Face Recognition Accuracy in Monitoring Video via Dual Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face recognition methods in monitoring video data are prone to misjudgment and missed judgments due to variable camera placements, distances, lighting conditions, and angles, leading to reduced accuracy.
Innovation Solution
A method involving two different face comparison models and a K-neighbor algorithm, combined with a screening mechanism using Euclidean distance and time-based analysis, to enhance face recognition accuracy in monitoring video data. The method includes collecting and preprocessing face data, training models, extracting high-dimensional features, and performing weighted fusion and K-neighbor calculations to achieve accurate recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single face comparison model is used, then the device complexity is low, but the face recognition accuracy deteriorates under variable monitoring conditions
Solution Approach 1:
The patent divides the face recognition task into multiple specialized comparison models (first face comparison model and second face comparison model), where each model is trained on different subsets of the training set. This segmentation allows each model to specialize in certain aspects of face recognition, improving overall accuracy while distributing the computational complexity across multiple models rather than requiring one overly complex model.
Solution Approach 2:
The patent combines the results from multiple face comparison models through a screening mechanism and weighted fusion. The K-neighbor algorithm aggregates predictions from different models, and the final recognition result is obtained by fusing these predictions with appropriate weights. This merging approach leverages the strengths of each individual model to achieve higher overall accuracy.
2Adaptability or versatility
If multiple different monitoring cameras are deployed to cover various angles and positions, then the adaptability of the system improves, but the quantity of video data and processing complexity increases
Solution Approach 1:
The patent extracts only the essential information from video data - specifically, face images are extracted from video frames using face detection algorithms. Rather than processing entire video streams, the system extracts and processes only the relevant face regions, significantly reducing the data volume that needs to be handled while maintaining recognition accuracy across multiple cameras.
Solution Approach 2:
The patent performs preliminary processing of video data by detecting and extracting face images before conducting the actual recognition task. Face detection algorithms identify and extract face regions from video frames in advance, filtering out irrelevant data. This preliminary action reduces the burden on subsequent recognition models and enables efficient handling of data from multiple cameras.
3Reliability
If manual screening and tagging of face pictures is performed thoroughly, then the quality of verification data improves, but the loss of time in data preparation increases
Solution Approach 1:
The patent applies partial manual screening rather than exhaustive manual processing of all face images. The system uses automated face detection to identify candidate faces, and manual screening is applied selectively to verify and tag these candidates. This partial action approach achieves sufficient verification data quality without the excessive time investment required for complete manual processing of all possible face images.
Data Source
AI summary
A method for recognizing a face from monitoring video data is disclosed. Two neural networks are used to compare and score high-dimensional face features of a face, and a K-neighbor algorithm and a screening mechanism with a Euclidean distance as a threshold are combined for face comparison recognition to obtain an accurate face recognition result. In addition, the present disclosure also performs further screening based on the time of video data and the frequency of face appearance, and finally obtains a recognition result, thereby obtaining a more accurate face recognition result. The present disclosure can perform relatively accurate face recognition on video data or picture data captured by a real-time or historical monitoring camera.