PCA/K-L transformation facial recognition method for vending systems

Lu Lin ¹, Yu Sang ¹, Erwei Li ²

⁰Faculty of Business and Technology, University of Cyberjaya, Cyberjaya, Selangor Darul Ehsan, Malaysia.

Plos One +

|

December 10, 2025

View abstract on PubMed

Summary

This study introduces an improved facial recognition model for automatic vending systems. The new method enhances accuracy and speed, offering robust performance even in noisy conditions.

Area Of Science

Computer Science
Biometrics
Artificial Intelligence

Background

Facial recognition technology is crucial for automatic vending systems, enabling fast checkout and personalized recommendations.
Existing facial recognition models face challenges in accuracy, processing speed, and robustness, particularly in varied conditions.

Purpose Of The Study

To develop an advanced facial recognition model for automatic vending systems.
To enhance both the accuracy and processing speed of facial recognition algorithms.
To improve the robustness of facial recognition against noise and varying angles.

Main Methods

A novel facial recognition model integrating an improved Principal Component Analysis (PCA) with the Hotelling transform.
Dimensionality reduction of facial features using sample partitioning and histograms within PCA.
Application of the Hotelling transform to processed, reduced-dimensional facial data for enhanced recognition.

Main Results

Achieved high recognition accuracies of 96.32% on renderMe-360 and 98.24% on VoxCeleb2 datasets.
Demonstrated an average accuracy of 94.388% for facial recognition from various angles.
Exhibited significant efficiency gains in feature construction and recognition times, with strong robustness in high noise environments.

Conclusions

The proposed model significantly improves facial recognition accuracy and processing speed for automatic vending systems.
The model's robustness to noise and varying angles offers new technical value for intelligent vending solutions.
This research contributes to the advancement of biometric technology in commercial applications.