| Currently, the urgent protection of endangered species, due to our wide range of protectedanimals, people have limited knowledge on the protection of animals, which for the publicsecurity organs to crack down on criminals caused some difficulty. Protection of animals facingsharp increase in the amount of image data, how to use the image retrieval technology, effectiveand efficient for mass image retrieval has become an urgent need. Traditional content-basedimage retrieval using low-level physical features to represent image content, but low-levelphysical characteristics of the image on the image content with people far from the high-levelunderstanding, people want the image retrieval on the semantic level. This in-depth study of thevisual semantic concept detection technology, the paper mainly involves the following threeaspects:(1) For the number of image features as the higher cost of making a great distancecalculation when calculating the need for dimensionality reduction of image features for theproblem, this chapter first describes the commonly used dimension reduction methods, includingtraditional dimension reduction methods include line of dimensionality reduction and nonlineardimensionality reduction, but these methods require large amounts of matrix operations,computation speed is slow. Position sensitive hashing algorithm is one of the recent researchfocus, through the analysis of random mapping-a position-sensitive hashing, random mappingmethod of characteristics for processing, we use it for dimensionality reduction and animalimages for quick retrieval. Experimental results show that the method reduces the number ofimages involved in the calculation, making the retrieval speed is greatly increased.(2) The local features can be better than the description of global characteristics of theimage of local information, so it is relevant in many areas of image processing are used,especially in object recognition, classification and so on. In image retrieval on the protectio n ofanimals, you can also use it to find images similar to the query target. In addition, the invertedindex in a text processing application has been more in the image search can also be calculatedusing similar techniques similarity of the two images. Based on these technologies, this chapterfeatures to achieve a partial inverted index of the animal image retrieval algorithm, the localcharacteristics combined with the inverted file, for image retrieval, the method first extracts localimage features, and then the feature vector used to drop E~2LSH dimensional map, and then mapthe results to establish the use of inverted index, and finally through the inverted file gives thenumber of barrels of CPC images is the similarity of image retrieval. The method of retrievalspeed and precision rate is also higher than the reference method.(3) in-depth study of the visual semantics makes to the visual dictionary-based word bagmethod for object recognition and gradually retrieved, so you can solve the major visual searchis similar to the visual image recognition, classification and search problem. Visual search withvisual words to express a class of objects or an image, first the concept of a semanticrepresentation to protect animals, expressed as a result of the use of visual word to describe the similarity between the images or images with a certain class of objects similar to the of. As withthe semantic lexicon representation of the image can be more descriptive of the image containingthe object information, so the visual dictionary has already begun for object classification andidentification, based on this, we propose a visual dictionary of opportunities for image retrievaltechnology to protect animals, although protection of animals images of objects in images thanmore complex, but still with a visual dictionary to some extent reflect the animal’s main message,so the image is used to retrieve the animal can achieve a certain effect, and all the features of themethod up there are more than a big improvement. |
PDF Full Text Request