文献类型： 外文期刊

作者： Gao, Rong-hua ¹ ; Wu, Hua-rui ¹ ;

作者机构： 1.Beijing Acad Agr & Forestry Sci, Beijing Res Ctr Informat Technol Agr, Beijing, Peoples R China

2.Natl Engn Res Ctr Ctr Informat Technol Agr, Beijing, Peoples R China

3.Minist Agr, Key Lab Informat Technol Agr, Beijing, Peoples R China

关键词： characteristics extraction;color;texture;shape;partial occlusion

期刊名称：INTERNATIONAL CONFERENCE ON COMPUTER SCIENCE AND ARTIFICIAL INTELLIGENCE (ICCSAI 2014)

ISSN：

年卷期： 2015 年

页码：

收录情况： SCI

摘要： Global characteristics are directly using a variety of standard machine learning and distance metric during image retrieval algorithm by feature vectors for image description, but there are significant limitations, such as the image is very sensitive to changes in global illumination, etc., while the lack of directly descriptive ability for objects in an image. Characteristics recognition is one of the primary image processing operations, which means it is the first operation of the processing of an image. Delete an occluded area from the image. There are two kinds of existing methods: one method is to manually remove the obstructions to be a uniform color fill or outline its contours, another way is to use the shortest path algorithm or the expansion of biology technique to extract optimal edges. In this paper, to solve the complex background images and the characteristics extraction error because of mutual occlusion, the partial characteristics extraction algorithm is proposed. Partial characteristics are extracted by deep analysis of the texture features, shape local features, and combined with the overall characteristics of the morphology, structure, etc. And achieve images efficient and fast retrieval.