生成对抗网络驱动下的北京蒙镶非遗图案智能生成研究  

RESEARCH ON INTELLIGENT GENERATION OF BEIJING MENGXIANG ICH PATTERNS DRIVEN BY GENERATIVE ADVERSARIAL NETWORKS

作　　者：朱荔丽 周威 陈妮 庞涛[2] ZHU LILI;ZHOU WEI;CHEN NI

机构地区：[1]北方工业大学机械与材料工程学院 [2]华北电力大学人文与社会科学学院

出　　处：《设计》2025年第6期138-142,共5页Design

基　　金：北京市社会科学基金一般项目“北京蒙镶技艺资料补遗与价值再生研究”(21YTB016)。

摘　　要：旨在通过生成对抗网络(GANs)实现北京蒙镶非遗图案的智能生成,突破传统工艺效率瓶颈,推动非遗的数字化保护与创新性传承。从数据采集、模型训练到图案生成与后期处理,系统探索GAN在北京蒙镶非遗图案智能生成中的应用。验证了GAN技术在非遗保护与创新设计中的可行性和优势,表明StyleGAN2模型在生成效率和质量上更优,能够更好地满足非遗图案生成的需求。SR(Super-Resolution)图像重建技术进一步提升了生成图案的分辨率和细节质量,使得生成的图案更加符合需求,为后续AIGC的创作提供了有力的技术支持。This study aims to achieve intelligent generation of Beijing Mongxiang ICH patterns using Generative Adversarial Networks(GANs),thereby overcoming the efficiency bottlenecks inherent in traditional craftsmanship and advancing the digital preservation and innovative inheritance of ICH.A systematic exploration of GAN in the intelligent generation of Beijing Mongxiang ICH patterns was conducted,encompassing data collection,model training,pattern generation,and post-processing.The feasibility and advantages of GAN technology in ICH preservation and innovative design were validated.The StyleGAN2 model demonstrated superior generation efficiency and quality,better aligning with the requirements of ICH pattern generation.Super-Resolution(SR)image reconstruction technology further enhanced the resolution and detail quality of generated patterns,ensuring their compliance with practical needs and providing robust technical support for subsequent AI-generated content(AIGC)creation.

关 键 词：生成对抗网络(GAN) 北京蒙镶技艺 非遗保护 智能生成 SR图像重建 

分 类 号：TB47[一般工业技术—工业设计]