This study, taking the Spring Festival folk culture of the Yellow River Delta as a case, initiates a behavior-symbol-semantic tri-level cultural gene analysis framework and constructs a human-machine collaborative digital preservation pathway by integrating generative adversarial networks (GANs). Through ethnographic fieldwork, the research comparatively analyzes the differentiated characteristics of New Year customs between canal and maritime cultural regions, revealing the shaping mechanisms of geographical environments and economic forms on cultural diversity, and establishes a hierarchical labeling system. To address the limitations of traditional GANs technology, a cGAN-ICH Inheritor Dual-Discriminator collaborative framework is designed, which synergizes machine vision discrimination with expert knowledge validation to ensure both visual fidelity and semantic authenticity for generation. By proposing a “digital technology-cultural ecology bidirectional empowerment” pathway, this study provides methodological insights for advancing the digital preservation and transmission of intangible cultural heritage.