Multi-scene adaptive video content recognition method and system based on artificial intelligence

Through the multi-scene adaptive video content recognition method based on artificial intelligence, the background elements and light elements in multi-scene videos are automatically identified and adjusted, which solves the problem of inconsistency in the existing technology and improves visual effects and efficiency.

CN118779490BActive Publication Date: 2024-12-03BEIJING MIAOYIN ANIMATION CULTURE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411281462.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-13
Publication Date
2024-12-03
Estimated Expiration
2044-09-13

AI Technical Summary

Technical Problem

In the prior art, the inconsistency between background elements and light elements caused by changes in camera angle and time during the production of multi-scene videos is difficult to efficiently adjust, affecting visual effects and efficiency.

Method used

Adopting multi-scene adaptive video content recognition method based on artificial intelligence, automatically identify background elements and light elements by obtaining video materials in multiple spatial scenes, and adjusting them using unified information to achieve color and light consistency between scenes. .

Benefits of technology

It significantly improves the overall visual effect and adjustment efficiency of multi-scene videos, ensuring the consistency of the scene at different angles and times.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118779490B_ABST
    Figure CN118779490B_ABST
Patent Text Reader

Abstract

The present invention discloses a multi-scene adaptive video content recognition method and system based on artificial intelligence, and relates to the technical field related to video recognition. The method comprises: obtaining scene video materials of a target space in multiple spatial scenes. Recognizing and extracting multiple background element images for multiple scene video materials, performing background element uniformity recognition, and obtaining multiple element uniformity information. Obtaining multiple light element images. According to multiple spatial angles, multiple light element images are recognized to obtain light angle information, and light element uniformity recognition is performed to obtain light uniformity information. According to multiple element uniformity information and light uniformity information, modification progressive analysis is performed to obtain modification amplitude information, and multiple scene video materials are uniformly modified and marked. The technical problem that the inconsistency of background elements and light elements caused by changes in camera angle and time in the multi-scene video production process in the prior art is difficult to efficiently adjust.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field related to video recognition technology, and specifically to a multi-scene adaptive video content recognition method and system based on artificial intelligence. Background Art

[0002] In virtual animation production and multi-scene video content processing, background elements and light elements in different scenes often show inconsistencies due to changes in camera angles and time. This inconsistency will lead to poor visual effects of the final video content, thus affecting the overall visual experience. In order to improve the visual consistency of multi-scene videos, it is usually necessary to uniformly identify and adjust background elements and light elements. However, the manual adjustment process in the prior art is time-consuming and error-prone, and it is difficult to ensure that the final rendering effect meets expectations.

[0003] Therefore, in the prior art, the inconsistency of background elements and light elements caused by changes in camera angle and time during the multi-scene video production process is difficult to adjust efficiently. Summary of the invention

[0004] This application solves the technical problem in the prior art that the inconsistency of background elements and light elements caused by changes in camera angle and time during multi-scene video production is difficult to adjust efficiently by providing a multi-scene adaptive video content recognition method and system based on artificial intelligence. It achieves the goal of automatically identifying background elements and light elements in multi-scene videos by using artificial intelligence technology, and adjusting them according to their uniformity information, so as to achieve color and light consistency between scenes, thereby significantly improving the overall visual effect and adjustment efficiency of multi-scene videos.

[0005] The present application provides a multi-scene adaptive video content recognition method based on artificial intelligence, the method comprising: obtaining scene video materials of a target space in multiple spatial scenes, wherein the multiple spatial scenes are constructed based on different multiple times and multiple spatial angles. Based on artificial intelligence, multiple background element images are identified and extracted from multiple scene video materials to obtain multiple background element image sets, and background element uniformity is identified based on the multiple spatial angles to obtain multiple element uniformity information. Light element extraction and identification are performed on the multiple scene video materials to obtain multiple light element images. According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information, and the light element uniformity is identified in combination with the multiple times to obtain light uniformity information. According to the multiple element uniformity information and light uniformity information, the modification progressive analysis of the multiple scene video materials is performed to obtain modification amplitude information, and the multiple scene video materials are uniformly modified as the multi-scene video content recognition result.

[0006] In the implementation method, multiple background element images are identified and extracted from multiple scene video materials to obtain multiple background element image sets, including: based on historical materials of animation spaces similar to the target space, a sample scene video material set is collected, and the multiple background element images in each sample scene video material are divided and labeled to obtain multiple sample background element image sets. Based on semantic segmentation, multiple background element image recognition branches are constructed to perform the identification and division of the multiple background element images. The sample scene video material set is used as input, and the multiple sample background element image sets are used as output respectively, and the multiple background element image recognition branches are supervised trained until convergence to obtain a background element image extractor. The background element image extractor is used to identify and extract multiple background element images from the multiple scene video materials, and classify them according to the multiple background elements to obtain multiple background element image sets of multiple background elements.

