Person-Product Interaction Estimation Using Depth Layer Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to accurately determine interactions between individuals and products in offline markets using video analysis, limiting the ability to collect valuable consumer interest data for marketing and sales strategies.
Innovation Solution
A method and system that analyze the relationship between a product region and a detected person in video frames to determine contact, estimating interaction information without requiring depth information, using a combination of keypoint estimation and object tracking algorithms, and validating the estimated interactions to filter out errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional video analysis methods are used to determine person-product interaction, then the system can process video data, but it cannot accurately determine whether contact is made between person and product
Solution Approach 1:
The method segments the video frame into multiple depth layers (first depth layer, second depth layer, third depth layer) to distinguish between different spatial zones. By dividing the measurement space into segments, the system can accurately determine whether a person's hand contacts the product by checking which depth layer the hand occupies, thereby resolving the accuracy problem in contact detection.
Solution Approach 2:
The patent introduces a depth dimension to the traditional 2D video analysis. By mapping hand positions to specific depth layers (first, second, or third depth layer), the system adds a third dimension (depth) to the detection problem, enabling accurate contact determination without requiring complex 3D reconstruction algorithms.
2Measurement precision
If depth information is collected to accurately determine contact, then measurement precision improves, but device complexity and computational load increase
Solution Approach 1:
The method extracts only the necessary depth information by dividing the video frame into discrete depth layers rather than processing complete 3D point cloud data. This extraction approach retains the essential depth discrimination capability while significantly reducing computational complexity and avoiding the need for complex depth estimation algorithms.
Solution Approach 2:
The system changes the parameter representation from continuous depth values to discrete depth layer assignments (first, second, or third depth layer). This parameter discretization simplifies the computational processing while maintaining the ability to accurately determine contact, effectively reducing device complexity without sacrificing measurement precision.
3Loss of information
If comprehensive interaction analysis is performed, then information completeness improves, but processing time increases
Solution Approach 1:
The method performs partial action by analyzing only the critical depth layer information needed to determine contact, rather than processing all possible interaction parameters. This selective analysis approach maintains sufficient information completeness for contact detection while significantly reducing processing time and computational resources required.
Data Source
AI summary
An interaction information estimation method is provided. The method includes the steps of: determining, on the basis of a relationship between a product region and a first person detected in a plurality of video frames, whether contact is made between the first person and a product included in the product region, thereby estimating information on an interaction between the first person and the product on the basis of a result of the determination; and validating the estimated information on the interaction.