Driver Gaze Heat Maps for Accurate Alertness Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately estimate a driver's alertness level based on their line of sight, particularly in varying environments, leading to potential safety issues during mobile object operation.
Innovation Solution
An electronic device equipped with an image-capturing unit, line-of-sight detector, and controller uses machine learning to generate heat maps representing the probability of a line of sight being at each position in an image, allowing for precise alertness level estimation by associating images with detected line of sights, and excluding data points where line of sight movement exceeds a threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning is used to estimate alertness level based on line of sight, then alertness level estimation accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the alertness estimation task into multiple components: an image capturing unit captures the driver's view, a line-of-sight detector tracks gaze position, and a controller with machine learning models processes the data. By dividing the complex estimation task into separate functional modules, the system achieves high accuracy while managing device complexity through modular architecture.
Solution Approach 2:
The patent applies preliminary action by pre-training machine learning models using learning data collected during a calibration phase. The controller learns the relationship between line-of-sight patterns and alertness levels before actual operation. This pre-learning process enables accurate real-time estimation without requiring complex runtime computations, thus improving accuracy while controlling complexity.
2Measurement precision
If learning data is collected and processed to create heat maps, then alertness level estimation accuracy is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary action by collecting learning data and training the machine learning model in advance during a calibration period. The controller stores the learned relationships as heat maps that represent probability distributions of line-of-sight positions for different alertness levels. During actual operation, the system only needs to query these pre-computed heat maps rather than performing complex real-time analysis, significantly reducing processing time while maintaining high accuracy.
3Measurement precision
If the system monitors line of sight continuously, then alertness level estimation accuracy is improved, but use of energy increases
Solution Approach 1:
The patent implements periodic action by capturing images and detecting line-of-sight positions at specific time intervals rather than continuously. The controller acquires images at predetermined frames per second and updates alertness estimation periodically. This periodic monitoring approach maintains sufficient accuracy for safety applications while significantly reducing energy consumption compared to continuous monitoring.
Data Source
AI summary
An electronic device 10 includes an image-capturing unit 11, a line-of-sight detector 12, and a controller 14. The image-capturing unit 11 generates an image corresponding to the view by performing image capturing. The line-of-sight detector detects a line of sight of a subject with respect to the view. The controller 14 functions as a first estimator 15. The first estimator 15 is capable of estimating a first heat map based on the image. The controller 14 calculates the alertness level of the subject based on the first heat map and the line of sight of the subject. The first estimator 15 is constructing using learning data obtained by machine learning the relationship between a learning image and a line of sight of a training subject when an alertness level of the training subject is in a first range.