[0007] In an implementation, based on the multiple spatial angles, background element uniformity identification is performed to obtain multiple element uniformity information, including: constructing a background element image restorer based on the historical image data of the multiple background elements, wherein the background element image restorer includes background element image restoration branches corresponding to the multiple background elements, and each background element image restoration branch includes multiple spatial angle background element image restoration paths corresponding to multiple spatial angles. Based on the multiple background elements and multiple spatial angles, the background element images in the multiple background element image sets are input into the corresponding spatial angle background element image restoration paths to obtain multiple restored background element image sets of preset spatial angles. Based on the multiple standard background element images of the multiple background elements at the preset spatial angles, similarity analysis and uniformity calculation are performed on the restored background element images in the multiple restored background element image sets to obtain multiple element uniformity information.

[0008] In an implementation, a background element image restorer is constructed based on the historical image data of the multiple background elements, including: based on the historical image data of the multiple background elements, a plurality of sample first spatial angle background element image sets of the first background element at multiple spatial angles are collected, and a sample first restored background element image set of the first background element at a preset spatial angle is collected. Based on a generative adversarial network, a plurality of background element image restoration paths corresponding to multiple spatial angles in a first background element image restoration branch corresponding to the first background element are constructed. A plurality of sample first spatial angle background element image sets are respectively used as input, and the sample first restored background element image set is used as supervision, and supervised training is performed on the plurality of background element image restoration paths in the first background element image restoration branch until convergence. The historical image data of other background elements are continued to be used to train and obtain a plurality of background element image restoration branches to obtain a background element image restorer.

[0009] In the implementation method, the light element recognition and extraction are performed on the multiple scene video materials to obtain multiple light element images, including: based on the historical materials of the animation space similar to the target space, a set of sample scene video materials is collected, and the light element images in each sample scene video material are divided and labeled to obtain a set of sample light element images. Based on semantic segmentation, a light element image extractor is constructed. The light element image extractor is supervised trained until convergence using the sample scene video material set and the sample light element image set. The light element image extractor is used to recognize and extract light elements on the multiple scene video materials to obtain multiple light element images.

[0010] In the implementation, the multiple light element images are identified according to the multiple spatial angles to obtain multiple light angle information, and the light element uniformity is identified in combination with the multiple times to obtain the light uniformity information, including: obtaining a sample spatial angle set and a sample light element image set, and marking the light angles in the sample light element images at different spatial angles to obtain a sample light angle information set. The sample spatial angle set and the sample light element image set are used as inputs, and the sample light angle information set is used as supervision to train a light angle identifier, identify the multiple light element images and the multiple spatial angles, and obtain multiple light angle information. Multiple standard light angle information at the multiple times is obtained, and the light uniformity information is calculated based on the multiple light angle information and the multiple standard light angle information.

[0011] In the implementation, based on the plurality of element uniformity information and light uniformity information, the modification progressive identification analysis of the plurality of scene video materials is performed, including: based on the plurality of element uniformity information, multiple background element modification amplitudes are classified and obtained, wherein, based on a decision tree, an element modification amplitude classifier is constructed using sample element uniformity information and sample background element modification amplitudes for classification. Based on the light uniformity information, the light element modification amplitude is classified and obtained. Based on the plurality of element uniformity information and light uniformity information, the modification classification of background element light occlusion is performed to obtain a combined modification amplitude. Based on the plurality of background element modification amplitudes, light element modification amplitudes and combined modification amplitudes, modification amplitude information is calculated and obtained as a multi-scene video content recognition result.

[0012] This application also provides a multi-scene adaptive video content recognition system based on artificial intelligence, including:

[0013] The scene video acquisition module is used to acquire scene video materials of the target space in multiple space scenes, wherein the multiple space scenes are constructed based on different multiple times and multiple space angles.

[0014] The element uniformity acquisition module is used to identify and extract multiple background element images from multiple scene video materials based on artificial intelligence, obtain multiple background element image sets, identify the uniformity of background elements based on the multiple spatial angles, and obtain multiple element uniformity information.

[0015] The light element image module is used to extract and identify light elements from the multiple scene video materials to obtain multiple light element images.

[0016] The light uniformity acquisition module is used to identify the multiple light element images according to the multiple spatial angles to obtain multiple light angle information, and to identify the uniformity of the light elements in combination with the multiple times to obtain light uniformity information.

[0017] The modification identification module is used to perform progressive modification analysis of the multiple scene video materials based on the multiple element uniformity information and light uniformity information, obtain modification amplitude information as the multi-scene video content recognition result, and uniformly identify the multiple scene video materials.

