the specific analysis process of step 4 further includes: calculating the peak signal-to-noise ratio σ of the synthesized video image P , structural similarity index σ S , feature similarity index σ F 、Information entropy σ I Substitute into the analysis formula The effect evaluation index ξ of video image synthesis is obtained, where ψ1, ψ2, ψ3, and ψ4 represent the preset peak signal-to-noise ratio, structural similarity index, feature similarity index, and information entropy weights, respectively, and ψ1+ψ2+ψ3+ψ4=1.

[0073] In a specific embodiment, the weights of the peak signal-to-noise ratio, the structural similarity index, the feature similarity index, and the information entropy are 0.25, 0.25, 0.25, and 0.25, respectively.

[0074] Step 5: Determine and process the optimization needs of video image synthesis: Determine whether the video image synthesis needs to be optimized based on the effect evaluation index of the video image synthesis, and perform corresponding processing.

[0075] As a preferred solution, the specific analysis process of step five is: compare the effect evaluation index of video image synthesis with the preset effect evaluation index threshold; if the effect evaluation index of video image synthesis is less than the preset effect evaluation index threshold, the video image synthesis needs to be optimized, and the optimization plan for video image synthesis is obtained and feedback is provided.

[0076] It should be noted that the optimization solution for video image synthesis can be obtained by adjusting the parameters of the artificial intelligence model of video image synthesis, improving the synthesis strategy, increasing the amount of data, etc.

[0077] In this embodiment, the present invention evaluates the effect of video image synthesis by detecting indicators such as the peak signal-to-noise ratio, structural similarity index, feature similarity index, information entropy, etc. of the video image synthesis, determines whether the video image synthesis needs to be optimized, and performs corresponding processing, thereby better ensuring the quality of the video image synthesis.

[0078] The above contents are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they shall all fall within the protection scope of the present invention.

Claims

1. A video image intelligent synthesis method based on artificial intelligence, characterized in that: The steps include: Step 1: Video image data acquisition and preprocessing: Acquire the video image data to be synthesized and perform preprocessing; Step 2: Selection of artificial intelligence models for video image synthesis: obtaining the amount and diversity of the video image data to be synthesized, as well as the objectives and requirements of the video image synthesis task, where the objectives and requirements include video style, content type, synthesis complexity, and synthesis accuracy, and obtaining the attribute information of each artificial intelligence model for video image synthesis, where the attribute information includes the computing resource requirement level and inference speed, analyzing the matching index of each artificial intelligence model for video image synthesis, and screening the adapted artificial intelligence model for video image synthesis; Step 3: Analyze and set video image synthesis parameters: According to the objectives and requirements of the video image synthesis task and the adapted artificial intelligence model of video image synthesis, analyze the basic parameters suitable for video image synthesis, where the basic parameters include resolution, frame rate, color space, contrast, brightness, saturation, and transparency, and further set the basic parameters of video image synthesis; Step 4: Video image synthesis effect detection and evaluation: Detect the effect index of video image synthesis, where the effect index includes peak signal-to-noise ratio, structural similarity index, feature similarity index, information entropy, and analyze the effect evaluation index of video image synthesis; Step 5: Determine and process the optimization needs of video image synthesis: Determine whether the video image synthesis needs to be optimized based on the effect evaluation index of the video image synthesis, and perform corresponding processing.

