本研究提出整合各種技術,發展新一代電腦閱卷系統的方法。新一代的電腦閱卷系統利用TWAIN協定作為個人電腦與掃瞄器或影印機之間的通訊介面。它可以自動對輸入的試卷影像作分析與計分。在此架構下可以改進傳統電腦閱卷系統的許多缺點。本研究結果整合的各種技術包括表格傾斜度校正、表格與文字分離、文字的特徵擷取與比對等。本研究結果整合各種技術的基礎在於充分利用輪廓鏈碼的特性,姿展出許多簡單快速的演算法,其中包括輪廓鏈碼形成演算法、鏈碼方向性判定、鏈碼特徵抽取與比對演算法等。本研究結果所提出的鏈碼特徵抽取與比對演算法是在離散系統中運算,沒有轉換成連續函數的困擾,藉由簡單的關係與邏輯運算,即可得到結果。若鏈碼上的點數為N,其鏈碼特徵抽取演算法的時間複雜度則為O(N)。本研究所提出的電腦閱卷系統,經過採用真實考試的試卷與模擬考試的試卷實驗後,證明本系統具有高達99.99%的閱卷正確率,同時具有將近每秒一張的快速閱卷速度,由此可知,本研究之結果具有高的實用性質。
The paper presents a method for integrating all technology to develop new generation computer scoring system. This new generation computer scoring system use TWAIN protocol as communication interface between computers with scanner or copy machine. The system can do paper image analysis and scoring automatically. Under this architecture, we improve the weakness of traditional computer scoring system. The outcome of this study integrates the calibration of the skew table, text/table separation, and character features extraction/comparison, etc. The basis to integrate all the technique of the result of this research is to use the characteristic of contour chain code to develop many simple and fast methods. These methods include contour chain code building algorithm, contour chain code direction decision algorithm, and chain code features extraction/comparison algorithm. This chain code features extraction/comparison algorithm of this research is to be computing in discrete system to avoid some trouble from continuous system. The algorithm can get by simple relational and logical operation. If the amount of the nodes of the chain code is N, then the time complexity of the chain code features extraction method is 0(N). This computer scoring system experimented with real test and simulated test papers. The correct rate reaches 99.99% and the scoring rate is almost one paper per minute. According to this result, the practicable ability of the outcome of this research is very well.