[0018] The multi-scene adaptive video content recognition method and system based on artificial intelligence proposed in this application is proposed. By acquiring scene video materials of the target space in multiple spatial scenes, wherein the multiple spatial scenes are constructed based on different multiple times and multiple spatial angles. Based on artificial intelligence, multiple background element images are identified and extracted from the multiple scene video materials to obtain multiple background element image sets. Based on the multiple spatial angles, the background element uniformity is identified to obtain multiple element uniformity information. Light element extraction and identification are performed on the multiple scene video materials to obtain multiple light element images. According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information. Combined with the multiple times, the light element uniformity is identified to obtain light uniformity information. According to the multiple element uniformity information and light uniformity information, the modification progressive analysis of the multiple scene video materials is performed to obtain the modification amplitude information. As the result of multi-scene video content recognition, the multiple scene video materials are uniformly modified. The technical problem that the inconsistency of background elements and light elements caused by the changes in camera angles and time in the multi-scene video production process in the prior art is difficult to efficiently adjust. By utilizing artificial intelligence technology, it is possible to automatically identify background and light elements in multi-scene videos, and adjust them based on their unified information to achieve color and light consistency between scenes, thereby significantly improving the overall visual effect and adjustment efficiency of multi-scene videos. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0020] Figure 1 A schematic diagram of a multi-scenario adaptive video content recognition method based on artificial intelligence provided in an embodiment of the present application;

[0021] Figure 2 A schematic diagram of the structure of an artificial intelligence-based multi-scene adaptive video content recognition system provided in an embodiment of the present application.

[0022] Explanation of the reference numerals: scene video acquisition module 11 , element uniformity acquisition module 12 , light element image module 13 , light uniformity acquisition module 14 , modification identification module 15 . DETAILED DESCRIPTION

[0023] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0024] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0025] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0026] The present application provides a method and system for multi-scenario adaptive video content recognition based on artificial intelligence, such as Figure 1 As shown, the method includes:

[0027] Obtain scene video materials of the target space in multiple spatial scenes, wherein the multiple spatial scenes are constructed based on different multiple times and multiple spatial angles. Based on artificial intelligence, identify and extract multiple background element images from the multiple scene video materials to obtain multiple background element image sets, and identify the uniformity of background elements based on the multiple spatial angles to obtain multiple element uniformity information. Extract and identify light elements from the multiple scene video materials to obtain multiple light element images.

[0028] Obtain scene video materials of the target space in multiple spatial scenes, where the target space is a space constructed by a virtual animation scene, such as a virtual living room. The multiple spatial scenes are constructed based on multiple different times and multiple spatial angles, and the multiple spatial angles are the angles of the virtual camera in the virtual scene. The scene video materials in multiple spatial scenes include video materials shot at different times and different shooting angles of the virtual scene. Subsequently, based on artificial intelligence, multiple background element images are identified and extracted from the multiple scene video materials through a background element image extractor to obtain multiple background element image sets, and background element uniformity is identified based on the multiple spatial angles, and restored images generated by multiple spatial angle images of each element are obtained, uniformity is obtained, the uniformity of all elements is completed, and multiple element uniformity information is obtained. Further, light element extraction and identification are performed on the multiple scene video materials, and the light element is an image expressing light in the animation, and multiple light element images are obtained.

[0029] The method provided by the embodiment of the present application also includes: based on the historical materials of the animation space similar to the target space, collecting a set of sample scene video materials, and dividing and annotating the various background element images in each sample scene video material to obtain multiple sample background element image sets. Based on semantic segmentation, construct multiple background element image recognition branches for performing the recognition and division of the multiple background element images. Using the set of sample scene video materials as input and the multiple sample background element image sets as output, supervised training is performed on the multiple background element image recognition branches until convergence to obtain a background element image extractor. Using the background element image extractor, the multiple scene video materials are identified and extracted for multiple background element images, and classified according to the multiple background elements to obtain multiple background element image sets of multiple background elements.

[0030] The method comprises: performing identification and extraction of multiple background element images from multiple scene video materials to obtain a set of multiple background element images, including: selecting historical materials of an animation space similar to the target space from a historical database, which materials should cover virtual scenes at different time periods and angles, collecting a set of sample scene video materials, and using a semantic segmentation tool or a manual annotation tool to divide the multiple background element images of the video frames in the sample scene video materials, and performing category annotation to obtain a set of multiple sample background element images, wherein the set of sample background element images includes sample background element segmentation images and corresponding annotation category information. Subsequently, based on semantic segmentation, multiple background element image recognition branches are constructed for performing the identification and division of the multiple background element images, that is, background element image recognition branches are constructed through a language segmentation model, and each background element image recognition branch corresponds to a background element sample category. The set of sample scene video materials is used as input, and the multiple sample background element image sets are respectively used as outputs, and the multiple background element image recognition branches are supervised and trained until convergence to obtain a background element image extractor. Finally, the background element image extractor is used to identify and extract multiple background element images from the multiple scene video materials, and the multiple background elements are classified to obtain multiple background element image sets of the multiple background elements.

