Face Recognition Device Using Segmented Detection and HOG Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face recognition systems using deep learning technology face challenges in processing high-resolution images in real-time due to high computational load, limiting their ability to detect faces accurately and efficiently, especially with varying image quality, angles, and lighting conditions.
Innovation Solution
A face recognition method and device that combines deep learning-based face detection with Histogram of Oriented Gradients (HOG)-based face detection and a face tracker, using Intersection over Union (IOU) values and feature vectors to update and track face objects across image frames, allowing for real-time face detection and recognition without reducing accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning-based face detection is used, then face detection accuracy is improved, but execution time increases and real-time processing becomes difficult
Solution Approach 1:
The patent segments the face detection process into two distinct stages: a first face detection unit that performs initial detection, and a second face detection unit that performs verification. This segmentation allows the system to benefit from the high accuracy of deep learning while reducing overall execution time by distributing computational tasks across multiple specialized components.
2Measurement precision
If deep learning-based face detection is used, then face detection accuracy is improved, but device complexity increases
Solution Approach 1:
The patent divides the complex deep learning-based face detection system into two separate face detection units with distinct functions. The first unit performs initial detection while the second unit performs verification, thereby reducing the computational burden and complexity of each individual unit while maintaining overall high accuracy.
3Reliability
If face detection rate is maintained across varying image qualities, then recognition reliability is improved, but computational load increases
Solution Approach 1:
The patent employs a two-stage detection architecture where the first face detection unit handles initial detection and the second face detection unit performs verification. This segmentation enables the system to maintain consistent face detection rates across varying image qualities, angles, and lighting conditions while optimizing computational resource usage by distributing processing tasks.
Data Source
AI summary
The present invention relates to a face recognition method and a face recognition device, the method including: a step for detecting one or more faces in an image frame; a step for updating a face object in a face list by using the one or more pieces of detected face information; and a step for recognizing an identity corresponding to the face object by using a plurality of feature vectors of the face object in the updated face list.