Design And Implementation Of Clothing Matching Recommendation System For Augmented Reality

With the popularization of computer and internet technology,more and more computer technology applied to the field of clothing.Among them,the research of clothing image processing,3D modeling,and the similar personalized recommendation of clothing are in the majority.And,there is few research for clothing collocation recommendation based on image content.At the same time the updated of clothing lead to garment data resources maintain a rapid growth rate,so it is difficult for users to have plenty of time to choose the right clothes.How quickly recommend the right for users and popular clothing collocation in the huge clothing resources become an urgent problem needed to solve.While the traditional clothing recommendation system is mostly similar clothing recommendation based on users data,the research about clothing collocation recommendation is few.Also the collaborative filtering recommendation system highly depending on user behavior data,although the performance is ideal,but it exists the cold start and work poorly when dealing with sparse data problem.Based on the above issues,this paper proposes a clothing matching recommendation system for augmented reality,the innovation points and achievements of this paper are as follows:1.Research the key technologies and the relevant theories,focus on the Kinect skeletal tracking technology,clothing recommendation system and support vector machine.2.Collect a large number of expert matching clothing data and general clothing data through internet resources;select four image features and extract image feature,which is used as the recommended data base.3.Combine augmented reality and Kinect somatosensory technology to complete the virtual try-on and scene experience;propose crowd-sensing of clothing design,complete the achieve the clothing design.4.Achieve the SVM algorithm,select the appropriate kernel function and parameters,dimension of feature vectors etc to train out a recommendation model and compared with other algorithms in recommendation accuracy.