[0031] The method provided in the embodiment of the present application also includes: constructing a background element image restorer based on the historical image data of the multiple background elements, wherein the background element image restorer includes background element image restoration branches corresponding to the multiple background elements, and each background element image restoration branch includes multiple spatial angle background element image restoration paths corresponding to multiple spatial angles. Based on the multiple background elements and multiple spatial angles, the background element images in the multiple background element image sets are input into the corresponding spatial angle background element image restoration paths to obtain multiple restored background element image sets of preset spatial angles. Based on the multiple standard background element images of the multiple background elements at the preset spatial angles, similarity analysis and uniformity calculation are performed on the restored background element images in the multiple restored background element image sets to obtain multiple element uniformity information.

[0032] Based on the multiple spatial angles, background element uniformity identification is performed to obtain multiple element uniformity information, including: collecting historical image data of multiple background elements at different spatial angles from a historical database, and constructing a background element image restorer. The background element image restorer includes background element image restoration branches corresponding to the multiple background elements, and each background element image restoration branch includes multiple spatial angle background element image restoration paths corresponding to multiple spatial angles. Further, based on the multiple background elements and multiple spatial angles, the background element images in the multiple background element image sets are input into the corresponding spatial angle background element image restoration paths, and the spatial angle background element image restoration paths perform image restoration of preset spatial angles on each spatial angle image, and obtain multiple restored background element image sets of preset spatial angles generated by restoration, and each restored background element image set contains images of preset spatial angles generated after restoring images at different spatial angles. Finally, based on the multiple standard background element images of the multiple background elements at the preset spatial angles, similarity analysis and uniformity calculation are performed on the restored background element images in the multiple restored background element image sets. When performing the similarity analysis, the similarity between each background element restored image and the standard background element image corresponding to the preset spatial angle is calculated to complete the similarity analysis. Subsequently, the mean calculation is performed based on the similarity parameters of the background element at different spatial angles to obtain the variance of the background element similarity parameters and obtain the uniformity information of the background element. The same processing is performed on other background elements to obtain uniformity information of multiple elements.

[0033] The method provided in the embodiment of the present application also includes: based on the historical image data of the multiple background elements, collecting and obtaining multiple sample first spatial angle background element image sets of the first background element at multiple spatial angles, and collecting the sample first restored background element image set of the first background element at a preset spatial angle. Based on the generative adversarial network, construct multiple background element image restoration paths corresponding to multiple spatial angles within the first background element image restoration branch corresponding to the first background element. Using multiple sample first spatial angle background element image sets as input respectively, using the sample first restored background element image set as supervision, the multiple background element image restoration paths within the first background element image restoration branch are supervised and trained until convergence. Continue to use the historical image data of other background elements, train to obtain multiple background element image restoration branches, and obtain a background element image restorer.

[0034] According to the historical image data of the multiple background elements, a background element image restorer is constructed, including: according to the historical image data of the multiple background elements, a plurality of sample first spatial angle background element image sets of the first background element at multiple spatial angles are collected, that is, images of a random background element at multiple spatial angles in different virtual spaces are obtained from the historical image data of the multiple background elements, and a plurality of sample first spatial angle background element image sets are obtained. And a sample first restored background element image set of the first background element at a preset spatial angle is collected, and the preset spatial angle is a preset fixed spatial angle, and images of the first background element at preset spatial angles in different virtual spaces are collected to obtain a sample first restored background element image set. Further, based on a generative adversarial network, a plurality of background element image restoration paths corresponding to multiple spatial angles in the first background element image restoration branch corresponding to the first background element are constructed. Wherein, each background element image restoration path is constructed by a generative adversarial network, and each background element image restoration path corresponds to a specific spatial angle, and the input image is restored to an image at a preset spatial angle. Further, multiple sample first spatial angle background element image sets are respectively used as input, and the sample first restored background element image set is used as supervision, and multiple background element image restoration paths in the first background element image restoration branch are supervised and trained until convergence. Finally, historical image data of other background elements are continued to be used to train multiple background element image restoration branches to obtain a background element image restorer. Each background element image restoration branch corresponds to a background element.

[0035] The method provided by the embodiment of the present application also includes: based on the historical materials of the animation space similar to the target space, collecting a set of sample scene video materials, and dividing and labeling the light element images in each sample scene video material to obtain a set of sample light element images. Based on semantic segmentation, a light element image extractor is constructed. Using the sample scene video material set and the sample light element image set, the light element image extractor is supervised trained until convergence. Using the light element image extractor, the multiple scene video materials are subjected to light element recognition and extraction to obtain multiple light element images.

[0036] Based on the historical materials of the animation space similar to the target space, a set of sample scene video materials is collected from the animation space similar to the target space. A semantic segmentation tool or a manual annotation tool is used for the light element images in each sample scene video material to separate the various light element images in the video frame of the sample scene video material, and the image categories are annotated to obtain a set of sample light element images. Further, based on semantic segmentation, a light element image extractor is constructed. Using the sample scene video material set and the sample light element image set, supervised training is performed on the light element image extractor until convergence. Finally, the light element image extractor is used to perform light element recognition and extraction on the multiple scene video materials to obtain multiple light element images, and the multiple light element images have angle identifications.

