曾金发.基于颜色信息与空间特征的自适应商标检索算法[J].包装工程,2018,39(9):212-219.
ZENG Jin-fa.The Adaptive Trademark Retrieval Algorithm Based on Color Information and Spatial Features[J].Packaging Engineering,2018,39(9):212-219. |
| 基于颜色信息与空间特征的自适应商标检索算法 |
| The Adaptive Trademark Retrieval Algorithm Based on Color Information and Spatial Features |
| 投稿时间：2017-11-29 修订日期：2018-05-10 |
| DOI：10.19554/j.cnki.1001-3563.2018.09.036 |
| 中文关键词: 商标检索 主颜色描述符 空间特征 k-均值聚类 模糊直方图分析 |
| 英文关键词: trademark retrieval dominant color descriptor spatial features k-mean clustering fuzzy histogram analysis |
| 基金项目:江西省教育厅科学技术研究项目（GJJ161401） |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 为了增强商标检索技术对商标特征的描述能力，改善其在外来干扰下的检索精度与鲁棒性。方法 提出一种基于颜色与空间特征自适应结合的商标检索算法。首先，引入主颜色描述符（DCD），将其作为颜色特征检测器，并在颜色特征提取时嵌入k-均值聚类算子，增强颜色区域，准确提取颜色特征。随后，每个商标被量化为8个显色的最大值，以便提取每个颜色分量中的空间分布信息。然后，通过利用不同的权重来平衡颜色与空间特征的重要性，定义一种基于模糊直方图分析技术，计算每个商标自适应系数，以准确描述彩色商标的图像特征。最后，通过Euclidean距离进行相似度量，输出检索到的商标。结果 实验结果表明，与当前商标检索方法相比，所提算法具有更高的检索精度与鲁棒性，呈现出更理想的P-R曲线，在召回率为0.7时，其检索准确率仍可达到90%。结论 文中检索方法具有较高的检索精度，在包装商标检测、商标版权保护等领域中具有良好的应用价值。 |
| 英文摘要: |
| The work aims to enhance the description ability of trademark retrieval technology for trademark features and improve its retrieval accuracy and robustness under external interference. A trademark retrieval algorithm based on the adaptive combination of color and space features was proposed. Firstly, the dominant color descriptor (DCD) was introduced as the color feature detector; and the k-mean clustering operator was embedded at the time of color feature extraction to enhance the color area and extract the color feature accurately. Secondly, each trademark was quantified as the maximum of eight colors, to extract the spatial layout information of each color component. Then, a technique based on fuzzy histogram analysis was defined by means of different weights to balance the importance of color and space features. The adaptive coefficient of each trademark was calculated to accurately describe the image features of a color trademark. Finally, the Euclidean distance was used to measure the similarity and output the retrieved trademark. The experimental results showed that, the proposed algorithm had higher retrieval accuracy and robustness compared with the current trademark retrieval methods, and presented a more ideal P-R curve. Its retrieval accuracy could still reach 90% when the recall rate was 0.7. The proposed retrieval method has higher retrieval accuracy and good application values in the field of package trademark detection and trademark copyright 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