Facial Recognition System Using Time-Based Feature Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing facial recognition systems require complex determination processing to identify customers and often incorrectly exclude customers who do not follow a predetermined action pattern, leading to repetitive counting of individuals.
Innovation Solution
A facial recognition system connected to a camera device, featuring a feature extractor, preservation unit, and statistical processing unit, which extracts and preserves facial features from video images, performing statistical analysis only when similarity exceeds a predetermined value over a set time, preventing repetitive counting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complicated determination processing is performed to identify customers based on action patterns, then customer identification accuracy is improved, but system complexity increases
Solution Approach 1:
The patent extracts only the essential identifying feature (facial features) from the complex action pattern analysis, eliminating the need for complicated determination processing while maintaining identification accuracy. The facial feature extractor isolates the key identifier from the entire video data, simplifying the system architecture.
Solution Approach 2:
The patent replaces the mechanical observation and analysis of human actions with an automated facial recognition system that uses algorithmic comparison of facial features. This substitution eliminates complex determination logic and provides automated, consistent identification.
2Speed
If facial recognition is performed without time-based filtering, then identification speed is improved, but repetitive counting of the same person occurs
Solution Approach 1:
The patent performs preliminary actions by extracting and storing facial features with their corresponding timestamps before the counting process. This preliminary organization of data with time information enables rapid retrieval and comparison while automatically preventing repetitive counting through time-based filtering.
Solution Approach 2:
The patent maintains continuous operation by processing video frames in real-time without interruption, extracting facial features continuously while applying time-based filtering. This ensures both high identification speed and accurate counting by continuously comparing current detections with historical data within the time window.
3Measurement precision
If action pattern analysis is used to exclude non-customers, then customer filtering accuracy is improved, but processing complexity increases
Solution Approach 1:
The patent extracts only the essential identifying feature (facial features) from the complex action pattern analysis, eliminating the need for complicated determination processing while maintaining identification accuracy. The facial feature extractor isolates the key identifier from the entire video data, simplifying the system architecture.
Solution Approach 2:
The patent changes the identification parameter from complex action patterns (multiple variables) to simple facial features (single unified parameter). This parameter transformation simplifies the processing logic while maintaining or improving filtering accuracy through reliable facial recognition.
Data Source
AI summary
A statistical processing is performed without performing complicated determination, so as not to repetitively include the same person. A facial feature extractor (32) in a facial recognition server (30) extracts features of facial image data including a face which is shown in an image obtained by a camera device (10) imaging an imaging area. A preservation unit (35) preserves features of facial image data including a face which has passed through the imaging area, in a pre-passing comparative original facial feature data memory (41). A statistics unit (37) performs statistical processing on the features of the facial image data, which are extracted by the facial feature extractor (32), in a case where feature having high similarity which is obtained by comparison with the features (extracted by the facial feature extractor (32)) of the facial image data imaged by the camera device (10) and is equal to or greater than a predetermined value are not preserved in the pre-passing comparative original facial feature data memory (41) until a predetermined time from an imaging time point by the camera device (10).