Cross-spectral palmprint recognition algorithm based on improved visual attention network

The global economy drives biometric technology development, with China's increasing population mobility boosting identity authentication and security needs. Palmprints are emerging as a reliable, non-intrusive alternative to traditional methods, enhancing privacy and accessibility. Technological advancements have integrated biometrics into commerce and public sectors, improving security, convenience, and privacy. This study used the VAN and domain generalization methods for cross-spectrum palmprint recognition, utilizing datasets from The Hong Kong Polytechnic University. The new MMVANNet network simulates human visual system characteristics, automatically focusing on crucial regions and features, maintaining robust performance across various spectral conditions, poses, and lighting, thereby enhancing accuracy and reliability.