CN-122027814-A - Method and device for compliance processing video and electronic equipment

CN122027814ACN 122027814 ACN122027814 ACN 122027814ACN-122027814-A

Abstract

The invention provides a method and a device for processing video in compliance and electronic equipment. The method comprises the steps of obtaining an original video frame sequence to be processed, detecting forbidden content of the original video frame sequence through an image compliance detection model to identify a content forbidden sub-area of the original video frame sequence, carrying out content restoration on the content forbidden sub-area of the original video frame sequence through an image compliance restoration model to generate a video frame after compliance restoration, and outputting a video stream based on the video frame after compliance restoration. The method and the device realize real-time, automatic and refined identification and generation type restoration of the forbidden visual content in the video stream, can seamlessly replace the illegal picture element with the compliant content under the condition of no interruption, and effectively solve the problems that the traditional video content supervision relies on manual auditing, processing lag and influences viewing experience.

Inventors

WU LI
ZHANG HAO
WENG WENMIN

Assignees

瑞芯微电子股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260123

Claims (14)

1. A method of compliance processing video, comprising: acquiring an original video frame sequence to be processed; Detecting forbidden content of the original video frame sequence through an image compliance detection model so as to identify a content forbidden subarea of the original video frame sequence; Performing content restoration on the content forbidden sub-region of the original video frame sequence through an image compliance restoration model to generate a video frame after compliance restoration, and And outputting a video stream based on the video frame after the compliance repair.
2. The method as recited in claim 1, further comprising: acquiring a preconfigured enhanced content region and enhanced content to be implanted, and Blending the enhancement content into an image position corresponding to the enhancement content region in the compliance-repaired video frame through an image compliance enhancement model to generate a compliance-repaired and enhanced video frame, Wherein outputting a video stream based on the compliance repaired video frame comprises outputting the video stream based on the compliance repaired and enhanced video frame.
3. The method of claim 1, wherein performing illicit content detection on the original video frame sequence by an image compliance detection model to identify content illicit sub-regions of the original video frame sequence comprises: And inputting the original video frame sequence into the image compliance detection model for analysis to identify the content forbidden sub-region, wherein the image compliance detection model comprises a YOLO model which is trained by adopting forbidden image library increment or a multi-mode visual language model which supports fuzzy search.
4. The method of claim 1, wherein content repairing the content forbidden sub-regions of the original video frame sequence by means of an image compliance repair model comprises: inputting the content forbidden sub-region and the original video frame sequence into the image compliance repair model to generate a replacement image of the content forbidden sub-region; Performing spatial fusion of the replacement image of the content forbidden sub-region with the texture, color and structure of the original video frame sequence, and And carrying out pixel-level synthesis on the replacement image of the content forbidden sub-area and the area which is not marked as forbidden in the original image.
5. The method of claim 1, wherein content repairing the content forbidden sub-regions of the original video frame sequence by means of an image compliance repair model comprises: The size of the content forbidden sub-area is expanded in space by a preset multiple to be used as a reference size, a reference area corresponding to the content forbidden sub-area is obtained according to the reference size, and And utilizing the image compliance restoration model to regenerate image contents of the content forbidden sub-areas based on the reference area and the original video frame sequence.
6. The method of claim 5, wherein using the image compliance repair model to regenerate image content for the content forbidden sub-region based on the reference region and the original video frame sequence comprises: extracting spatial features from a current video frame containing the content forbidden sub-region; Extracting timing features from the reference region of at least one video frame nearest to the current video frame, and And taking the spatial characteristics and the time sequence characteristics as condition information, inputting the condition information into an image generation model, and generating the repair content for replacing the content forbidden sub-area by the image generation model.
7. The method of claim 6, wherein inputting the spatial feature and the temporal feature as condition information to an image generation model comprises: Fusing the spatial features and the temporal features to generate a condition vector, and The condition vector is input to the image generation model.
8. The method of claim 2, wherein blending the enhancement content into the image locations corresponding to the enhancement content regions in the compliance-repaired video frame via an image compliance enhancement model comprises: Determining the enhanced content area as a target area; Based on background characteristics corresponding to a target area of a current video frame, carrying out spatial characteristic fusion on the enhanced content; performing timing feature consistency processing on the enhanced content based on neighboring video frames of the current video frame, and And fusing the enhanced content with the background of the target area by using an image generation model.
9. The method as recited in claim 2, further comprising: Receiving and analyzing configuration information provided by the system, wherein the configuration information comprises coordinate, size, transparency parameter and implantation time point of the blended area, and And based on the analyzed configuration information, merging the enhancement content to be implanted into the image position corresponding to the enhancement content region through the image compliance enhancement model.
10. The method as recited in claim 1, further comprising: and performing incremental training on the image compliance detection model based on the newly added forbidden image sample and the corresponding label.
11. The method of claim 2, wherein outputting the video stream based on the compliance repaired and enhanced video frames comprises: Performing hardware video coding on the video frames subjected to the compliance repair and enhancement to generate a coded video stream; packaging the coded video stream according to a streaming media format to generate a compliant live stream, and Pushing the compliant live stream to a live server.
12. An apparatus for compliance processing video, comprising: The image acquisition module is configured to acquire an original video frame sequence to be processed; the image compliance detection module is configured to detect forbidden contents of the original video frame sequence through an image compliance detection model so as to identify a content forbidden sub-area of the original video frame sequence; An image compliance restoration module configured to perform content restoration on the content forbidden sub-regions of the original video frame sequence through an image compliance restoration model to generate a compliance restored video frame, and And the image output module is configured to output a video stream based on the video frame after the compliance repair.
13. The apparatus as recited in claim 12, further comprising: An image compliance enhancement module configured to obtain a preconfigured enhancement content region and enhancement content to be implanted, and to blend the enhancement content into an image location corresponding to the enhancement content region in the compliance-repaired video frame via an image compliance enhancement model to generate a compliance-repaired and enhanced video frame, Wherein the image output module is configured to output the video stream based on the compliance repaired and enhanced video frames.
14. An electronic device, comprising: A memory configured to store an executable program, and A processor configured to execute the program to perform the method according to any one of claims 1 to 11.

