Occupant Observation Device Adaptive Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing occupant observation devices face reduced accuracy in condition estimation due to a uniform proportion of eye and mouth conditions, which can lead to inaccuracies in detecting occupant states such as drowsiness, especially for individuals with small eyes or those with challenging eye features.
Innovation Solution
The device incorporates an imager, eye detector, mouth detector, and condition estimator that dynamically adjust the weighting of eye and mouth detection results based on eye opening rates and specific events, such as prolonged low eye opening, to refine condition estimation accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a uniform proportion is used to reflect eye and mouth detection results in occupant condition estimation, then the device complexity is reduced and ease of operation is improved, but the measurement precision and reliability of condition estimation deteriorate
Solution Approach 1:
The patent applies dynamics by making the weighting proportions adaptive rather than fixed. The condition estimator dynamically adjusts the weighting of eye detector results versus mouth detector results based on detection quality metrics. When eye detection quality is high, the system weights eye results more heavily; when eye detection quality is low (e.g., small eyes, challenging features), it automatically increases the weight of mouth detection results, thereby maintaining high measurement precision without requiring complex manual configuration
Solution Approach 2:
The system implements feedback by using the detection results themselves to determine how those same results should be weighted. The condition estimator evaluates the quality of eye detection (through the eye detector's analysis of eye features) and uses this feedback to adjust the proportion of eye versus mouth detection results in the final condition estimation. This self-regulating feedback mechanism improves measurement precision while keeping the system relatively simple
2Reliability
If eye detection results are always weighted equally or uniformly, then the processing simplicity is maintained, but the reliability of condition estimation deteriorates for occupants with small eyes or challenging eye features
Solution Approach 1:
The patent applies parameter changes by dynamically modifying the weighting parameters based on detection conditions. Instead of using fixed weights, the system changes the weighting parameters adaptively - adjusting the proportion of eye detector results versus mouth detector results based on the actual quality of eye detection for each occupant. This allows reliable condition estimation for diverse occupants (including those with small eyes) while keeping the weighting adjustment mechanism integrated into the existing detection framework
3Measurement precision
If the system uses only eye detection results, then the device complexity is minimized, but the measurement precision deteriorates for certain occupant types
Solution Approach 1:
The patent applies local quality by allowing different detection methods to contribute differently based on local conditions. Rather than uniformly relying on one detector type, the system evaluates the local quality of each detection source (eye detector versus mouth detector) and weights them accordingly. For occupants with good eye detection quality, eye results dominate; for those with poor eye detection quality, mouth results gain local importance. This localized adaptation improves overall measurement precision while managing multi-detector coordination through automated quality assessment
Data Source
AI summary
An occupant observation device includes an imager configured to capture an image of a head of an occupant of a vehicle; an eye detector configured to detect at least a part of eyes of the occupant in an image captured by the imager; a mouth detector configured to detect at least a part of the mouth of the occupant in the image captured by the imager; and a condition estimator configured to estimate a condition of the occupant on the basis of a detection result of the eye detector and a detection result of the mouth detector, in which the condition estimator changes a ratio of reflecting each of the detection result of the eye detector and the detection result of the mouth detector in an estimation of the condition of the occupant, on the basis of the detection result of the eye detector or a result of a process performed on the basis of the detection result of the eye detector.