[0037] According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information, and the light element uniformity is identified in combination with the multiple times to obtain light uniformity information. According to the multiple element uniformity information and light uniformity information, the modification progressive analysis of the multiple scene video materials is performed to obtain modification amplitude information as the multi-scene video content recognition result, and the multiple scene video materials are uniformly marked for modification.

[0038] According to the multiple spatial angles, the multiple light element images are identified, and the multiple light element images corresponding to different spatial angles are obtained, and the multiple light angle information of the image is obtained. Combined with the multiple times, the standard light angles at the corresponding time points are obtained, and the uniformity of the light elements is identified to obtain the uniformity information of the light elements. Finally, according to the multiple element uniformity information and the light uniformity information, the modification progressive analysis of the multiple scene video materials is performed, and the modification progressive analysis includes element modification amplitude analysis, light modification amplitude analysis and combined modification amplitude analysis. The above modification amplitude analysis is weighted and calculated to obtain the modification amplitude information. As the result of multi-scene video content recognition, the multiple scene video materials are uniformly modified and marked. The final scene video will visually show consistent color and light effects, ensuring that the scene is consistent at different angles and times. The technical problem that the inconsistency of background elements and light elements caused by changes in camera angles and time in the multi-scene video production process in the prior art is difficult to efficiently adjust is solved. By utilizing artificial intelligence technology, it is possible to automatically identify background and light elements in multi-scene videos, and adjust them based on their unified information to achieve color and light consistency between scenes, thereby significantly improving the overall visual effect and adjustment efficiency of multi-scene videos.

[0039] The method provided in the embodiment of the present application also includes: obtaining a sample space angle set and a sample light element image set, and annotating the light angles in the sample light element images at different space angles to obtain a sample light angle information set. Using the sample space angle set and the sample light element image set as input, and using the sample light angle information set as supervision, a light angle identifier is trained to identify the multiple light element images and multiple space angles to obtain multiple light angle information. Multiple standard light angle information at the multiple times is obtained, and light uniformity information is calculated based on the multiple light angle information and the multiple standard light angle information.

[0040] According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information, and combined with the multiple times, the uniformity of the light elements is identified to obtain the uniformity information of the light, including: obtaining a sample spatial angle set and a sample light element image set, and marking the light angles in the sample light element images under different spatial angles to obtain a sample light angle information set, wherein the sample light angle information set includes the sample spatial angle and the corresponding light element image and the specific light angle identifier. The sample spatial angle set and the sample light element image set are used as input data, and the sample light angle information set is used as supervision data to perform supervised training on the neural network model until the output accuracy of the model meets the threshold, and a trained light angle identifier is obtained. The multiple light element images and the multiple spatial angles are identified to obtain multiple light angle information. The light angle information is the light angle of the light element image obtained under different spatial angle conditions. In order to ensure the uniformity of the light elements under different viewing angles, the light angles of the light element images obtained under different spatial angle conditions are obtained, and the angle uniformity is determined by the standard light angle to ensure the uniformity of the images under different viewing angles in the virtual environment. Finally, multiple standard light angle information at the multiple times is obtained, that is, different standard light angle information at different times in the virtual space is obtained. Light uniformity information is calculated based on the multiple light angle information and the multiple standard light angle information at the corresponding time. When calculating light uniformity, the time corresponding to the light angle information is obtained to obtain standard light angle information, and the standard light angle information is used as the mean parameter to calculate the standard deviation of the light angle information at different viewing angles to obtain light uniformity information.

[0041] The method provided by the embodiment of the present application also includes: classifying and obtaining multiple background element modification amplitudes according to the plurality of element uniformity information, wherein, based on a decision tree, an element modification amplitude classifier is constructed using sample element uniformity information and sample background element modification amplitudes for classification. Classifying and obtaining light element modification amplitudes according to the light uniformity information. Classifying and obtaining background element light occlusion modification according to the plurality of element uniformity information and light uniformity information to obtain a combined modification amplitude. Calculating and obtaining modification amplitude information according to the plurality of background element modification amplitudes, light element modification amplitudes and combined modification amplitudes as a multi-scene video content recognition result.

[0042] According to the multiple element uniformity information and light uniformity information, the modification progressive identification analysis of the multiple scene video materials is performed, including: according to the multiple element uniformity information, multiple background element modification amplitudes are classified and obtained, and the background element modification amplitude is the proportion of adjustment to the background element. When obtaining multiple background element modification amplitudes, they are obtained through an element modification amplitude classifier. The element modification amplitude classifier is constructed based on a decision tree, and the decision tree model is trained using sample element uniformity information and sample background element modification amplitude to complete the construction of the element modification amplitude classifier. According to the light uniformity information, the light element modification amplitude is classified and obtained, wherein each light uniformity information corresponds to a specific light element modification amplitude, and the light element modification amplitude is the amplitude of adjustment to the light element.

