【作者】 张丹；

【导师】 蒋高明；

【作者基本信息】 江南大学 ， 纺织工程， 2011， 硕士

【摘要】 贾卡经编针织物是由带贾卡装置的经编机器生产的具有独特花纹效应的织物,其花纹图案精致多变,花型效应层次丰富,手感柔软,备受消费者青睐,广泛用于室内装饰、高档妇女内衣面料及花边辅料。但织物花纹图案的工艺设计需要设计师进行描图勾边,耗时费力,根本不能满足产品日益更新的市场要求。本课题针对这一现状,从实际应用角度出发,首先介绍了纹理图像分割的基础知识,分析了贾卡经编针织物花纹图案特有的纹理特点,其次针对三种不同纹理类别的织物,在Visual C++.NET编程平台上,对多种纹理图像分割算法进行了尝试,最终提出了三种针对少梳贾卡经编针织物花纹图案的自动识别方法。在对识别方法的详细介绍中,文章首先介绍了贾卡经编针织物图像获取和预处理流程,然后阐述如何运用三种纹理图像分割算法对针织物花纹图案进行自动识别。第一种是迭代阈值分割方法,针对纹理细致的贾卡经编针织物,中值滤波后图像经过阈值分割、边缘光滑,便可得到精确的花纹图;第二种是基于纹理能量分析的方法,针对带两种纹理、底网纹理粗糙的贾卡经编针织物,织物图像与Laws纹理模板进行卷积,再结合迭代阈值分割方法,便生成清晰的识别图像。第三种是基于小波变换的方法,针对包含三种及以上纹理区域、纹理特征粗细多变的贾卡经编针织物,利用Mallat金字塔小波变换尺度的不同,提取贾卡经编针织物的平均能量和方差作为纹理特征,结合K均值聚类识别织物花纹图案。接下来对识别算法进行了试验和评价,识别结果证明,本文提出的三种算法均可以精确、快速地获取相应纹理类别织物的花纹图案。最后将所提出算法应用于经编针织物CAD系统中,经验证,可以有效地减少设计人员的工作量,加快贾卡经编针织物花纹图案意匠图的设计速度,而又不失其正确性。本文所提出算法亦可适用于其他相似纹理图像的识别处理中,具有一定的推广价值。本文的最后,对纹理图像分割技术的发展方向也做了一定展望。更多还原

【Abstract】 Jacquard warp knitted fabric which has a unique fabric pattern is produced by warp knitting machine with the jacquard device. For its delicate and varied patterns, soft feel, it’s to be loved by many consumers, and is widely used as interior decoration, high-end lingerie fabrics and lace accessories. However, the process for designing this fabric patterns needs tracing by designers which is time-consuming, it can not meet the product requirements of the growing market updates.The topic for this situation, from a practical point of view, first mastered the basics of texture image segmentation, analyzed the unique texture characteristics of Jacquard warp knitted fabric patterns, followed by trying a wide range of texture image segmentation algorithms in the Visual C++.NET programming platform, against the texture for three different types of fabric, finally put forward three methods for automatic recognition of jacquard warp-knitted fabric patterns.In a detailed description of the identification method, the paper introduced the Jacquard warp-knitted fabric image acquisition and pre-process, then describes how to use the three texture image segmentation algorithm knitted for automatic identification. The first method is the iterative threshold segmentation method, for Jacquard warp knitted fabric with detailed textures, the filtered image then was treated by threshold segmentation, edge smooth, can be changed to an accurate pattern image; the second is based on the texture energy analysis, contrary to the more rough textured jacquard warp-knitted fabric with two texture regions, after convolution of fabric image and the Laws texture template, combined with the iterative threshold segmentation method, we generate a clear recognition image. The third is based on the wavelet transform, for fabric with three or more texture regions, texture changes from rough to meticulous. Using Mallat pyramid different scale wavelet transform, extraction the average energy and variance of Jacquard warp-knitted fabric as texture features, combined with K means clustering to identify the fabric patterns.Then tested and evaluated these methods, recognition results show that the proposed three algorithms can accurately and quickly obtain the fabric texture patterns of the appropriate type. Finally, the proposed algorithm is applied to warp knitted fabric CAD system, proven to be effective in reducing the workload of designers, and to accelerate design speed of Jacquard warp knitted fabric drafted patterns, but not lost its validity.The proposed algorithm is also applicable to other similar texture image recognition processing, it has some promotional value. The final paper also shows some prospects of the development of texture image segmentation technology.更多还原