Research On Plane Geometry Information Extraction And Automatic Drawing

Understanding the meaning of a problem is a prerequisite for achieving computerlike human problem-solving.Planar geometric graphic is an important component of geometric problems,and information extraction of geometric graphic is an important means of understanding the meaning of geometric propositions.Similar to human proof of geometric propositions,automatic composition of planar geometry is also one of the most vital auxiliary means for computer automatic proof.Generally,planar geometric propositions contain not only textual information but also geometric graphic information.Combining the two information can improve the accuracy of understanding the meaning of geometric questions;At the same time,in the process of computer automatic proof,combined with the constructed geometric figures,it can assist reasoning to optimize intermediate information and improve reasoning efficiency.This article mainly realizes the optimization of information extraction algorithm for planar geometric images,and implements an automatic drawing system to assist in generating image information for geometric problems without image information.The main content includes:(1)Collect datasets with different noise levels and sizes for different preprocessing experiments,and select the optimal preprocessing method.(2)Use line segment merging and corner recognition to improve the Hough transform of lines and circles,and compare the geometric element extraction experiments of the pre improved and post improved Hough transform methods to verify the effectiveness of the improvement;(3)Build a cross language OCR platform using two methods to extract non geometric element information,compare performance and accuracy,and combine the extracted geometric elements to obtain the information that can be obtained from the entire image based on input;(4)The framework design optimization of the automatic drawing system fully utilizes Java’s reflection properties and abstract factory design patterns,reduces the coupling degree of each module,and implements a recursive iterative algorithm to update constraint values,making the generated static graphics have an appropriate aspect ratio and size.The experimental results show that the improved geometric pattern recognition algorithm has a recognition accuracy of 85.0% for shapes without curves,with a perfect recognition rate of 94.1% for thin line images and 58.3% for thick line images;The perfect recognition rate for graphics that contain both circles and line segments is 52.9%.If there are slight flaws on the image that do not affect the overall effect,the recognition rate is 98.3%.After extracting the geometric information successfully,combining OCR system to identify letters and mathematical notation,the relative position information of points can be obtained,and their feature attributes can also be obtained according to the input.For the automatic drawing system,it draws on the architecture design of Io C and AOP ideas in the Spring framework,improving the code reuse rate and enhancing its maintainability and the generated image conforms to the input description with an aspect ratio and moderated area.