国土资源遥感Issue(4)：8-13,6.DOI:10.6046/gtzyyg.2015.04.02
一种参数化影像多目标检索的相似度评估模型
A parameterized evaluation model of similarity on multi-target retrieval for remote sensing image
摘要
Abstract
In content - based image retrieval ( CBIR ) , the feature selection of target retrieval and similarity computing is a hot issue of current research. As the object-oriented image processing methods are conducive to describing the spatial relationships between objects, the authors proposed an parameterized evaluation model of similarity for multi-target retrieval on the basis of an analysis of multi-feature descriptor. This model not only takes into account the global spatial relationship between image objects but also considers the local features of image targets. The model organizes the multi -dimensional features for an image using the analytic hierarchy process ( AHP) method under moving picture expert group ( MPEG7 ) standard, then the eigenvalues are calculated using existing methods, and finally the weighting coefficients are set for calculating the overall similarity on the basis of user preferences so as to evaluate the effect of multi-target retrieval. Thus it can provide a powerful reference and experience for multi-target retrieval in an image database. The experimental results show that target retrieval is not only related to the method of similarity calculation but also related to the complexity of the image itself.关键词
遥感影像/多目标检索/相似度评估模型Key words
remote sensing image/multi-target retrieval/accuracy evaluation model分类
计算机与自动化引用本文复制引用
曾志,周永福,杜震洪,刘仁义..一种参数化影像多目标检索的相似度评估模型[J].国土资源遥感,2015,(4):8-13,6.基金项目
国家自然科学基金项目“内容-过程的海气二氧化碳源汇格局及时空演变研究”(编号:4101356)和博士启动基金项目“云环境下大数据遥感影像高效处理关键技术”(编号：C513.0201)共同资助。 （编号:4101356）
