In order to resolve the problems of bag of visual words model(BoVWM) based object retrieval methods,such as low time efficiency,low distinction of visual words and weakly visual semantic resolution because of missing spatial information and quantization error,a novel object retrieval method was proposed.Firstly,E2LSH is used to identify and eliminate the noise key points and similar key points,consequently,the efficiency and quality of visual words was improved.Then,the stop words of dictionary were eliminated by Chi-square model to improve the distinguish ability of visual dictionary.Finally,the spatially-constrained similarity measurement was introduced to accomplish object retrieval,and a robust re-ranking method with the K-nearest neighbors of the query for automatically refining the initial search results was introduced.Experimental results on Oxford5K and Flickr1 datasets indicated that the distinguish ability of visual semantic expression is effectively improved and the object retrieval performance is substantially boosted compared with the traditional methods.