2. The method for intelligent synthesis of video images based on artificial intelligence according to claim 1, characterized in that: The specific analysis process of step 2 includes: Obtaining the accumulated data volume of the video image data to be synthesized; Obtain the number of scene changes, the number of lighting condition changes, and the number of shooting angle changes of the video image data to be synthesized, and substitute them into a preset relationship function between the number of scene changes, the number of lighting condition changes, the number of shooting angle changes, and the degree of diversity of the video image data to obtain the degree of diversity of the video image data to be synthesized; Extract the data volume range and diversity range of video image data acceptable to each artificial intelligence model for video image synthesis stored in the database; If the cumulative data volume of the video image data to be synthesized falls within the data volume range of the video image data acceptable to a certain artificial intelligence model, then the data volume of the video image data acceptable to the artificial intelligence model meets the requirements; otherwise, the data volume of the video image data acceptable to the artificial intelligence model does not meet the requirements. The compliance coefficients of the video image data volumes acceptable to the artificial intelligence model corresponding to the data volumes of the video image data acceptable to the artificial intelligence model that meet the requirements and do not meet the requirements are set, and the compliance coefficients of the video image data volumes acceptable to each artificial intelligence model for video image synthesis are obtained by screening, and recorded as δ i , i represents the number of the i-th artificial intelligence model of video image synthesis, i = 1, 2, ..., n; If the diversity degree of the video image data to be synthesized falls within the range of the diversity degree of video image data acceptable to a certain artificial intelligence model for video image synthesis, then the diversity degree of video image data acceptable to the artificial intelligence model meets the requirements; otherwise, the diversity degree of video image data acceptable to the artificial intelligence model does not meet the requirements. The compliance coefficient of the diversity degree of video images acceptable to the artificial intelligence model corresponding to the diversity degree of video image data acceptable to the artificial intelligence model that meets the requirements and does not meet the requirements is set, and the compliance coefficient of the diversity degree of video images acceptable to each artificial intelligence model for video image synthesis is screened and recorded as ε i ; By analyzing the formula Get the first matching index of each artificial intelligence model for video image synthesis Wherein γ1 represents a correction factor of a preset first matching index, β1 and β2 represent weights of a preset acceptable video image data volume conformity coefficient and an acceptable video image diversity conformity coefficient, respectively, and β1+β2=1.

3. The method for intelligent synthesis of video images based on artificial intelligence according to claim 2, characterized in that: The specific analysis process of step 2 also includes: Obtain the video style, content type, synthesis complexity and synthesis accuracy of the video image synthesis task; Extract the matching factors of each artificial intelligence model for video image synthesis and various video styles and the matching factors of each artificial intelligence model and each content type stored in the database, screen and obtain the matching factors of each artificial intelligence model for video image synthesis and the video style and content type of the video image synthesis task, and record them as Extract the acceptable complexity threshold and achievable accuracy threshold of each artificial intelligence model for video image synthesis stored in the database; If the synthesis complexity of the video image synthesis task is less than or equal to the acceptable complexity threshold of a certain artificial intelligence model for video image synthesis, then the acceptable complexity of the artificial intelligence model meets the requirements; otherwise, the acceptable complexity of the artificial intelligence model does not meet the requirements. The acceptable complexity compliance coefficients corresponding to the acceptable complexity of the artificial intelligence model that meets the requirements and does not meet the requirements are set, and the acceptable complexity compliance coefficients of each artificial intelligence model for video image synthesis are screened and recorded as If the synthesis accuracy of the video image synthesis task is less than or equal to the accuracy threshold of a certain artificial intelligence model for video image synthesis, then the accuracy of the artificial intelligence model meets the requirements; otherwise, the accuracy of the artificial intelligence model does not meet the requirements. The accuracy compliance coefficients corresponding to the accuracy of the artificial intelligence model that meets the requirements and does not meet the requirements are set, and the accuracy compliance coefficients of each artificial intelligence model for video image synthesis are screened and recorded as By analyzing the formula Get the second matching index of each artificial intelligence model for video image synthesis Wherein γ2 represents the correction factor of the preset second matching index, e represents the natural constant, η1, η2, η3, η4 represent the weights of the preset video style matching factor, content type matching factor, acceptable complexity compliance coefficient and precision compliance coefficient respectively, and η1+η2+η3+η4=1.

