Store Gaze Tracking Using Virtual Raycasting From Head Pose
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies lack effective methods to track customer interactions with products in a store environment without requiring costly hardware installations and intrusive image processing, which can compromise privacy and are prone to transient events.
Innovation Solution
A computer-implemented method using a camera with known position and orientation to determine a human's gaze location by analyzing images and raycasting within a virtual representation of the store environment, allowing for reduced hardware requirements and privacy preservation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If vision-based solutions with multiple cameras are used to track customer gaze, then measurement precision is improved, but device complexity and cost increase significantly
Solution Approach 1:
The patent creates a virtual copy of the store environment with virtual structures representing physical products and shelving. Instead of using multiple physical cameras to track gaze, the system uses a single camera feed combined with a virtual representation of the store, where raycasting techniques determine gaze location by tracing virtual rays from the customer's head position and orientation to intersect with virtual product structures. This copying approach eliminates the need for multiple expensive cameras while maintaining measurement precision.
Solution Approach 2:
The patent introduces a virtual representation of the store environment as an intermediary between the single camera and the gaze tracking function. The virtual structures serve as a mediator that translates the simple camera feed into detailed gaze location information by raycasting through the virtual model. This intermediary layer allows accurate gaze tracking without requiring complex multi-camera hardware installations.
2Measurement precision
If detailed image analysis is performed to determine gaze location, then measurement precision is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-establishing the virtual representation of the store environment, including virtual structures, product locations, and shelving configurations, before actual gaze tracking begins. The virtual model is prepared in advance with all geometric relationships defined, so that during operation, the system only needs to perform efficient raycasting operations rather than analyzing complex images in real-time. This preliminary setup significantly reduces processing time while maintaining precision.
Solution Approach 2:
The patent replaces the mechanical/image-based analysis system with a computational geometry approach using raycasting in a virtual environment. Instead of analyzing actual images to determine where customers are looking, the system substitutes image analysis with mathematical ray-tracing through a pre-built virtual model, dramatically reducing processing time while preserving measurement precision.
3Loss of information
If camera systems are installed to monitor customer interactions, then information quality is improved, but privacy concerns increase
Solution Approach 1:
The patent extracts only the essential information needed for gaze tracking - specifically head position and orientation - from the camera feed, while deliberately excluding detailed image data that would compromise privacy. The system processes minimal geometric information to determine gaze location and product interaction, leaving out facial features, personal identifiers, and other sensitive visual data. This extraction approach maintains information quality for retail analytics while reducing privacy intrusion.
4Measurement precision
If multiple cameras are deployed to cover store areas, then measurement precision is improved, but ease of manufacture and deployment deteriorate
Solution Approach 1:
The patent makes a single camera system universal by combining it with a comprehensive virtual representation of the entire store environment. The single camera performs multiple functions: capturing customer head position, determining orientation, and enabling gaze tracking across the full store area through the virtual model. This multi-functionality eliminates the need for multiple specialized cameras, greatly simplifying hardware installation and deployment while maintaining full coverage and measurement precision.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to a computer implemented method for determining a gaze location of a human interacting with products in a store environment. The method comprising: obtaining, from a camera monitoring at least a portion of the store environment, one or more images of human interaction with products in the store environment, the camera having a known position and orientation within a coordinate system of the store environment; determining a position and orientation of a head of the human in the coordinate system of the store environment by analyzing the one or more images; and determining the gaze location of the human as an intersection between a raycast from a position and orientation of the head of the human in a virtual representation of the store environment corresponding to the position and orientation of the head of the human in the coordinate system of the store environment and a virtual structure in the virtual representation of the store environment.