Research Of Semantic Annotation Based On Object Extraction

With the continued popularity of the Internet and the continued development of imagecapturing device, image as an important information source of human perception of the world,in real-world applications have become more important.How to quickly find the target from amassive collection of digital images image is a study spot.Compared with the increasingamount of data and the development of large-scale image database, traditional based textretrieval technology have not been able to adapt to the user’s research needs, traditionalinformation analysis and retrieval technology is facing new challenges.The emergence ofcontent-based image retrieval, nor fully met through high-level image semantic retrievalintention.As established in the lower visual image similarity measure does not fully expressthe characteristics of high-level semantics.Image retrieval based on semantic is a betterchoice.Semantic-based image retrieval key issues are how to correspond to the underlyingsemantics of the Visual characteristics of human cognition.Current semantic extractionmethods, divided by high-level semantics of the following three: is based on the semantics ofexternal sources of information extraction, semantic extraction based on human interactionand semantic extraction based on knowledge.Image semantic information can be obtained byautomatic image annotation technology, so you can achieve image retrieval on the semanticlevel.In General, underlying characteristics combine top keywords for semanticinformation.Using machine learning methods,Match images and text description.UnderlyingVisual feature is the cornerstone of high level semantic information.Introduction this article describes the background of research on automatic objectextraction and semantic annotation of images;And then gave a presentation on objectextraction using color image processing method, image filtering, image segmentation, imagequantization and BP neural network of artificial intelligence.The3rd chapter, under the basis of pretreatment, introduced an extraction method of theprincipal object in the image.Image is made up of various objects, and different objects, the color distribution and the image will have greater differences in texture.The4th chapter, presented this automatic semantic annotation methods.Verify the validityof the method.Finally, we have the article summary and Outlook.