4. The method for intelligent synthesis of video images based on artificial intelligence according to claim 3, characterized in that: The specific analysis process of step 2 also includes: Extracting attribute information of each artificial intelligence model for video image synthesis stored in the database, and obtaining the computing resource requirement level and reasoning speed of each artificial intelligence model for video image synthesis; The computing resource level that the video image synthesis device can provide is obtained. If the computing resource requirement level of a certain artificial intelligence model of video image synthesis is equal to or lower than the computing resource level that the video image synthesis device can provide, then the computing resource requirement of the artificial intelligence model can be met. Otherwise, the computing resource requirement of the artificial intelligence model cannot be met. The computing resource requirement compliance coefficient of the artificial intelligence model corresponding to the computing resource requirement that can be met and cannot be met is set, and the computing resource requirement compliance coefficient of each artificial intelligence model of video image synthesis is screened and recorded as κ i ; The inference speed of each artificial intelligence model for video image synthesis is denoted as v i ; By analyzing the formula Get the third matching index of each artificial intelligence model for video image synthesis Wherein γ3 represents the correction factor of the preset third matching index, λ1 and λ2 represent the weights of the preset computing resource requirement compliance coefficient and inference speed respectively, λ1+λ2=1, and n represents the number of artificial intelligence models for video image synthesis.

5. The method for intelligent synthesis of video images based on artificial intelligence according to claim 4, characterized in that: The specific analysis process of step 2 also includes: The first matching index of each artificial intelligence model synthesized from video images Second matching index The third matching index Substitute into the analysis formula Get the matching index of each artificial intelligence model for video image synthesis Wherein, μ1, μ2, and μ3 represent the weights of the preset first matching index, second matching index, and third matching index respectively, and μ1+μ2+μ3=1.

6. The method for intelligent synthesis of video images based on artificial intelligence according to claim 1, characterized in that: The specific analysis process of step 2 also includes: The matching indexes of the various artificial intelligence models for video image synthesis are compared with each other, and the artificial intelligence model corresponding to the maximum matching index is used as the adapted artificial intelligence model for video image synthesis.

7. The method for intelligent synthesis of video images based on artificial intelligence according to claim 1, characterized in that: The specific analysis process of step three is as follows: Extract the case information of historical video image synthesis stored in the database, obtain the goals and demand information of each historical video image synthesis case, the selected artificial intelligence model and the basic parameters of video image synthesis, use the machine learning algorithm to train and obtain the basic parameter analysis model of video image synthesis, substitute the goals and demand information of the video image synthesis task and the adapted artificial intelligence model of video image synthesis into the basic parameter analysis model of video image synthesis, and obtain the appropriate basic parameters for video image synthesis, where the basic parameters include resolution, frame rate, color space, contrast, brightness, saturation, and transparency, and further set the basic parameters of video image synthesis.

8. The method for intelligent synthesis of video images based on artificial intelligence according to claim 1, characterized in that: The specific analysis process of step 4 includes: The synthesized video image is detected to obtain the peak signal-to-noise ratio, structural similarity index, feature similarity index, and information entropy of the video image synthesis, which are denoted as σ P , σ S , σ F , σ I .

9. The method for intelligent synthesis of video images based on artificial intelligence according to claim 8, characterized in that: The specific analysis process of step 4 also includes: The peak signal-to-noise ratio σ of the video image synthesis P , structural similarity index σ S , feature similarity index σ F 、Information entropy σ I Substitute into the analysis formula The effect evaluation index ξ of video image synthesis is obtained, where ψ1, ψ2, ψ3, and ψ4 represent the preset peak signal-to-noise ratio, structural similarity index, feature similarity index, and information entropy weights, respectively, and ψ1+ψ2+ψ3+ψ4=1.

10. The method for intelligent synthesis of video images based on artificial intelligence according to claim 1, characterized in that: The specific analysis process of step five is as follows: The effect evaluation index of the video image synthesis is compared with the preset effect evaluation index threshold. If the effect evaluation index of the video image synthesis is less than the preset effect evaluation index threshold, the video image synthesis needs to be optimized, and the optimization plan of the video image synthesis is obtained and feedback is given.

Images