Description

Method and device for compliance processing video and electronic equipment Technical Field The present invention relates to the field of video processing technologies, and in particular, to a method and an apparatus for processing video in compliance, and an electronic device. Background With the rapid development of the network live broadcast technology, the network live broadcast industry is accelerated, and the network live broadcast is increasingly widely applied in social media, entertainment platforms, live broadcast by electronic commerce, educational live broadcast, live broadcast with goods, live broadcast of news media, product release meeting and other technical scenes. There are also several problems in the administration of live video content, such as illegal bad information, false advertising, disclosure of corporate sensitive information, exposure of personal privacy, insufficient consciousness of the user's discipline, lag in technological development and administration means, etc. Disclosure of Invention The invention provides a method and a device for compliance processing video and electronic equipment, which can realize real-time compliance processing of video data. In one aspect of the invention, a method of compliance processing video is provided. The method comprises the steps of obtaining an original video frame sequence to be processed, detecting forbidden content of the original video frame sequence through an image compliance detection model to identify a content forbidden sub-area of the original video frame sequence, carrying out content restoration on the content forbidden sub-area of the original video frame sequence through an image compliance restoration model to generate a video frame after compliance restoration, and outputting the video frame based on the video frame after compliance restoration. In another aspect of the invention, an apparatus for compliance processing video is provided. The device comprises an image acquisition module, an image compliance detection module, an image compliance restoration module and an image output module, wherein the image acquisition module is configured to acquire an original video frame sequence to be processed, the image compliance detection module is configured to carry out forbidden content detection on the original video frame sequence through an image compliance detection model so as to identify a content forbidden sub-area of the original video frame sequence, the image compliance restoration module is configured to carry out content restoration on the content forbidden sub-area of the original video frame sequence through an image compliance restoration model so as to generate a video frame after compliance restoration, and the image output module is configured to output a video stream based on the video frame after compliance restoration. In yet another aspect of the present invention, an electronic device is provided. The electronic device comprises a memory configured to store a computer program and a processor configured to execute the computer program to perform the method of compliance processing video described above. According to the technical scheme, the original video frame sequence to be processed is obtained, and the image compliance detection model is utilized to carry out real-time and automatic forbidden content detection, so that the forbidden content subareas in the image can be accurately identified, the efficient discovery of the visual illegal content which is difficult to cover in traditional manual auditing or simple rule filtering is realized, and the content supervision loopholes caused by auditing timeliness and subjectivity are effectively eliminated. The identified forbidden sub-areas and the corresponding video frames are input into an image compliance restoration model to carry out intelligent content restoration, so that the refinement and the generation type replacement of the illegal picture elements are realized. Under the condition that manual real-time intervention is not needed, stable, objective and efficient automatic real-time compliance processing of live video streams is realized, content risks and operation cost are reduced, and an end-to-end solution taking safety compliance and high-quality impression into consideration is provided. Drawings FIG. 1 is a flow chart of a method of compliance processing video according to an embodiment of the present invention; FIG. 2 is an overall environmental architecture diagram of a method of compliance processing video in accordance with an embodiment of the present invention; FIG. 3 is a flow chart of a method of compliance processing video according to an embodiment of the present invention; FIG. 4 is a flow chart of repairing a content violation sub-area using an image compliance repair model according to an embodiment of the present invention; FIG. 5 is a flow chart of an image compliance enhancement model enhancement content binding sub-area