[0043] Further, in some scenes, there may be interactions between background elements and light elements, and the percentage of interaction is based on artificial preset identification. Based on the interaction relationship, the modification amplitude of the background element and the modification amplitude of the light element are combined, and the product calculation of the uniformity information of multiple elements and the uniformity information of light is performed with the interaction relationship as the influencing parameter. The influencing parameter of the interaction is the percentage of interaction, and the modification classification of the light occlusion of the background element is performed to obtain the combined modification amplitude. By adjusting the combined modification amplitude, the background elements and the light elements are not only consistent in different scenes, but also can maintain the overall visual effect consistency after the combination. For example, if the color and light intensity of the building need to be adjusted in multiple scenes, these two parameters need to be adjusted at the same time to ensure the consistency of the final rendering effect. Finally, according to the modification amplitudes of the multiple background elements, the modification amplitudes of the light elements and the combined modification amplitudes, a weighted calculation of the modification amplitude is performed, wherein the weight is set based on the preference of the technician. If the preference is the overall effect, the corresponding combined modification amplitude weight is higher, and the modification amplitude information is obtained by weight calculation as the multi-scene video content recognition result. Ensure that the elements and light performances between the scenes are more consistent, thereby improving the overall quality of the multi-scene video.

[0044] In the above, refer to Figure 1 The multi-scene adaptive video content recognition method based on artificial intelligence according to an embodiment of the present invention is described in detail. Figure 2 The present invention describes an artificial intelligence-based multi-scene adaptive video content recognition system according to an embodiment of the present invention.

[0045] The multi-scene adaptive video content recognition system based on artificial intelligence according to the embodiment of the present invention solves the technical problem in the prior art that the inconsistency of background elements and light elements caused by changes in camera angle and time during the multi-scene video production process is difficult to adjust efficiently. By utilizing artificial intelligence technology, the background elements and light elements in the multi-scene video are automatically identified, and adjusted according to their uniformity information, the color and light consistency between the scenes is achieved, thereby significantly improving the overall visual effect and adjustment efficiency of the multi-scene video. The multi-scene adaptive video content recognition system based on artificial intelligence includes: a scene video acquisition module 11, an element uniformity acquisition module 12, a light element image module 13, a light uniformity acquisition module 14, and a modification identification module 15.

[0046] The scene video acquisition module 11 is used to acquire scene video materials of the target space in multiple space scenes, wherein the multiple space scenes are constructed based on different multiple times and multiple space angles.

[0047] The element uniformity acquisition module 12 is used to identify and extract multiple background element images from multiple scene video materials based on artificial intelligence, obtain multiple background element image sets, identify the uniformity of background elements based on the multiple spatial angles, and obtain multiple element uniformity information.

[0048] The light element image module 13 is used to extract and identify light elements from the multiple scene video materials to obtain multiple light element images.

[0049] The light uniformity acquisition module 14 is used to identify the multiple light element images according to the multiple spatial angles to obtain multiple light angle information, and to identify the uniformity of the light elements in combination with the multiple times to obtain light uniformity information.

[0050] The modification identification module 15 is used to perform modification progressive analysis of the multiple scene video materials according to the multiple element uniformity information and light uniformity information, obtain modification amplitude information as the multi-scene video content recognition result, and uniformly identify the multiple scene video materials.

[0051] The specific configuration of the element uniformity acquisition module 12 will be described in detail below. The element uniformity acquisition module 12 may further include: identifying and extracting multiple background element images from multiple scene video materials to obtain multiple background element image sets, including: based on historical materials of animation spaces similar to the target space, collecting a sample scene video material set, and dividing and annotating the multiple background element images in each sample scene video material to obtain multiple sample background element image sets; based on semantic segmentation, constructing multiple background element image recognition branches for performing the identification and division of the multiple background element images; using the sample scene video material set as input, and using the multiple sample background element image sets as output, respectively, and performing supervised training on the multiple background element image recognition branches until convergence to obtain a background element image extractor; using the background element image extractor, identifying and extracting multiple background element images from the multiple scene video materials, and classifying them according to the multiple background elements to obtain multiple background element image sets of multiple background elements.

[0052] The specific configuration of the element uniformity acquisition module 12 will be described in detail below. The element uniformity acquisition module 12 further includes: based on the multiple spatial angles, background element uniformity recognition is performed to obtain multiple element uniformity information, including: according to the historical image data of the multiple background elements, a background element image restorer is constructed, wherein the background element image restorer includes background element image restoration branches corresponding to the multiple background elements, and each background element image restoration branch includes multiple spatial angle background element image restoration paths corresponding to multiple spatial angles; based on the multiple background elements and multiple spatial angles, the background element images in the multiple background element image sets are input into the corresponding spatial angle background element image restoration paths to obtain multiple restored background element image sets of preset spatial angles; according to the multiple standard background element images of the multiple background elements at the preset spatial angles, similarity analysis and uniformity calculation are performed on the restored background element images in the multiple restored background element image sets to obtain multiple element uniformity information.

