Machine Learning Iris Image Appropriateness Determination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing iris authentication systems require iris detection from photograph images, making it difficult to determine if an image is appropriate for authentication unless an iris is detected.
Innovation Solution
A determination apparatus using a machine learning model to assess whether iris information can be extracted from a photograph image, reducing the need for explicit iris detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If iris detection is performed from photograph image, then iris authentication can be executed, but it is difficult to determine image appropriateness without detecting iris
Solution Approach 1:
The system performs preliminary assessment of image quality using multiple detection models (eye detection, iris detection, face detection) before committing to full iris authentication processing. This allows the system to evaluate whether an image is suitable for authentication without completing the entire iris extraction pipeline, thereby determining image appropriateness efficiently.
2Measurement precision
If multiple detection models are used to assess image quality, then determination accuracy is improved, but processing load increases
Solution Approach 1:
The system dynamically adjusts the processing pipeline based on intermediate detection results. For example, if eye detection fails or the eye region is insufficient, the system terminates further processing early. If face detection succeeds but iris detection fails, the system can still determine image inadequacy without attempting full iris extraction. This dynamic adaptation maintains high assessment accuracy while optimizing processing efficiency by avoiding unnecessary computational steps.
Data Source
AI summary
A determination apparatus includes an acquisition unit that acquires a photograph image obtained by photographing a user, a storage unit that stores a first learning model constructed through machine learning using teacher data in which a plurality of images of eyes are associated with labels each indicating whether or not an iris can be extracted, and a first determination unit that inputs the photograph image into the first learning model and determines whether or not iris information can be extracted based on an output from the first learning model.