摘要
针对传统"视觉词袋模型"在进行场景分类时只利用图像的特征域,忽略其空间域中上下文语义信息的问题,提出一种基于图像上下文语义信息的场景分类方法.在传统"视觉词袋模型"的基础上,引入马尔科夫随机场模型对图像上下文语义信息进行建模,利用潜在的狄利克雷分布学习场景的主题分布,且利用支持向量机构造场景分类器.对15类场景的分类实验证明该方法能够有效提高分类精确度.
A novel approach was proposed to categorize the scenes. Based on the traditional Bag of Visual words （BOV） model, the Markov Random Field （MRF） was introduced to combine the feature field and the spatial field in order to quantify the image into a set of unordered visual words. And then the Latent Dirichlet Allocation （LDA） was used to learn the topic distribution. At last, the Support Vector Ma- chine（SVM） was applied to identify a new image. The experimental results on 15 nature scenes show that the introduction of the contextual semantic information on the basis of the traditional method can enhance the classification accuracy.
出处
《四川大学学报（自然科学版）》
CAS
CSCD
北大核心
2013年第6期1223-1229,共7页
Journal of Sichuan University(Natural Science Edition)
基金
教育部"春晖计划"(z2011149)