[0053] The specific configuration of the element uniformity acquisition module 12 will be described in detail below. The element uniformity acquisition module 12 may further include: constructing a background element image restorer based on the historical image data of the multiple background elements, including: collecting and obtaining multiple sample first spatial angle background element image sets of the first background element at multiple spatial angles, and collecting a sample first restored background element image set of the first background element at a preset spatial angle based on the historical image data of the multiple background elements. Based on the generative adversarial network, construct multiple background element image restoration paths corresponding to multiple spatial angles in the first background element image restoration branch corresponding to the first background element. Use multiple sample first spatial angle background element image sets as inputs respectively, use the sample first restored background element image set as supervision, and supervise the multiple background element image restoration paths in the first background element image restoration branch until convergence. Continue to use the historical image data of other background elements, train to obtain multiple background element image restoration branches, and obtain a background element image restorer.

[0054] The specific configuration of the light element image module 13 will be described in detail below. The light element image module 13 further includes: performing light element recognition and extraction on the multiple scene video materials to obtain multiple light element images, including: based on historical materials of animation spaces similar to the target space, collecting a set of sample scene video materials, and dividing and labeling the light element images in each sample scene video material to obtain a set of sample light element images. Based on semantic segmentation, a light element image extractor is constructed. Using the sample scene video material set and the sample light element image set, supervised training is performed on the light element image extractor until convergence. Using the light element image extractor, light element recognition and extraction are performed on the multiple scene video materials to obtain multiple light element images.

[0055] The specific configuration of the light uniformity acquisition module 14 will be described in detail below. The light uniformity acquisition module 14 further includes: obtaining multiple light angle information, combining the multiple times, identifying the uniformity of light elements, and obtaining light uniformity information, including: obtaining a sample spatial angle set and a sample light element image set, and marking the light angles in the sample light element images at different spatial angles to obtain a sample light angle information set. Using the sample spatial angle set and the sample light element image set as input, using the sample light angle information set as supervision, training a light angle identifier, identifying the multiple light element images and multiple spatial angles, and obtaining multiple light angle information. Obtain multiple standard light angle information at the multiple times, and calculate and obtain light uniformity information based on the multiple light angle information and the multiple standard light angle information.

[0056] Below, the specific configuration of the modification identification module 15 will be described in detail. The modification identification module 15 further includes: performing modification progressive identification analysis of the multiple scene video materials according to the multiple element uniformity information and the light uniformity information, including: classifying and obtaining multiple background element modification amplitudes according to the multiple element uniformity information, wherein, based on a decision tree, an element modification amplitude classifier is constructed using sample element uniformity information and sample background element modification amplitudes for classification. According to the light uniformity information, the light element modification amplitude is classified. According to the multiple element uniformity information and the light uniformity information, the modification classification of the background element light occlusion is performed to obtain the combined modification amplitude. According to the multiple background element modification amplitudes, the light element modification amplitudes and the combined modification amplitudes, the modification amplitude information is calculated and obtained as the multi-scene video content recognition result.

[0057] The artificial intelligence-based multi-scene adaptive video content recognition system provided in the embodiment of the present invention can execute the artificial intelligence-based multi-scene adaptive video content recognition method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0058] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or the server, and the various units and modules included are only divided according to functional logic, but are not limited to the above divisions, as long as the corresponding functions can be achieved. In addition, the specific names of the functional units are only for the convenience of distinguishing from each other, and are not used to limit the scope of protection of the present invention.

[0059] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A multi-scene adaptive video content recognition method based on artificial intelligence, characterized in that: The method comprises: Acquire scene video materials of a target space in multiple space scenes, wherein the multiple space scenes are constructed based on different multiple times and multiple space angles; Based on artificial intelligence, multiple background element images are identified and extracted from multiple scene video materials to obtain multiple background element image sets, and the uniformity of the background elements is identified based on the multiple spatial angles to obtain multiple element uniformity information; Extracting and identifying light elements from the multiple scene video materials to obtain multiple light element images; According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information, and the light element uniformity is identified in combination with the multiple times to obtain light uniformity information; According to the uniformity information of the multiple elements and the uniformity information of the light, a progressive modification analysis is performed on the multiple scene video materials to obtain modification amplitude information as a multi-scene video content recognition result, and a unified modification mark is performed on the multiple scene video materials; Based on the multiple spatial angles, background element uniformity is identified to obtain multiple element uniformity information, including: According to the historical image data of the multiple background elements, a background element image restorer is constructed, wherein the background element image restorer includes background element image restoration branches corresponding to the multiple background elements, and each background element image restoration branch includes multiple spatial angle background element image restoration paths corresponding to multiple spatial angles; Based on the multiple background elements and multiple spatial angles, input the background element images in the multiple background element image sets into the corresponding spatial angle background element image restoration path to obtain multiple restored background element image sets of preset spatial angles generated by restoration; According to a plurality of standard background element images of the plurality of background elements at a preset spatial angle, performing similarity analysis and uniformity calculation on the restored background element images in the plurality of restored background element image sets to obtain a plurality of element uniformity information; According to the multiple spatial angles, the multiple light element images are identified to obtain multiple light angle information, and the light element uniformity is identified in combination with the multiple times to obtain light uniformity information, including: Obtaining a sample spatial angle set and a sample light element image set, and marking the light angles in the sample light element images at different spatial angles to obtain a sample light angle information set; Using the sample spatial angle set and the sample light element image set as input, using the sample light angle information set as supervision, training a light angle identifier, identifying the multiple light element images and multiple spatial angles, and obtaining multiple light angle information; A plurality of standard light angle information at the plurality of times is acquired, and light uniformity information is calculated based on the plurality of light angle information and the plurality of standard light angle information.

