纺织学报 ›› 2018, Vol. 39 ›› Issue (05): 137-143.doi: 10.13457/j.fzxb.20170711007

• 管理与信息化 • 上一篇    下一篇

基于边缘和颜色特征的织物印花花型检索

  

  • 收稿日期:2017-07-31 修回日期:2018-01-23 出版日期:2018-05-15 发布日期:2018-05-10

Printed fabric pattern retrieval based on edge and color features

  • Received:2017-07-31 Revised:2018-01-23 Online:2018-05-15 Published:2018-05-10

摘要:

针对织物印花花型同型不同色和同形异构的特点,提出了一种综合花型边缘和颜色特征的图像检索算法。首先利用中值滤波对花型图像进行滤波去噪,并采用Canny算子获取花型边缘;以边缘像素坐标加权中心点为圆心,将各边缘点划分到若干等距环面内,提取花型边缘距离直方图,并在对边缘像素点直线拟合基础上,提取边缘方向直方图;然后在HSV空间内对彩色边缘进行颜色量化,提取各颜色分量构成颜色直方图;最后对3种特征分别进行相似性度量,并赋予三者不同权重予以结合,获得图像间相似度。结果表明,该算法具有平移、缩放、旋转不变性,并能兼顾边缘特征和颜色特征对检索结果的影响,准确区分同型不同色和同形异构的花型图案,该算法平均查准率可达93.17%。

关键词: 印花织物, 花型检索, 图像检索算法, 边缘特征, 颜色特征

Abstract:

The same shape filled with different solors or the same external shape comprising different intrenal shape is wery common in fabric printing. Aiming at precisely retrieval printed fabric patterns, a retrieval algorithm based on pattern edge and color features was proposed. Firstly, pattern edge image was obtained from the pretreated result with an image median filter process, and the mean value of all the edge points of the edge image from canny operate result was calculated and set as its center, then the edge points were divided into some equidistant ring regions from the center for extracting edge distance histogram as the image edge feature. Based on the line fitting results from the edge pixels, the edge direction histogram was extracted too. Secondly, the color edge was quantized in the HSV color space, and the edge color histogram was extracted as the image color feature. Finally, edge and color features were combined with different weights after normalization. Experiment results show that the algorithm has influence proposed of image translation, scaling and rotation, and the influence of edge and color features can clearly reflect in the retrieval results, so that patterns with the same shape but different colors, or patterns with the same external shape but different internal structures, can be accurately distinguished. The average precision of the algorithm is higher than 93.17%.

Key words: fabric printing, pattern retrieval, image retieval algorithm, edge feature, color features

[1] 向军 张杰 潘如如 韩要宾 张继东 高卫东 . 采用纹理平滑的印花织物图案轮廓提取[J]. 纺织学报, 2017, 38(11): 162-167.
[2] 张成梁 李蕾 董全成 葛荣雨. 应用区域颜色分割的机采棉杂质检测方法[J]. 纺织学报, 2017, 38(07): 135-141.
[3] 曹丽 胡旭东. 基于多特征融合的织物印花图像分割[J]. 纺织学报, 2016, 37(08): 149-153.
[4] 景军锋 杨盼盼 李鹏飞. 采用距离匹配函数的印花织物图案周期测定[J]. 纺织学报, 2015, 36(12): 98-0.
[5] 李敏 崔树芹 谢治平. 高斯混合模型在印花织物疵点检测中的应用[J]. 纺织学报, 2015, 36(08): 94-98.
[6] 张海波 黄铁军 修毅 赵野军 章江华. 基于神经网络的男西装图像情感语义识别[J]. 纺织学报, 2013, 34(12): 138-0.
[7] 潘如如;高卫东;钱欣欣;张晓婷. 基于互相关的印花织物疵点检测[J]. 纺织学报, 2010, 31(12): 134-138.
[8] 罗勇;王璐;张新龙;贾舜华;周启澄;胡晓峰;越国瑛. 基于H2O2氧化青黛染色新方法[J]. 纺织学报, 2009, 30(08): 92-96.
[9] 周平.;汪亚明;朱森勇. 时-空域多特征证据学习与增强的印染疵点在线检测[J]. 纺织学报, 2006, 27(5): 1-5.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!