 ==>> 点击下载文档 |
本文首先分析了商标检索管理中存在的问题以及产生的原因,在此基
础上提出对商标检索管理进行策略优化,特别提出改进商标检索方法,对
此,本文提出了一种基于深度学习的商标检索方法,该方法首先建立一个
包含85800张商标图像的商标库,然后通过深度学习实验提取特征和进行
特征匹配得到最终的检索结果,实验的具体做法是先将商标库中的商标图
像输入两个已经训练好的深度学习模型中分别进行特征提取,为了提高运
算效率对提取的特征分别进行哈希编码,然后将不同模型所得的特征进行
融合,最后通过K-means聚类获得与待检索图像相似度较高的一组图像
来作为商标检索的结果。通过大量的深度学习实验和对比试验验证本文方
法在检索的准确度和效率上较传统方法有了进一步的提升。接着为了验证
本文提出的基于深度学习的商标检索方法的有效性,本文建立相关评价指
标和评价模型,对商标注册申请的负贵人进行实验和问卷调查,让他们分
别采用本文提出的检索方法和商标总局官方检索方法进行了商标检索并
填写问卷,然后对问卷的结果进行分析,从而本文方法的冇效性得到验证
本文将计算机图像处理的方法1、;/:川于解决商标检索管理问题1丨1。本文
I
J匕京化.1:人学硕士学位论文
提出得基于深度学习的商标相似度检索方法,通过深度学习网络层层特征
提取和抽象,在网络最高层输出的特征与传统的图像检索方法相比,既有
包含局部信息的低层特征又有高层语义特征;通过局部感知和权值共享,
减少变量的数量级,提尚了检索效率;为了进一步提尚商标检索的精度,
应用集成学习的思想,对多个深度模型的结果进行投票得到最终检索结
果,获得在检索精度和效率上均优于传统方法的新的商标检索方法
本文通过将深度学习方法应用于商标检索管理中,对当前的商标检索
方法进行改进,有利于商标检索管理水平的提升,有利于商标枷锁管理中
问题的解决
关键词:商标,深度学习,商标检索管理,策略研宂
II ^
THE APPLICATION RESEARCH IN THE MANAGEMENT
OF TRADEMARK RETRIEVAL BASED ON DEEP
LEARNING
ABSTRACT
Trademark is the product of modern economy, it is so closely related to
the image, culture and brand of company. Trademark image is just an intuitive
performance of trademark, it is composed by the graphics, text, letters, which
is related to the image of product and brand. In view of the importance of the
trademark to the enterprise, it is necessary to search the trademark image
similarity in the process of trademark similarity examination.
Inefficient problems in the process of trademark retrieval management
trigger a low efficiency of trademark review and trademark application
backlog. At the same time, inefficient problems of trademark retrieval method
and incomplete search results cause problems in trademark similarity
examination. Therefore, putting forward a kind of high accuracy fast
trademark retrieval methods to improve effectiveness and fairness of the
trademark management has an important role.
Deep learning as a hot method in current artificial intelligence field, is
widely used in image processing field. The method of applying deep learning
to learn the characteristics of the image automatically, is to put trademark
images into the deep learning model which has been trained, and then conduct
mi
北京化I:大学硕丨:学位论文
feature extraction. Finally, a set of images with high similarity to the image to
be retrieved is obtained by K-means clustering as a result of trademark search.
In order to improve the efficiency of the trademark retrieval and the accuracy
of the trademark search, we hash code to characteristics, vote for the retrieval
results with the method of integrated learning so as to get the MAP (on
average) retrieval rate higher trademark retrieval method.
In order to validate the proposed trademark retrieval method based on the
deep learning effectiveness, we conducted an experiment with the person in
charge of the trademark registration application. They used the similarity
retrieval system proposed in this paper and the official similarity retrieval
method, and filled out the questionnaire, then we analyzed the validity of the
questionnaire, which verified the validity of the method.
Therefore, the method of trademark retrieval proposed in this paper can
improve the efficiency and effectiveness of trademark retrieval management.
In addition, the paper also put forward other method to research on the
strategy of trademark retrieval management from the view of increasing
publicity efforts to expand the use of channels, and call on the government,
enterprises, consumers together to improve the effectiveness of trademark
management.
KEY WORDS: trademark , deep learning , trademark retrieval,, strategy
research
IV
目录
胃一章绪论 1
1.1研宄背景 1
1.2研宄意义 3
1.3主要研宄工作 3
1.3.1主要研宄方法 3
1.3.2主要研宄内容 4
1.3.3石开宄思路 4
第二章研究现状 7
2.1深度学习方法的研宄综述2.2商标检索管理研究综述2.2.1商标相似度判定2.2.2商标检索方法研宄2.2.3商标检索管理策略研宄2.3本章小结 14
第三章基于深度学习的商标检索方法的提出3.1商标检索管理问题研宄3.2商标检索管理策略优化3.2.1加大宣传力度,提高商标检索意识3.2.2拓展使用渠道3.2.3改进商标检索的方法一一基于深度学习的商标检索方法的提出3.3本章小结 19
v
北京化工大学硕士学位论文
第四章深度学习方法在商标检索中的应用设计
21
4.1深度学习方法 21
4.1.1深度卷积神经网络的原理和特点
21
4.1.2深度卷积神经网络的网络结构
22
4.2迭代量化方法(ITQ)
24
4.3基于深度学习的商标检索方法的具体实现
25
4.3.1特征提取阶段
25
4.3.2特征匹配阶段
26
4.4实验设计
27
4.4.1商标库的建立
27
4.4.2实验思路 28
4.4.3实验平台简介
28
4.4.4实验过程 28
4.4.5对比实验设计
30
4.5实验结果对比与分析
32
4.6本章小结 33
第五章深度学习方法的有效性分析
35
5.1商标检索方法评价指标
35
5.2商标检索方法的评价模型
36
5.3本章小结 39
献章结论 41
6.1商标图像检索方法方面
41
6.1.1目前方法存在的不足与改进
41
6.1.2方法的创新点
41
VI
tl＊
6.2商标检索管理方面
41
6.2.1企业在角度 41
6.2.2消费者角度 426.2.3政府角度 43
参考文献 45
附录 49
賴 53
研究成果及发表的学术论文
55
作者和导师简介 57
VII
Contents
Chapter 1 Introduction1.1 Research background1.2 Research meaning1.3 Main research work1.3.1 Main research ideas1.3.2 Main research content1.3.3 Main research ideasChapter 2 Related research progress2.1 Deep learning. 7
2.2 Trademark retrieval2.2.1 Trademark similarity judgment2.2.2 Trademark images retrieval methods2.2.3 Trademark retrieval management strategy researchChapter 3 Proposed the trademark retrieval method based on deep
learning 15
3.1 Trademark retrieval management3.2 Trademark retrieval strategy optimization3.2.1 Enhance the propaganda3.2.2 Change the trademark retrieval method3.2.3 Expand use channel proposed the trademark retrieval method based on deep
learning 18
3.3 Summary 19
Chapter 4 Trademark retrieval based on deep learning
21
4.1 The basic principle of deep learning
21
4.1.1 Features of the deep convolutional layers
21
v
0^
4.1.2 Structures of the convolutional layers
22
4.2 The introduction of ITQ
24
4.3 The introduction of the trademark retrieval based on deep learning
25
4.3.1 Feature extr
。。。以上简介无排版格式，详细内容请下载查看
