Approach To Intelligently Generating We-map Symbols

Maps play an essential role in people’s lives today,and traditional cartography methods have become increasingly mature,with extremely well-established production rules.However,with the increasing demand for traditional maps in terms of their large-scale and accuracy,more and more cartographic methods are being put on the agenda.Strictly speaking,cartography is subject to many constraints,such as strict scale,map projection,and map symbols.In order to meet the production needs of existing maps,cartographic methods are constantly improving,from relying on manual maps created by experienced map makers,to semi-automatic mapping that combines algorithms,to fully automatic mapping that is now possible.The intelligent generation of fully automatic cartographic techniques has also become a new development direction.However,as modern society’s demand for maps becomes more diversified,cartographic methods are also constantly evolving and improving.Therefore,in the actual process of cartographic,strict map standards are no longer universally applicable.In the fast-paced internet era,there have been many map requirements that focus on writing "meaning" rather than "form".With the explosive growth of self-media applications,maps have transformed from physical objects to electronic media,focusing on point-to-point information transmission,expression,and understanding between users,such as the subway transportation map,which focuses on expressing the order of stations rather than strict distance and direction.In this case,the semantic function of the map is greater than the strict geometric accuracy,the cartographic cycle is greatly shortened,the threshold is lowered,making it easier for map-makers and users to collaborate,and point-to-point real-time interaction and dissemination are more convenient.To this end,a lightweight map,the we-map,has emerged.The we-map is a "grassroots" map aimed at the general public,with low production requirements.The maker does not need to go through rigorous professional training,and users can easily participate in cartography and application and interact on their personal electronic devices.The we-map focuses on the understanding and expression of semantic information,so the most important aspect for the we-map is the map symbol as the basic element.However,the production of map symbols also poses difficulties,as traditional map symbols require manual drawing.For experienced map-makers,there are still challenges in extracting prominent features of land and expressing them in a beautiful and reasonable manner.The users’ participation in cartography is low,the threshold of we-map cartography is high in the step of map symbols generation,a method of intelligently rapid making symbols based on deep learning——“intelligent generation of we-map symbols” is proposed in this paper.So,we can maximally fulfill the requirement that each we-map user has ability to become a cartographer.Specifically,this can be divided into: self-service and customization for the intelligent generation of we-map symbols.The details are as follows:(1)Self-service intelligent generation of we-map symbols —— an intelligent symbol matching solution for we-maps through image recognition.Firstly,the preprocessed small homemade dataset is trained by transfer learning on the models VGG16 to finish classifying the dataset.Next,the completed classified data is matched with the map symbols in the mini symbol library that has been constructed in advance.Finally,the corresponding map symbols are obtained and applied to the real situation,in other words,users can match symbols in the mini symbol library by the easiest way(e.g.,by taking a picture),helping them to produce maps.The experimental results show that the method can finish the task of intelligent symbol recognition on images basis,meanwhile,it is useful to simplify the cartographic process in the step of symbols generation.As a whole,this method improves the users’ participation in cartography and achieves the goal that lower the threshold of we-map cartography in the step of map symbols generation,which effectively meets the demand that each we-map user has the ability to become a cartographer.Of course,the method also saves the user’s time of cartograph,then provides a feasible channel for the subsequent rapid cartography of we-map.(2)Customized symbols for we-map cartography: To address the issue of insufficient personalization in symbol selection in the research of an intelligent symbol matching solution for we-maps through image recognition,a method of generating customized symbols for wemap cartography is proposed.First,the idea of user story mapping is borrowed to guide users to express their needs for exclusive symbols in the form of stories and create symbol stories that belong to them.Secondly,a deep learning-based saliency detection model Bas Net is adopted as the core,and an algorithm is designed to implement the technical process,thereby realizing the idea of generating customized symbols.Finally,user cognitive surveys and analyses are conducted on the generated customized symbols.The results show that:(a)generating customized symbols can facilitate the point-to-point instant sharing function of wemap cartography;(b)providing users with a convenient way to generate customized symbols enables them to efficiently and quickly create maps and receive personalized services.By generating map symbols through this customization method,users do not need professional skills,in-depth learning,or bear basic cartographic costs.The general public can use their favorite images and mobile devices to quickly and conveniently create their own satisfactory maps,thereby reducing the threshold for cartographic and increasing the participation and personalization in the symbol generation process.In this paper,we take advantage of the opportunities provided by deep learning(as for transfer learning,visual saliency detection,VGG16 and Bas Net)for intelligent map making,and propose two intelligent mapping methods for we-map symbol generation: “an intelligent symbol matching solution for we-maps through image recognition” and “customized symbols”.Using transfer learning,visual saliency detection and other related technologies in deep learning,we implement an image-based input method,which can intelligently recognize,match or generate personalized symbols that meet user expectations.The experimental results show that these two methods can improve user participation and personalization in symbol generation,lower the mapping threshold for users,meet the demand of everyone can map in symbol generation,and enhance the important role of map symbols in information transmission.