Enhanced Identity Recognition Post-Cosmetic Surgery Using Convolutional Neural Networks and Extreme Learning Machine

Abstract

Facial recognition systems face significant challenges in accurately identifying individuals who have undergone cosmetic surgery, as these procedures often result in substantial alterations to facial features. This study introduces a novel approach combining Convolutional Neural Networks (CNN) for feature extraction and Extreme Learning Machine (ELM) for classification, specifically designed to enhance identity recognition after cosmetic surgeries. The method was evaluated using the IIITD Plastic Surgery Face Database, which includes images from both local and global surgeries. Preprocessing techniques, such as histogram equalization and noise reduction, were employed to improve image quality and ensure robust feature extraction. The experimental results demonstrate that the proposed method significantly outperforms traditional CNN-based approaches, achieving recognition rates above 95% across both local and global surgeries. Notably, ear surgery (otoplasty) achieved 98.84% accuracy, and eyelid surgery (blepharoplasty) reached 98.20% accuracy. The Extreme Learning Machine (ELM) component played a crucial role in reducing overfitting and improving generalization, making the system highly efficient for handling large datasets. These findings highlight the effectiveness of the proposed method in clinical, legal, and security applications, where accurate post-surgery identity recognition is critical. The ELM-based system offers a reliable and efficient solution for identity verification in complex post-surgical scenarios, demonstrating its potential for broader application in real-world settings.