Research On Optimization Method Of Lending Service In Library Based On Collaborative Filtering And AR Technology

Library is the treasure house of human knowledge,more and more people choose to borrow books to improve themselves,but the traditional library lending service consumes a lot of time for borrowers to search and locate books,which brings a lot of inconvenience to borrowers.Therefore,libraries can provide a proactive service of personalized book recommendation and navigation of recommended books to optimize library lending services which can help borrowers find satisfactory books more quickly to improve the lending experience.However,for sparse book lending data,most of the collaborative filtering models have difficulties in starting the recommendation system due to insufficient initial data.How to perform low-cost user indoor positioning to obtain user location information and how to collect book location information for book navigation in complex scenarios like libraries is a problem that needs to be solved in the implementation of book navigation services.Firstly,to address the problem of insufficient data in the early stage of book recommendation system startup resulting in weak performance of recommendation,this paper proposes a coupled collaborative filtering model based on attention mechanism.The coupled collaborative filtering model is used to lift the assumption that traditional collaborative filtering is independently and identically distributed,and at the same time,the attention mechanism is used to deeply mine the key degree of explicit attributes,which improves the recommendation effect of the model.Experiments prove that the proposed model outperforms the traditional collaborative filtering method and coupled collaborative filtering model in two evaluation indexes,Hit-on Recall and NDCG.Secondly,to solve the problem of how to obtain the initial location of users in navigation services,the research completed in this paper includes:(1)proposing a gridded offline keyframe map construction method,keyframe acquisition by smartphones,and localization by image matching,which reduces the cost while realizing an image matching method based on direction sensors and Bo W(bag-of-words model),which improves the image matching The speed is improved,and the effectiveness of the method is verified through experiments.(2)To solve the problem of image mis-matching caused by high similarity of indoor scenes in libraries,an improved TD-IDF similarity scoring rule is proposed in this paper,and the map key frames of adjacent directions are acquired by using the orientation sensor to assist in scoring,and the method is proved to improve the robustness of image matching in library scenes through experiments.(3)To address the scale uncertainty problem of the library scene images captured by monocular cell phones for positional estimation,this paper proposes a method to use gridded offline keyframe map information for scale recovery,which provides more accurate positioning information.Finally,to solve the problem of book location collection in navigation services,this paper proposes a book marker methods based on AR(augmented reality)technology to support users to collect book location information and build a map of book location information.And the AR navigation function is realized by the above-mentioned user initialized location and book location information,which verifies the feasibility of the research.