摘要
在计算机评卷过程中,通常方法是将试卷扫描成灰度图像,由评卷人员对扫描图像进行评卷。有些试卷是空白试卷,可以通过计算机识别后,直接给分。在一些情况下,很难直接用像素灰度值来区分空白试卷和非空白试卷。研究表明,像素灰度值的列向量或行向量的标准差可以将空白试卷的特征表现出来。神经网络具有自组织、自学习、非线性逼近能力,应用神经网络可以有效地识别出空白试卷。为了便于神经网络进行空白试卷的识别,减小神经元的数量,可将图像像素灰度值列向量标准差的标准差和行向量标准差的标准差作为区分空白试卷和非空白试卷的影响因子。
In the process of grading examination paper by computer, the common method is to scan the paper into gray - scaie image, and let grade staffs give the paper a mark. Some examination paper are blank examination paper, can directly give the examination paper a mark by computer recognize. But in some condition, can hardly use pixel gray value to distinguish blank examination paper and non - blank examination paper. People can availably recognize the blank examination paper by making use of the neural network, which is able to do self- organization, self- learning and non - linear approximation. Experiment expresses the influence factor classifying the blank examination paper should have： the standard deviation of standard deviation of image pixel gray value column vector and the standard deviation of standard deviation of image pixel gray value row vector.
出处
《计算机技术与发展》
2009年第8期208-211,共4页
Computer Technology and Development
关键词
空白试卷
识别
神经网络
blank examination paper
recognization
neural network
作者简介
贾志先（1958-），男，山西临猗人，教授，主要研究方向为神经网络、算法、计算机安全。
