刘婷,王茜娟.基于泽尼克矩耦合颜色空间加权度量的商标检索算法[J].包装工程,2018,39(23):216-223.
LIU Ting,WANG Xi-juan.Trademark Retrieval Algorithm Based on Zernike Moment Coupled Color Space Weighted Measure[J].Packaging Engineering,2018,39(23):216-223. |
| 基于泽尼克矩耦合颜色空间加权度量的商标检索算法 |
| Trademark Retrieval Algorithm Based on Zernike Moment Coupled Color Space Weighted Measure |
| 投稿时间：2018-08-15 修订日期：2018-12-10 |
| DOI：10.19554/j.cnki.1001-3563.2018.23.035 |
| 中文关键词: 商标检索 泽尼克矩 颜色空间度量 加权Euclidean 距离测量 特征匹配 |
| 英文关键词: trademark retrieval Zernike moment color space measure weighted Euclidean distance measurement feature matching |
| 基金项目:河北省高等学校科学技术研究项目（Z2015085）；江西省教育厅科学技术研究项目（GJJ160750） |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 针对商标检索系统中利用单一特征进行识别和度量时，往往难以充分表征商标特征，易出现检索精度和鲁棒性不高等问题，文中拟设计一种泽尼克（Zernike）矩耦合颜色空间加权度量的商标检索方案。方法 首先，利用Zernike矩作为商标的形状描述符，充分描述商标的形状信息。随后，利用颜色空间来描述图像中像素空间信息的颜色分布特征。然后，分别将输入商标的Zernike矩特征、颜色空间特征与存储在数据库中的特征进行匹配，以计算Zernike矩特征的加权Euclidean距离与颜色空间度量。最后，联合颜色空间度量与Euclidean距离，综合考虑形状与颜色特征，形成新的距离测量规则，输出与查询商标相似的商标。结果 实验数据表明，与当前商标检索算法相比较，所提算法具有更高的检索准确率与鲁棒性，表现出更为理想的Precision-Recall以及平均准确率（Mean Average Precision, MAP）。结论 所提算法返回的图像与查询图像相似度较高，在商标注册、侵权保护等方面中具有一定的参考价值。 |
| 英文摘要: |
| The work aims to propose to design a trademark retrieval scheme based on Zernike moment coupled color space weighted measure with respect to the problem that it is difficult to fully characterize the trademark features, which can easily lead to poor retrieval accuracy and robustness during identification and measurement by the use of a single feature in the trademark retrieval system. Firstly, the Zernike moment was used as the shape descriptor of the trademark to fully describe the shape information of the trademark. Subsequently, a color space was used to describe the color distribution features of pixel spatial information in the image. Then, the Zernike moment feature and color space of the input trademark were matched with the features stored in the database to calculate the weighted Euclidean distance and the color space measure of the Zernike moment feature. Finally, combined with color space measure and Euclidean distance, the shape and color characteristics were comprehensively considered to form new distance measurement rules and output the trademark similar to the query trademark. The experiment data showed that, compared with the current trademark retrieval algorithm, the proposed algorithm had higher retrieval accuracy and robustness, indicating more ideal Precision-Recall and Mean Average Precision (MAP). The image retrieved by the proposed algorithm has a higher similarity to the query image, which has a certain reference value in the aspects of trademark registration and infringement protection, etc. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
渝公网安备 50010702501716号