手寫紙本是非及選擇題答案辨識系統之實作

• Main Author: 林瑞明
• Other Authors: 陳桂霞
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/58852950024820169170

碩士 === 臺中師範學院 === 教育測驗統計研究所 === 90 === This dissertation is based on the theory of digital image processing and Matlab software to develop a recognition algorithm for the answers of paper-based handwriting Yes/No and choice questions. Finally, the target of this dissertation is to design an answer-sheet reading system. . Regarding to the recognition system for handwriting Yes/No and choice questions, it is utilized digital image processing to implement digital preprocessing, including noise reduction, binarization, cutting line processing, area maximum, and smoothing, etc. After preprocessing the answers for Yes/No questions, the images are thinning, and then the number of end-points is counted as the feature for image recognition. In the aspect of the answers for choice questions, it is according to some structural features to recognize the answers, such as the judgment of whether the images of characters include looping, and the vertex number of left/right indent of character images, plus the number of end-points after the thinning of character images. . In terms of physically designing an answer-sheet reading system, the GUI design allows the system to be used easily and user-friendly. Also, the system capability of processing images is enhanced with image toolbox of Matlab, and its powerful matrix computing. Finally, in this dissertation, An answer-sheet reading system based on the recognition algorithm for handwriting answers of Yes/No and choice questions is developed and this system is sometimes need to be helped by human when the result has some troubles. . Via the experiments in the elementary primary school, it is higher than 98% recognition rate for recognizing students’ pencil-writing answers for Yes/No and choice questions. The system also has apparently high tolerance toward difficult-recognized and slanted characters.