随着人工智能技术的发展与应用，智能化设计受到了越来越多的关注。本研究以藏毯图案设计为对象，提出了基于美学的藏毯图案生成框架及智能设计系统，旨在通过智能技术优化传统人工设计流程中大量耗时耗力的环节，提升生产过程中设计迭代的效率。为实现此目标，提出了由用户输入层、元素生成层、个性修正层和交互应用层组成的层次化智能设计规则，以指导智能系统各功能模块的设计与开发。 针对用户输入层，系统地梳理了藏毯设计中的需求描述类型，提炼出文本描述和灵感图像两种典型输入方式，通过开展用户实验总结了不同用户输入类型的使用特点，提出了使用增强现实辅助用户进行灵感挖掘的绘图工具。针对元素生成层，归纳了藏毯主要风格的特征和对应元素，进一步构建了美学藏毯数据集进行生成训练，从而实现了生成与需求相对应的图案元素，同时基于经典图案构成理论，设计了多种图案拼接方法以满足主流藏毯机织设备的不同尺寸要求。针对个性修正层，运用心理学和美学相关原理，建立了使用情感词汇对图案色彩组合进行变换的方法，为用户提供了藏毯个性化修正的解决方案。针对交互应用层，基于网页对智能系统进行了设计与搭建，同时探索了藏毯图案的预览方式。分别对生成图案满意度和智能设计系统可用性进行了用户实验评估，结果表明该系统易于学习，能够有效生成美观且符合设计需求的藏毯图案，显著减少了人工设计所需的时间成本。本工作为人工智能应用设计领域提供了一种新的发展思路，同时促进了藏毯艺术的保护和发展。 本研究的主要贡献包括： （1）提出了层次化的藏毯图案智能设计规则，通过用户研究对比了图像生成的多种输入方式，设计并开发了增强现实灵感绘图工具，丰富了输入形式的可能性。 （2）构建了场景化美学藏毯图案生成框架，通过文本描述和灵感图像搭配生成对应艺术风格的藏毯图案，设计了快速拼接功能以生成不同尺寸的藏毯，同时提出了情感化的色彩组合修正方法，以满足用户的个性化需求。 （3）搭建了图形化交互的藏毯图案生成系统，并通过用户实验对生成图案满意度及系统可用性进行评估。该系统有效提升了藏毯设计流程的效率，探索了藏毯智能设计的解决方案。

With the development and application of artificial intelligence technology, intelligent design has gained increasing attention. This study focuses on the design of Tibetan carpet patterns, proposing an aesthetically-based Tibetan carpet pattern generation framework and intelligent design system, aiming to optimize the time-consuming and labor-intensive aspects of traditional manual design processes and improve the efficiency of communication and design iterations during production. To achieve this goal, a hierarchical intelligent design rules comprising user input layer, element generation layer, personalized modification layer, and interactive application layer is proposed to guide the design and development of various functional modules of the intelligent system. For the user input layer, based on the main demand description types in Tibetan carpet design, two typical input methods, textual description and inspiration image, are sorted out. User experiments are conducted to refine the usage characteristics of different inspiration types and propose a drawing tool that uses augmented reality to assist designers in mining inspiration. For the element generation layer, the characteristics and elements of the main styles of Tibetan carpets are summarized, and an aesthetic Tibetan carpet dataset is further constructed for generative training to generate pattern elements corresponding to the demand. Meanwhile, based on the classical pattern composition theory, various rapid tiling methods are designed to meet the different size requirements of mainstream weaving equipments. For the personalized modification layer, psychological and aesthetic principles are applied to establish a method for modifying the color combination of Tibetan carpets using emotional vocabulary, providing users with a personalized adjustment solution. For the interactive application layer, An intelligent system is designed and built based on web pages, and the possibilities of previewing the generated Tibetan carpets are explored. User experiments are conducted to evaluate the satisfaction of the generated patterns and the usability of the intelligent design system. The results show that the system is easy to learn, effectively generates beautiful and demand-compliant Tibetan carpet patterns, significantly reducing the time and effort required for manual design. This research provides a new development idea for the application of artificial intelligence in the field of design and promotes the protection and development of Tibetan carpet art. The main contributions of this study include: （1） Proposing a hierarchical Tibetan carpet pattern intelligent design rules, comparing various image generation input methods through user research, and designing and developing an augmented reality inspiration drawing tool to enrich the possibilities of input content. （2） Constructing a scenario-based aesthetic Tibetan carpet pattern generation framework, generating corresponding artistic style Tibetan carpet patterns through the combination of textual description and inspiration image, designing a rapid tiling function to generate Tibetan carpets of different sizes, and proposing an emotion-based color combination modification method to meet users‘ personalized needs. （3） Building a graphical interactive system for Tibetan carpet pattern generation and conducting user experiments to evaluate image satisfaction and system usability, effectively improving the efficiency of the Tibetan carpet design process and exploring solutions for intelligent Tibetan carpet design.