Research On Visual Feature Analysis And Classification Of Network Education Graphics Resources

In recent years,with the continuous development of educational informationization,the demand for Education and information resources is more and more large.As an important form of media resources,image which is intuitive and vivid,is playing more and more important role in the educational and teaching practice.In order to conduct unified management of image resources to facilitate the retrieval of people use,we need to category the range of images,so the urgent problem in the image resources storehouse construction is How to seek the effective image classification way to achieve the better classified effect.The graph is a kind of quite simple image,which is Chart or shape Constituted by geometry essential factors like spot,line,surface and non-geometry essential factor like light and shade,gradation,color,and so on.Because its goal is clear about,its background is unitary, its characteristic is obvious,and its wide applies in the education teaching,the research on graph classification has important meaning.This article has done the dissect of the visual characteristic of graphic,and has conducted the graph classified research based on visual characteristic.The main jobs are as follows:Firstly,we analyzed the description of graphs' visual characteristic and realized many kinds of description parameters of shape feature,Line characteristic and spatial relations characteristic,including relative dispersion,probability and variance of distance distribution, region degree of saturation,line laxity,eccentricity,the marginal density,area of Approaches polygon,the slender length,the concave convexity,invariant moment.Secondly,we built an educational graphics library which contains about 7000 graphics, through the observation and the investigation of 100,000 digital pictures,we has established the image classification system which relying on the vision and the semantics.Firstly, according to whether the graph contains color,we divide it into it colored graphs or black and white graphs,considered the color characteristic of graphs is not very rich,the experiment mainly has carried on the classification of black and white graphs.The black and white graphs carries on the classification according to levels,which have 2 first-level classifications, 4 second-level classifications,8 third-level classifications and subclass.Thirdly,according to graph classification system we limited each kind of graphs' concept,analyzed the available visual characteristic,and provided the flame and the basis for the later image classification experiment. Fourthly,on the foundation of visual characteristic's extraction,using threshold classification method for graphs whose visual character is obvious,and SVM fuzzy classification method for graphs which cannot be classified by sole parameter,we has experimented on the Colored graph/black and white graph,single object graph/multi-object graph,Line graph/packing graph,radiation graph/polymerization graph,polymerization graph's subclass,outline graph/striated pattern,Multi-object graph and its subclass and so on, achieving the high accuracy and correct.In addition,we analyzed the result of the experiment for each class,pointed out reasons of the mistake classification and giving example pictures, and proposed the feasible improvement direction.This article taking VC++6.0 as the developing platform,unifying the Matlab 6.5 as data simulation tool,has developed "graph classification system based on visual characteristic" based on the Windows XP operating system,carry on the classified experiment on many levels in the system,which has obtained quite satisfying result.The findings indicated that the research on the visual characteristic of graphs is advantageous in enriching the description method on simple images and enhancing the the effect of graph classification;the establishment of graph classification system based on visual characteristic has enriched the graph storehouse of network educational resources,which is advantageous in the resources' cross platform,trans-regional sharing;the classified experiment of all levels of graphs has obtained the satisfactory rate of accuracy and served the experiment purpose,simultaneously it has also provided the convenience for the education graph resources' retrieval.