Blink Detection via Aligned Eye Image Difference Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image recognition technologies face challenges in accurately and efficiently detecting blinks from video recordings, which is crucial for determining the state of a biological individual and enhancing security measures.
Innovation Solution
A method and apparatus that extract and align eye images from frames of video, generate a difference image, and utilize a pre-trained blink detection model to identify blink events by analyzing pixel differences, improving the speed and accuracy of blink detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional image recognition technologies are used to detect blinks, then the detection can be performed, but the speed and accuracy of blink detection are insufficient
Solution Approach 1:
The patent segments the face image into multiple eye images (first eye image and second eye image) and processes them separately. By dividing the complex task of blink detection into smaller sub-tasks of extracting and comparing individual eye images, the system achieves both high accuracy and efficient processing speed.
Solution Approach 2:
The patent extracts only the necessary eye region information from the complete face image, discarding irrelevant facial features. This extraction of critical eye image data enables the system to focus computational resources on the most informative elements, thereby improving detection speed without sacrificing accuracy.
2Measurement precision
If eye images are extracted and aligned from video frames, then blink detection accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining marking points on the eye images and pre-establishing the alignment methodology. This preparation before actual blink detection allows the system to quickly compare and align eye images without time-consuming computations during the detection process itself.
Solution Approach 2:
The patent creates a simplified representation of the eye images by extracting only the relevant regions and aligning them based on marking points. This copying of essential features rather than processing the entire original images reduces processing time while maintaining detection accuracy.
3Productivity
If difference images are generated and processed through blink detection models, then detection speed increases, but the complexity of the system increases
Solution Approach 1:
The patent introduces difference images as an intermediary representation that bridges the raw eye images and the final blink detection result. By transforming the complex task of blink detection into a simpler difference image comparison problem, the system achieves faster processing while the complexity is managed through this intermediate step rather than requiring complex direct analysis.
Data Source
AI summary
A method and apparatus for detecting a blink. An embodiment includes: extracting two frames of face images from a video recording a face; extracting a first to-be-processed eye image and a second to-be-processed eye image respectively from the two frames of face images, and aligning the first to-be-processed eye image with the second to-be-processed eye image through a set marking point, the marking point being used to mark a set position of an eye image; acquiring a difference image between the aligned first to-be-processed eye image and the second to-be-processed eye image, the difference image being used to represent a pixel difference between the first to-be-processed eye image and the second to-be-processed eye image; and importing the difference image into a pre-trained blink detection model to obtain a blink detection label, the blink detection model being used to match the blink detection label corresponding to the difference image.