Research On Automatic Color Transfer Method Of Map Based On Manifold Learning

With the wide application and development of geographic information technology and cartography tools,more and more people use and make maps.Map users and cartographers are gradually blurred.The era of public cartography is coming,and the demand for public personalized cartography is becoming stronger.As the essential element of visual and graphic communication,color is an important part in the process of map cartography.Effective map colors can improve the efficiency and accuracy of geographic information transmission,enhance the aesthetic perception of the map to attract readers' attention;otherwise,invalid map colors will confuse users and influence the transmission of geographic information.At present,many cartographers have researched on the color design of maps,such as Color Brewer and other tools also make it easier.Nevertheless,most of these tools are suitable for specific tasks,cannot be personalized customized design,and have limitations in automation.For novice cartographers and ordinary cartographers who lack background knowledge of color theory and the rules of map color design,there are problems with unscientific color design and low efficiency.Therefore,designing high-quality and aesthetic map colors is still very challenging.Under the background of public cartography,creating a scientific and aesthetic map has become a new problem in the field of cartography.Among them,designing scientific map colors needs to follow many color design rules(such as conventional color rules,color-recognizable rules,etc.),design aesthetic map colors also requires visual perception such as color harmony.In recent years,color transfer methods in computer graphics can transfer the color of one or more images to another.In map color design,some researches also try to transfer the colors of artistic paintings to maps.However,these methods suffer from the limitation of insufficient automation or due to the transfer using raster maps and lack of consideration of relevant map color design rules,making the transfer result more like painting rather than map.Therefore,it cannot satisfy the user needs.Based on the concept of color transfer in computer graphics,this paper studies the method of automatic map color transfer based on manifold learning.The main research contents and achievements are as follows:(1)At present,the image color transfer method does not take into account the problems of color space distribution,color harmony visual perception characteristics.This paper analyzes the color visual variable coding rules and color visual perception theory in the map color design theory.Define the map color transfer as multi-objective and multi-constrained optimization problems.Introducing information entropy and color harmony model to quantify map color space distribution features and color visual perception features,objective functions are constructed.In addition,constraints are color-recognizable rules and conventional color design rules.So a map color transfer model is constructed that considering the scientific and aesthetic of map color design.(2)To solve the problem that discrete color space is difficult to search and the search efficiency is low,a manifold learning method is introduced.The method can reduce the dimensionality of multi-variable,nonlinear,and high-dimensional map color factors.Moreover,Artificial Bee Colony search algorithm is introduced based on color manifolds to solve the constrained search problem,and as to realize the rapid solution of the model.(3)Conducted experimental verification.This article takes four thematic maps as examples,refers to artistic painting pictures,and uses the proposed automatic map color transfer method to automatically transfer the color of the pictures to the map.In addition,uses eye movement experiments and satisfaction questionnaires to compare the generated results with the standards Map color scheme and results of NVIDIA's Fast PhotoStyle image color transfer method.Evaluate from objective and subjective indicators.The experimental results show that the overall effect of the color schemes generated by the method proposed in this paper were not worse than that of the two comparative experimental color schemes,and some indicators were even better than the comparative experimental color schemes.