摘要
提出一种基于粒子群优化的多特征融合的商标图像检索方法,该方法可自动优化多特征融合的权重,提高图像检索系统的自适应性,解决了多特征商标图像检索中的权重分配问题。在1000幅图像构成的商标图像库进行检索实验,实验结果表明,与基于单一特征的检索方法和一些多特征融合的检索方法相比,提出方法的检索性能最优。
This paper develops a method of trademark image retrieval based on particle swarm optimization in multi-feature fusion.It can optimize the weights of multi-feature fusion automatically,improve self-adaptive of the image retrieval system,and solve the allocation problem of feature weights in the trademark image retrieval.After retrieving in the trademark image database constituted of 1000 images,the results show that the proposed method has a better retrieval performance than the single-feature-based retrieval methods and some multi-feature-fusion retrieval methods.
出处
《计算机工程与应用》
CSCD
2012年第21期186-190,共5页
Computer Engineering and Applications
基金
国家自然科学基金(No.61073138
No.61103141)
江苏省高校自然科学研究重大项目(No.11KJA520004)
关键词
粒子群优化算法
多特征融合
特征权重分配
商标图像检索
Particle Swarm Optimization（PSO）algorithm
multi-feature fusion
feature weight assignment
trademark image retrieval
