Hybrid Iris Tracking Using Cascaded Regression and Circle Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for eye center localization in facial images face challenges in robustness, especially in low-resolution or noisy conditions, and require annotated training data, which can be cumbersome to obtain.
Innovation Solution
A hybrid approach combining cascaded regression and hand-crafted model fitting for eye center localization, using a cascade regression forest engine with Histogram of Oriented Gradients features and a circle fitting post-processing step, which allows for training without manually annotated data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If hand-crafted model fitting methods are used for eye center localization, then measurement precision is improved, but reliability deteriorates in challenging settings such as low resolution or noisy images and poor illumination
Solution Approach 1:
The patent combines hand-crafted model fitting methods with machine learning-based cascaded regression methods into a hybrid approach. The hand-crafted methods provide accurate circle fitting for eye center localization, while the machine learning component provides robust feature detection and initialization that works reliably in challenging conditions. This merging allows the system to achieve both high precision and reliability across diverse imaging conditions.
2Reliability
If machine learning based methods are used for eye center localization, then reliability is improved, but measurement precision deteriorates compared to model fitting approaches
Solution Approach 1:
The patent segments the eye center localization process into distinct stages: (1) machine learning-based cascaded regression for robust initialization and feature detection, (2) hand-crafted circle fitting for precise measurement. Each stage performs its specialized function optimally, with the machine learning component handling the challenging aspects of robust detection and the hand-crafted component handling the precise measurement, thereby achieving both reliability and precision.
3Reliability
If machine learning based methods are used for eye center localization, then reliability is improved, but device complexity increases due to requirement of annotated training data
Solution Approach 1:
The patent implements a self-service mechanism where the system uses detected facial features and geometric relationships to automatically generate training data annotations. The cascaded regression detector identifies facial landmarks, and circle fitting algorithms automatically annotate eye center positions, eliminating the need for manual annotation. This allows the machine learning component to be trained without external annotation resources, reducing device complexity while maintaining reliability.
Data Source
AI summary
This document relates to hybrid eye center localization using machine learning, namely cascaded regression and hand-crafted model fitting to improve a computer. There are proposed systems and methods of eye center (iris) detection using a cascade regressor (cascade of regression forests) as well as systems and methods for training a cascaded regressor. For detection, the eyes are detected using a facial feature alignment method. The robustness of localization is improved by using both advanced features and powerful regression machinery. Localization is made more accurate by adding a robust circle fitting post-processing step. Finally, using a simple hand-crafted method for eye center localization, there is provided a method to train the cascaded regressor without the need for manually annotated training data. Evaluation of the approach shows that it achieves state-of-the-art performance.