Online Hand-Drawn Geometric Shape Recognition for the Application of Mathematics E-Teaching Environments

• Main Authors: Po-How Chang, 張柏豪
• Other Authors: Tun-Wen Pai
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/88404174588534092965

碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 95 === Abstract For the goal of rising knowledge-based economy, Taiwan government plans to launch the applications of e-schoolbag and e-learning ideas based on the information and networking technologies. It is the most important item in the e-Taiwan project from the “Challenge 2008”, the Six-year National Development Plan. In order to conform with the national digital eLearning program, we have developed the digital ink document teaching system that is based on the Tablet PC platform. The key feature of the developed system is digitally pen-based computing which enhances the teaching and interactive learning activities among teachers and students. In order to strengthen the features of fluency and accuracy in teaching mathematics, we have proposed an online geometric shape recognition model which automatically transforms a sketched geometric shape into its corresponding digital shape object. This thesis presents geometric models which define various shapes consisting of fundamental line segments. Prior to performing recognition processes, models of individual geometrics must be created and reserved in the system. There are five major processes in the proposed algorithm including pre-processing, line segment extraction, line segment optimization, line segment refined allocation, and feature matching. In the pre-processing step, the digital point data array of the original sketch will be filtered and normalized. The extraction, optimization, and refined allocation processes effectively generate representative line segments, remove redundant segments, and provide the optimal representation respectively. At the last, the pattern matching process selects the best candidate from possible geometrical shapes. In this research, we have chosen eight frequently drawn geometrics and collected abundant testing samples for system evaluation. The results show that the proposed recognition kernel for hand drawn geometrical shapes achieves a recognition rate of 88.87%.