Video Advertisement Insertion on Blank Walls Using Human Body Intersection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep convolutional neural networks struggle to automatically insert advertisements in videos without specific targets or objects, particularly on blank walls, leading to unsatisfactory user experiences due to inappropriate advertisement placement.
Innovation Solution
A method using deep learning to identify human body and blank wall areas in videos, determining an advertisement insertion position that intersects with the human body area, ensuring a natural and unobtrusive placement by adjusting the rectangular area's position and size to meet specific conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If DCNN is used to detect targets/objects for advertisement replacement, then automatic advertisement insertion can be achieved, but advertisement insertion fails when no specific targets or objects are present in the video
Solution Approach 1:
Instead of detecting objects to place advertisements, the patent inverts the approach by detecting blank wall areas and using human body positions to determine advertisement placement. The system identifies regions without objects and uses the presence of human bodies as the criterion for advertisement insertion, rather than requiring specific detectable targets.
Solution Approach 2:
The patent changes the detection parameters from object-specific features to spatial and contextual features. It uses blank wall area detection and human body position detection as new parameters, replacing the traditional object detection parameters. This allows the system to work in scenarios where no specific objects are present by focusing on spatial relationships and background characteristics.
2Measurement precision
If advertisement is inserted on blank wall, then advertisement visibility is improved, but user experience deteriorates when advertisement blocks important video content
Solution Approach 1:
The patent applies local quality by making the advertisement placement dependent on local video characteristics. It analyzes each region of the video frame to determine if it contains a blank wall and whether a human body is present, then makes placement decisions based on these local conditions. This ensures advertisements are placed in appropriate locations without blocking important content.
Solution Approach 2:
The patent introduces the human body as an intermediary element for advertisement placement. Instead of directly placing advertisements on blank walls, it uses the human body's position and shape as a mediator to determine the optimal placement location. The advertisement is positioned to intersect with the human body area, allowing the body to naturally frame or partially obscure the advertisement, making it less intrusive.
3Productivity
If random position is selected for advertisement insertion, then insertion speed is maintained, but placement reasonableness deteriorates
Solution Approach 1:
The patent performs preliminary analysis of the video frame by detecting blank wall areas and human body positions before determining advertisement placement. This preliminary action of identifying suitable regions and understanding the spatial context allows for rapid yet reasoned placement decisions, maintaining speed while improving placement quality.
Solution Approach 2:
The patent changes from random position selection to position selection based on detected parameters (blank wall area coordinates and human body position). By using these detected parameters to guide placement, the system maintains computational efficiency while achieving more reasonable and context-appropriate advertisement positions.
Data Source
AI summary
A method for searching an AD insertion position includes following steps: S1. identifying a human body area in a video image; S2. identifying a blank wall area in the video image; and S3. determining the AD insertion position in the blank wall area in the video image, so that the AD insertion position intersects the human body area. A method for automatically inserting an advertisement AD in a video includes following steps: determining an AD insertion position in the video by searching a product AD insertion position based on deep learning, and inserting the AD on the position. Through the method, an appropriate position for an AD can be automatically searched on a blank wall part in the video. In addition, the inserted AD can be presented in the video in a natural way, and thereby avoiding or reducing the impact of AD insertion on impression of the video.