Design Of Image Retrieval System Based On Cuckoo Search Algorithm

The rapid development of the Internet has closely connected the massive data information and people's lives,as well as contributed to a great increase in pictures,videos and other multimedia information.In the information age,it has been a hot issue that how to find the information we need accurately and efficiently.Texts,by the means of keywords,work as the search object for information search by traditional search engines.Text-based retrieval technology has been very mature,however,the drawbacks of text search lie in that it is difficult to retrieve images that are difficult to describe in words.In addition,sometimes words can hardly express people's search intention intuitively and comprehensively.However,content based image retrieval(CBIR)technology can solve this problem well.Content based image retrieval searches images by uploading pictures instead of typing texts,and then the computer extracts the features of the images automatically,and then match similar images from the image database.At present,the content based image search technology has two main problems that need to be improved: improving the search efficiency and reducing the semantic gap to improve the search accuracy.In this paper,the following work is done based on the CBIR:(1)Extracting image features to construct image feature database,building a basic CBIR system.This paper uses corel1000 as image database.Extracting color moment feature,color correlation feature and LBP texture feature of images to compose a feature vector library,and the use MATLAB as a tool,building a basic CBIR system,realized the function to search for similar images by uploading images.(2)Presenting a search algorithm for image search based on content and CuckooSearch,use the CuckooSearch algorithm,which optimizes in contiguous space,to solve image search problems in discrete feature vector space,improved the search efficiency of CBIR system.CuckooSearch(CS)algorithm is also called “cuckoo search”,it is a swarm intelligence optimization algorithm proposed by Professor YANG from Cambridge University in 2009.The strengths of this algorithm lie in few parameters,good search path and strong global search ability.In this paper,the CS algorithm is applied to CBIR system,regarding the image search as finding optimal solution problem,taking advantage of the CS algorithm.At last,the experiments show that the algorithm has higher search efficiency than the traversal search algorithms in CBIR.(3)Proposing an relevance feedback algorithm based on the CuckooSearchalgorithm adjustment support vector machine parameters dynamically,reduced the semantic gap in the CBIR system.Firstly,regarding relevance feedback as a binary classification problem,support vector machine is used to divide the image into two categories through the result of relevance feedback.Searching for the optimal SVM parameters by CS algorithm,adaptive adjustment of support vector machine parameters.Experiments show that this algorithm is faster and more accurate than traditional CS,PSO,GA algorithms in SVM classification.So,the relevance feedback of image search can get higher accuracy under less feedback times,improved the search accuracy.