Scholarlink

<p>The preservation of traditional embroidery, particularly the intricate techniques of Miao craftsmanship, faces growing challenges due to the diminishing number of skilled artisans and the increasing demand for scalable reproduction. This study presents a novel vision-based artificial intelligence framework designed to capture, model, and replicate traditional embroidery skills with high fidelity. Leveraging high-resolution image processing, trajectory modeling, and demonstration-based learning, the proposed system integrates Dynamic Movement Primitives (DMPs) and Mixture Density Networks (MDNs) to encode and generalize the spatial-temporal patterns of expert demonstrations. The embroidery trajectories are extracted and refined through visual tracking and B-spline fitting, while robotic reproduction is achieved using a six-degree-of-freedom manipulator guided by optimized kinematic mapping. Experimental evaluations demonstrate the system’s capacity to reproduce complex stitching behaviors with high accuracy and consistency, maintaining both the artistic integrity and structural precision of the original craftsmanship. This framework offers a scalable and adaptable solution for the digital preservation and robotic inheritance of intangible cultural heritage.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Vision-Based AI framework for skill transfer and robotic replication of Miao embroidery techniques

  • 期刊论文

  • 发表日期：

    2025年6月20日

查看全文
  • Ling Chen,
  • Jing Chen,
  • Zhi Su,
  • Xiaotong He,
  • Chuanlin Zuo,
  • Xiangjun Li

摘要

The preservation of traditional embroidery, particularly the intricate techniques of Miao craftsmanship, faces growing challenges due to the diminishing number of skilled artisans and the increasing demand for scalable reproduction. This study presents a novel vision-based artificial intelligence framework designed to capture, model, and replicate traditional embroidery skills with high fidelity. Leveraging high-resolution image processing, trajectory modeling, and demonstration-based learning, the proposed system integrates Dynamic Movement Primitives (DMPs) and Mixture Density Networks (MDNs) to encode and generalize the spatial-temporal patterns of expert demonstrations. The embroidery trajectories are extracted and refined through visual tracking and B-spline fitting, while robotic reproduction is achieved using a six-degree-of-freedom manipulator guided by optimized kinematic mapping. Experimental evaluations demonstrate the system’s capacity to reproduce complex stitching behaviors with high accuracy and consistency, maintaining both the artistic integrity and structural precision of the original craftsmanship. This framework offers a scalable and adaptable solution for the digital preservation and robotic inheritance of intangible cultural heritage.