纺织品纹理双面融合与共轭特征识别关键技术研究

Digital Reverse Engineering (RE) of textile products can improve the quality of product design, shorten the design cycle and enhance the quick response ability of textile companies. The image acquisition and analysis of textile texture is one of the key technologies for the reverse engineering of textile products, digital image processing and analysis is widely used to solve this problem. As the traditional fabric has the nature of conjugated structure, that is, one individual yarn tracing up and down through both sides of the fabric alternately, it endows the fabric texture restricted by dual side conjugated features, one single side image is not enough to reconstruct the complete fabric texture any more. This project focuses on the key fundamental issue of automatic identification of fabric texture, presents a new method to implement the said dual side compound scanning and conjugated texture analysis, sets up an integrated imaging system based on the registration and matching of marking points on the single object with double surfaces, digitizes both upper and bottom surface image of the fabric texture, establishes a conjugated texture model for the fabric texture identification based on weave pattern coding and colour clustering techniques. This research presents a new solution for the automatic, quick and accurate identification of textile texture. The fundamental research on the yarn location based on edge detection, the modelling of fabric texture based on weave pattern and colour justification, both of them can provide a theoretical and technological support for the quick sorting and identification of textile texture.

纺织品数字化逆向工程，可有效提高纺织企业的产品设计质量，缩短设计周期，降低设计成本，增强企业对市场的快速响应能力。纺织品纹理信息的采集与分析是纺织品数字化逆向工程的核心技术之一，目前普遍采用数字图像处理和识别的手段。由于传统织物具有双面共轭结构，即单根纱线交替沉浮于织物正反两个表面，使得织物纹样完备信息受到双面纹理特征约束，单一表面的织物纹理图像难以重现织物纹样建构。本课题针对自动识别织物纹样的关键基础问题，提出双面图像融合与共轭纹理特征识别的新方法，通过建立单目标空间定位约束点匹配复合成像系统，获得织物纹样正反定位匹配的图像对，建立纹样共轭纹理模型，并结合组织编码和颜色聚类分析技术识别织物纹样类型，为实现纺织品纹样的自动准确快速识别提供新的研究思路；研究织物图像分析中识别定位纱线边缘，基于纹理与颜色判别纹样类型的基本图像理解与识别问题，有望为纺织品纹样快速智能识别提供理论依据和技术基础。

纺织品纹理信息的采集与分析是纺织品数字化逆向工程的核心技术之一，可有效提高纺织企业的产品设计质量， 缩短设计周期，降低设计成本，增强企业对市场的快速响应能力。本课题针对自动识别织物纹样的关键基础问题，提出双面图像融合与共轭纹理特征识别的新方法，通过建立单目标空间定位约束点匹配复合成像系统，获得织物纹样正反定位匹配的图像对，建立纹样共轭纹理模型，并结合组织编码和颜色聚类分析技术识别织物纹样类型，实现纺织品纹样的自动准确快速识别；通过课题的研究，完成了纺织品纹理和颜色的双面数字化建模、双面数字化扫描、双面数字化识别，达到表征织物结构和测量织物颜色的目的，开发了适用于纺织品外观分析的数字化设备，获得了完备的织物组织结构和颜色的原始信息群，实现了对织物特别是色织物的纱线交织状态和颜色排列的准确识别。本课题的顺利完成，对于研究织物图像分析中判别纹样类型等基本图像理解与识别问题，具有一定的科学研究价值，有望为纺织品纹样快速智能识别提供理论依据和技术基础。