Line-of-Sight Detection Using Movement Vector Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing line-of-sight detection technologies face challenges in accurately detecting the line of sight of individuals wearing glasses due to reflections from glasses lenses, especially when the image-capturing interval is long or frame rate is low, leading to difficulties in identifying corneal reflections.
Innovation Solution
A line-of-sight detection device that captures infrared images and uses a processor to extract candidate corneal reflections by calculating movement vectors and threshold vectors based on reference points, distinguishing between corneal reflections and glass reflections to accurately determine the line of sight.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If infrared light is radiated to capture an image including the user's head for line-of-sight detection, then line-of-sight detection can be performed, but when the user is wearing glasses, infrared light reflects onto the lens creating false reflections that prevent accurate corneal reflection extraction
Solution Approach 1:
The patent segments the reflection detection process into multiple temporal snapshots (first image and second image captured at different times). By dividing the detection into sequential frames and analyzing movement between them, the system can distinguish stationary glasses reflections from moving corneal reflections, thereby extracting accurate corneal reflection locations despite the presence of glasses.
Solution Approach 2:
The patent performs preliminary action by capturing a first image before the second image, establishing a reference frame with candidate corneal reflections. This preliminary capture allows the system to predict expected corneal reflection positions in subsequent frames based on eye movement patterns, enabling it to identify and filter out false reflections from glasses lenses.
2Productivity
If the image-capturing interval is extended or frame rate is reduced to decrease processing load, then processing speed improves, but the ability to distinguish corneal reflections from glasses reflections deteriorates
Solution Approach 1:
The patent changes the parameter of image capturing interval to an extended duration, capturing images at lower frame rates. By adjusting this temporal parameter, the system reduces processing load while maintaining detection accuracy through the movement vector analysis that works effectively even with sparser temporal sampling.
Solution Approach 2:
The patent substitutes the mechanical approach of increasing frame rate with a computational approach using movement vector analysis. Instead of relying on higher temporal resolution to distinguish reflections, the system uses mathematical analysis of position changes between frames, replacing the need for frequent capturing with intelligent processing of fewer frames.
3Measurement precision
If multiple images are captured and processed to distinguish corneal reflections from glasses reflections, then detection accuracy improves, but processing time and computational load increase
Solution Approach 1:
The patent extracts only the essential information needed for line-of-sight detection by identifying and focusing on corneal reflection positions while filtering out glasses reflections. By extracting only the relevant candidate reflections and their movement characteristics, the system reduces processing complexity and time while maintaining high detection accuracy.
Solution Approach 2:
The patent performs partial action by processing only the necessary number of images (first and second images) rather than continuously processing all available frames. This selective processing of minimal required images reduces computational load and processing time while still achieving sufficient accuracy for line-of-sight detection.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate detection of the line of sight even when individuals wear glasses, reducing installation costs and improving detection speed by using a simple configuration, even with low frame rates or reduced image numbers.
Implementation Method 1
an infrared image is captured that includes the head of a user onto which infrared light is radiated
Implementation Method 2
detect a line of sight of the user on the basis of a location of a corneal reflection
Implementation Method 3
infrared light is reflected onto a lens of the glasses, and a reflection of infrared light that is different from a corneal reflection may be seen on a portion of the lens of the glasses
Data Source
AI summary
When a plurality of candidate corneal reflections are extracted from an image of a person, a line-of-sight detection device calculates a movement vector of a reflection location on the basis of the candidate corneal reflection extracted from a first image and a second image, calculates a movement vector of a reference point on the person on the basis of the reference point extracted from the first image and the second image, identifies a corneal reflection from among the plurality of candidate corneal reflections on the basis of the movement vector of the reflection location and the movement vector of the reference point so as to identify a location of the corneal reflection, and extracts a location of a pupil from the image, and calculates a line of sight of the person on the basis of the identified location of the corneal reflection and the extracted location of the pupil.