基于SIFT的甘肃庆阳刺绣针法检测增强算法研究  

Research on Enhancement Algorithm of Gansu Qingyang Embroidery Needle Detection Based on SIFT

作　　者：田欢[1] TIAN Huan(Department of Modern Service,Lanzhou Vocational Technical College,Lanzhou Gansu 730070,China)

机构地区：[1]兰州职业技术学院现代服务系,甘肃兰州730070

出　　处：《甘肃科技纵横》2024年第4期42-47,共6页Scientific & Technical Information of Gansu

基　　金：2022年兰州市科技计划项目“基于计算机视觉的兰州市文物资源三维重建方法与数字化应用研究”(2022-2-77);2023年甘肃省自然科学基金项目“甘肃非遗刺绣数字化保护技术研究”(23JRRA1471);2022—2023年兰州职业技术学院院级科研项目“数字化新农村背景下计算机专业人才培养研究”(2022XY-19)。

摘　　要：甘肃庆阳刺绣源于黄河流域,制品式样繁复种类多样,是陇东地区特色文化的重要载体。对庆阳刺绣的数字化保护可以作为中华文化传承和活化保护的有力手段。由于刺绣图案和针法具有较为复杂的边缘结构和重复率较高的纹理特征,因此对其进行区域检测是一项颇具挑战性的任务。文章选取5种庆阳刺绣中经典针法并建立用于学习训练的图像数据集,通过导向滤波搭建改进检测网络模型SIFT-E,对针法图案进行实时检测。与SIFT相比,改进后的方法识别匹配率更高,极大地提升了检测效率。Qingyang embroidery in Gansu Province,originating from the Yellow River Basin,boasts intricate de-signs and diverse varieties,serving as an important carrier of characteristic culture in the eastern part of Gansu Prov-ince.The digital protection of Qingyang embroidery can be used as a powerful means for the inheritance and activa-tion of Chinese culture.However,due to embroidery patterns and stitches have complex edge structures and texture features with high repetition rates,region detection is a challenging task.In this paper,five classical stitches from Qingyang embroidery are selected and an image dataset for learning and training is established.An improved detec-tion network model SIFT-E was constructed through guided filtering to perform real-time detection of stitch pat-terns.Compared with SIFT,the improved method achieved a higher recognition matching rate,greatly enhancing the detection efficiency.

关 键 词：庆阳刺绣 卷积神经网络 导向滤波 SIFT 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]