The Design And Implementation Of Clothing Retrieval System Based On Image

With the rise in the number of domestic Internet users year by year and thedevelopment of e-commerce technology,the online shopping market has expanded eachyear.Clothing,shoes and bags are the most traded commodity in the entire onlineshopping market.The shopping model that buying clothing goods with clothingshopping website is increasingly being recognized.Buying clothing online is animportant way for people to buy clothes.Nowadays,online clothing shopping isbecoming an increasingly popular model.Customers can conveniently search for their favorite clothing by typing somekeywords such as “white short-sleeved chiffon dress” in online shopping websites suchas Taobao.com, VANCL and amazon.cn.By browsing the search records,customers filterout the clothing which they most satisfied with and buy it.However，the buyers’inaccurate description of the clothing will cause buyers to spend a lot of time to pick outthe target clothing, in addition, the clothing products on the Internet are more and morenowdays, so this situation often happens. In order to improve people’s search experience,many Internet companies have put in the study of search using image visual features,but using the the underlying visual features simply would cause the absence of imagesemantics. In view of this, this paper make use of the online shopping sites’ clothingimage as well as the marking information the sellers provided, and achieve mappingimage to the keywors through the process of feature extraction and image clustering,andfinally complete the design and implementation of clothing i retrieval system based onthe image. The main contents of this paper are as follows:(1) This paper has carried out clothing image clustering make use of image featureextraction and clustering algorithm on the condition of the clothing image and their textdescription of the online shopping website has been already collected.Based on theresults of image clustering as well as the text description，this paper has carried out themapping of image to keywords and establish the contact between image visual featureand semantics.(2) The traditional k-means clustering algorithm is improved after the research and analysis on it. The dissimilarity matrix of the dataset is used to determine the initialclustering centers and the K value，and the improved method is applied to clusterclothing image.Experiments show that the improved method is more effective relative totraditional k-means method on the clustering of clothing image feature extracted in thispaper.(3) This paper utilize mature keyword-based image retrieval technology，integratethe modules of image feature extraction and keyword mapping，design and implementthe clothing retrieval system based on image. The system map the image provided byusers to keywords and use the ksywors to achieve retrival, so the system is a kind ofretrival system based on image semantics.The development of this system is completed using java language in the Eclipsedevelopment environment.