面向多源数据集的人脸老化研究

Research on Face Aging Based on Multi-Source Datasets

摘要: 针对公安跨年龄人脸识别中图像老化失真、模型泛化差、身份特征易丢失的问题，研究基于StyleGAN2与开源代码，整合三大公开数据集，通过标准化预处理优化数据质量，改进生成器与判别器，加入身份与性别约束，构建复合损失函数提升模型性能。实验表明，所提方法在受控数据集上生成图像保真度较高，为公安跨年龄人脸识别实战提供了有益探索与实验支撑。

Abstract: Aiming at the problems of image aging distortion, poor model generalization, and easy loss of identity features in cross-age face recognition for public security practice, this study integrates three public datasets based on StyleGAN2 and open-source codes, optimizes data quality through standardized preprocessing, improves the generator and discriminator, adds identity and gender constraints, and constructs a composite loss function to enhance the model performance. Experiments show that the proposed method has high fidelity of generated images on controlled datasets, providing useful exploration and experimental support for public security cross-age face recognition practice.