2. The multi-scene adaptive video content recognition method based on artificial intelligence according to claim 1 is characterized in that: Identify and extract multiple background element images from multiple scene video materials to obtain multiple background element image sets, including: Based on historical materials of animation spaces similar to the target space, a set of sample scene video materials is collected, and a plurality of background element images in each sample scene video material are divided and labeled to obtain a plurality of sample background element image sets; Based on semantic segmentation, construct multiple background element image recognition branches for performing the multiple background element image recognition divisions; Using the sample scene video material set as input, using the multiple sample background element image sets as output respectively, performing supervised training on the multiple background element image recognition branches until convergence, and obtaining a background element image extractor; The background element image extractor is used to identify and extract multiple background element images from the multiple scene video materials, and the multiple background elements are classified to obtain multiple background element image sets of the multiple background elements.

3. The multi-scene adaptive video content recognition method based on artificial intelligence according to claim 1 is characterized in that: According to the historical image data of the multiple background elements, a background element image restorer is constructed, comprising: According to the historical image data of the multiple background elements, a plurality of sample first spatial angle background element image sets of the first background element at a plurality of spatial angles are acquired, and a sample first restored background element image set of the first background element at a preset spatial angle is acquired; Based on a generative adversarial network, construct multiple background element image restoration paths corresponding to multiple spatial angles in a first background element image restoration branch corresponding to the first background element; Using a plurality of sample first spatial angle background element image sets as input, using the sample first restored background element image set as supervision, and performing supervised training on a plurality of background element image restoration paths in the first background element image restoration branch until convergence; Continue to use historical image data of other background elements to train multiple background element image restoration branches to obtain a background element image restorer.

4. The multi-scene adaptive video content recognition method based on artificial intelligence according to claim 1 is characterized in that: The light element recognition and extraction is performed on the multiple scene video materials to obtain multiple light element images, including: Based on historical materials of animation spaces similar to the target space, a sample scene video material set is collected, and light element images in each sample scene video material are divided and labeled to obtain a sample light element image set; Based on semantic segmentation, build a light element image extractor; Using the sample scene video material set and the sample light element image set, supervised training is performed on the light element image extractor until convergence; The light element image extractor is used to identify and extract light elements from the multiple scene video materials to obtain multiple light element images.

5. The multi-scene adaptive video content recognition method based on artificial intelligence according to claim 1 is characterized in that: According to the plurality of element uniformity information and light uniformity information, modification progressive recognition analysis of the plurality of scene video materials is performed, including: According to the plurality of element uniformity information, a plurality of background element modification amplitudes are obtained by classification, wherein based on a decision tree, an element modification amplitude classifier is constructed by using the sample element uniformity information and the sample background element modification amplitudes for classification; According to the light uniformity information, classify and obtain the light element modification amplitude; According to the plurality of element uniformity information and light uniformity information, modification classification of light occlusion of background elements is performed to obtain a combined modification range; According to the modification amplitudes of the multiple background elements, the modification amplitudes of the light elements and the combination modification amplitudes, modification amplitude information is calculated and obtained as a multi-scene video content recognition result.

6. The multi-scene adaptive video content recognition system based on artificial intelligence is characterized by: The system is used to execute the method according to any one of claims 1 to 5, and the system comprises: A scene video acquisition module, used to acquire scene video materials of a target space in multiple space scenes, wherein the multiple space scenes are constructed based on different multiple times and multiple space angles; An element uniformity acquisition module is used to identify and extract multiple background element images from multiple scene video materials based on artificial intelligence, obtain multiple background element image sets, and identify the uniformity of background elements based on the multiple spatial angles to obtain multiple element uniformity information; A light element image module is used to extract and identify light elements from the multiple scene video materials to obtain multiple light element images; A light uniformity acquisition module, configured to identify the plurality of light element images according to the plurality of spatial angles to obtain a plurality of light angle information, and identify the uniformity of the light elements in combination with the plurality of times to obtain light uniformity information; The modification identification module is used to perform progressive modification analysis of the multiple scene video materials based on the multiple element uniformity information and light uniformity information, obtain modification amplitude information as the multi-scene video content recognition result, and uniformly identify the multiple scene video materials.