基于深度卷积网络和局部主动轮廓模型的贵州苗族服饰图腾图案自动分割方法研究

贵州苗族服饰图腾图案自动分割对苗族服饰文化数字化抢救、保护和传承具有重要作用，能为挖掘苗族服饰文化内涵和发现苗族服饰传承机理研究提供可靠依据。为解决传统人工勾画苗族服饰图腾图案存在耗时且受文化传承人经验和主观性影响等问题，本项目针对贵州苗族服饰图腾图案自动分割开展系统研究，拟提出基于深度卷积网络和局部主动轮廓模型的贵州苗族服饰图腾图案自动分割方法。研究内容包括：（1）基于FCN的贵州苗族服饰图腾图案区域定位模型；（2）基于FCN的贵州苗族服饰图腾图案区域提取模型；（3）基于局部主动轮廓的贵州苗族服饰图腾图案自动分割模型。本项目的研究目标是借鉴并利用多学科的理论和方法，为提高贵州苗族服饰图腾图案自动分割准确性和鲁棒性，构建基于深度卷积网络和局部主动轮廓模型的贵州苗族服饰图腾图案自动分割基本理论与方法，并探索其在实际中的应用，最终提供更为有效、更为便捷的贵州苗族服饰图腾图案自动分割解决方案。

Automatic segmentation of Guizhou Miao clothing totem patterns plays an important role in digital rescuing, protection and inheritance of Miao clothing culture, which can provide a reliable basis for these researches about excavating its cultural connotation and discovering its cultural inheritance mechanism. Currently, traditional manual delineation of Miao clothing totem pattern is time-consuming, and the results of delineation depend on cultural inheritors' experience and subjectivity. To deal with these problems, we propose an automatic segmentation method of Guizhou Miao clothing totem patterns based on deep convolution network and local active contour model by a comprehensive study. The research content includes: (1) The model for the Guizhou Miao clothing totem patterns region localization based on full convolution network is explored; (2) The model for Guizhou Miao clothing totem patterns region extraction based on full convolution network is explored; (3) The model for Guizhou Miao clothing totem patterns automatic segmentation based on local active contour is explored. Our final objective is to utilize theories from multiple research areas and improve the effectiveness and practicability of Guizhou Miao clothing totem patterns automatic segmentation with a better solution.