Facial Recognition via Landmark Measurement Reproducibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face recognition technologies face challenges with errors due to ambient illumination, facial position, aging, hairstyle, and facial wear, which are difficult to control, leading to reduced accuracy and increased information load, especially in non-standardized conditions like group photos or surveillance videos.
Innovation Solution
A method using cumulative standard deviations and standard errors to differentiate reproducible and non-reproducible measurements between facial anthropological landmarks, allowing for population-verified reproducibility in 2D or 3D photographs, enabling unique feature identification for improved recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If facial topological morphing methods are used for face recognition, then the system can process a wide range of facial images, but accuracy is compromised due to errors from ambient illumination, facial position, aging, hairstyle, and facial wear
Solution Approach 1:
The patent transforms the face recognition approach from topological morphing to linear and angular measurements between anthropological landmarks. This parameter change enables quantitative, objective measurements that are less sensitive to environmental factors like illumination and position, while also capturing subtle individual differences through precise metric data
Solution Approach 2:
The patent replaces the mechanical/topological morphing system with a statistical measurement system. Instead of comparing facial shapes through morphing algorithms, the system uses linear measurements (distances between landmarks) and angular measurements (angles formed by landmark connections), analyzed through statistical methods to identify reproducible versus unique features
2Measurement precision
If standardized positioning and illumination are applied to minimize errors, then measurement accuracy improves, but the method becomes impossible to apply in non-standardized conditions such as group photos, motion video, or surveillance camera
Solution Approach 1:
The patent changes from requiring controlled environmental parameters (standardized lighting, positioning) to using invariant anatomical parameters (landmark positions, inter-landmark distances, angles). These anatomical parameters remain consistent regardless of external conditions, enabling accurate measurements in surveillance, video, and group photo scenarios
Solution Approach 2:
The patent pre-identifies and catalogs anthropological landmarks and their measurement protocols before actual face recognition occurs. By establishing a framework of reproducible measurements across populations in advance, the system can quickly apply these pre-defined measurement patterns to any facial image without requiring standardized capture conditions
3Reliability
If linear and angular measurements between anthropological landmarks are used, then errors from illumination and positioning are overcome, but the knowledge of which measurements are reproducible versus unique within ethnic groups is lacking
Solution Approach 1:
The patent implements a feedback mechanism where measurements from multiple individuals within ethnic groups are collected and statistically analyzed. The system identifies which linear and angular measurements show high reproducibility (low variation) across the population versus those that are unique to individuals. This feedback loop builds a database of ethnic-specific measurement patterns that improves recognition accuracy
Solution Approach 2:
The patent performs preliminary statistical analysis on large datasets of facial measurements from various ethnic groups before deployment. This pre-characterization of population-specific reproducible measurements creates a reference framework that guides the recognition process, allowing the system to distinguish between common ethnic features and individual unique features
Data Source
AI summary
A method of facial recognition has been developed by the application of a statistical method, standard deviations or standard errors versus sample number plots, to differentiate the degree of reproducibilities of various measurements between facial anthropological landmarks in individual ethnic groups. Reproducible measurements between facial anthropological landmarks in a particular ethnic group mean they are common features shared by individuals of that ethnic group. Non-reproducible measurements are unique features of each individual in that ethnic group which may be used for individual facial recognition purposes. Such methodology may be computerized for automatic facial recognition. A large amount of data of each ethnic group is needed for facial recognition. In turn, the development of databases of each ethnic group will result in a large amount of data of human faces.