非单一视角下复杂织物纹理拓扑结构建模与图像特征识别研究

The project aims to investigate the topological modeling of fabric texture image, multi directional image fusion, texture feature recognition from a non-single perspective, considering both texture regularity and irregularity. The topological information of fabric texture could be analyzed by means of image recognition, the general topological model is established for both regular and irregular textile patterns, and texture characteristic analysis could be conducted on the fabric image with the topological characteristics. The methodology used for non-single view imaging of fabric image recognition, image fusion of multi-directional image, grid generation of topological element, topological modeling, and topological texture element analysis is investigated systematically. In this project, the algorithms structure and software implementation of topology analysis of the textile texture image are developed for yarn location detection, the weaving type recognition, the meshing of the topology and the color representation of the yarn body. Based on theoretical analysis and experimental study, the final completion of a practical and robust texture topology modeling from a non-single perspective and multi-directional image fusion, texture feature analysis of complex fabric patterns is done to explore a new method of fabric texture image recognition, it could provide a basic application research platform for large-scale fabric pattern collection and analysis, fabric defect detection, ancient fabric inpianting and intelligent manufacturing.

本项目致力于非单一视角下复杂织物纹理图像的拓扑结构建模、多角度图像融合、纹理特征识别的研究，综合考虑织物纹理的规则性与随机性，通过图像识别的手段解析得到织物纹理的拓扑性信息，并建立适用于规则及不规则纺织品纹样的一般拓扑结构模型。在分析具有纹理拓扑特性的纺织品图像的纹理特性的基础上，研究可用于织物图像识别的非单一视角下的图像采集、多角度图像融合、拓扑点网格化、拓扑结构建模、拓扑性纹理基元分析的基本方法及其实现手段。研究纺织品纹理图像的拓扑解析过程中纱线定位检测、组织点交织类型识别、拓扑网格划分、纱线本体颜色表征的算法结构与软件实现。通过算法开发和实验研究，实现非单一视角下基于纹理拓扑建模、多角度图像融合、纹理特征解析的织物纹理结构自动识别的研究目标，探索提高织物纹理图像识别鲁棒性的新方法，为建立大规模织物纹样采集分析、织物疵点检测、古代织物图像修复、纺织品智能制造奠定较好的应用研究基础。

本项目致力于非单一视角下复杂织物纹理图像的拓扑结构建模、多角度图像融合、纹理特征识别的研究，综合考虑织物纹理的规则性与随机性，通过图像识别的手段解析得到织物纹理的拓扑性信息，并建立适用于规则及不规则纺织品纹样的一般拓扑结构模型。在分析具有纹理拓扑特性的纺织品图像的纹理特性的基础上，研究可用于织物图像识别的非单一视角下的图像采集、多角度图像融合、拓扑点网格化、拓扑结构建模、拓扑性纹理基元分析的基本方法及其实现手段。研究纺织品纹理图像的拓扑解析过程中纱线定位检测、组织点交织类型识别、拓扑网格划分、纱线本体颜色表征的算法结构与软件实现。通过算法开发和实验研究，实现非单一视角下基于纹理拓扑建模、多角度图像融合、纹理特征解析的织物纹理结构自动识别的研究目标，探索提高织物纹理图像识别鲁棒性的新方法，为建立大规模织物纹样采集分析、织物疵点检测、古代织物图像修复、纺织品智能制造奠定较好的应用研究基础。