Research And Development On Recognition Control Of Graphs In Homework

The exercise is an important teaching process. The problem of innovation of exercise means is raised with the rapid development of modern educational technology and the slowly lag of traditional means of exercise. Researchers make great effort in the reformation of exercise means by distance homework system and can solve some problems such as automatic recognition and correction of mathematical expression. However, for the basis of science and engineering courses such as basic mechanics, homewoks that often contain graphic content and students often need draw many kinds of graphs. This paper in-depth researches on how to automatic recognize and correcting the exercise graph.There are many methods of pattern recognition, the process are generally completed by two steps-feature extraction and recognition, also with the graph recognition in homework. Different with the common graph recognition, graph recognition in homework has its own characters, which need extract the characters of teachers' and students' graphs unit at the same time, including the space characters and attribute characters. Attribute characters are the features of their own and space characters are the features of location and topology relationship in the teachers' and students' graphs. Recognition step is the process of classifying and understanding graph characters that has extracted. There is no accurate calculation model about homework graphs recognition. Based on fuzzy recognition, the paper raises a multi-factor similitude mode-MFModel, which identifies the similarities between students' graphs unit and teachers' key graphs unit. Firstly, it computes the local similarity, and then it computes the overall similarity if the local similarity is bigger than the local threshold. It can be recognized correctly only if the overall similarity between the teachers' key graphs unit and students' graphs unit is bigger than the overall threshold, otherwise it is considered wrong.Several key issues of graph identification in homework have been studied in this paper. The main creative results in the paper are:(1) This paper proposes a model DCone to identify the position relation, which expandes the Cone model, integrates the direction and distance characteristics, that makes the position relationship of the graphs exacter.(2) This paper proposes a new algorithm EGC recognizing the symbols and graphs together. EGC algorithm can solve some special problems that the symbolic expressions recognized are different if students suppose different directions. We need consider the directions of graphs that students have used.(3) This paper proposes a similarity model-MFModel that considers many factors of graphs in homework. It makes up of 4 important factors of graphs recognition in homework including graphs type GUType graphs attributes GUAttri position relationship GUPos and topology relationship GUTop. None of them can independently decides similarity between two graphs, but jointly decides by their own rights.(4) This paper designs a novel object index- OIS, that makes relationship between GTD and graph objects. Based on MFModel and OIS, we propose an algorithm to recognize and control graphs in homework generally, called OIS-GIC.