基于个性化本体的图像语义标注和检索

Personalized ontology-based semantic annotation and retrieval of image model

针对目前图像检索系统较难实现语义检索的问题,提出了一种新的以本体为核心的图像语义标注和检索模型。构建个性化本体描述图像语义,继而提取基于概念集的图像语义特征并利用本体中"Is-A"关系设计相似性度量方法最终实现语义扩展检索。其难点在于顶级本体向个性化本体进化,以及基于概念集和"Is-A"关系实现语义相似度量的方法。通过系统的初步实现与相关实验的验证,该模型的检索准确度可达88.6%,明显高于传统的基于关键字和基于通用本体的图像检索,实现了图像智能检索功能。

In order to solve the problem that it is difficult to implement semantic retrieval of image, a novel ontologycored semantic annotation and retrieval model of image was proposed, which constructed a personalized ontology to describe image, extracted the image semantic features based on the concept set and used inter-relationship ＂ Is-A＂ of ontology to measure the similarity between the images and query. The difficulty lies in the personalized evolution of the top ontology, as well as the semantic similarity measure based on concept set and inter-relationship ＂ Is-A ＂. This model was implemented and some related experiments were carried out. The experimental results validate the retrieval accuracy as high as 88.6%, significantly higher than the traditional keyword-based and generic ontology-based image retrieval models, which realized intelligent retrieval of image.