一种基于决策树的选项识别方法  

A method for option recognition based on the decision tree

作　　者：雷俊杰 张伟池 余荣[1] 张浩川[1] 

机构地区：[1]广东工业大学自动化学院,广东广州510000

出　　处：《无线互联科技》2018年第1期113-115,138,共4页Wireless Internet Technology

摘　　要：文章将机器学习中的决策树算法和图像处理技术相结合,提出了一种基于决策树的选项识别方法,该方法首先需要通过人工标注的方式从答题卡中抽取选项构造训练集和测试集,训练集和测试集都包括填涂的选项和未填涂的选项两类,接着将训练集中的答题卡选项切割成n个大小相同的小矩形,通过计算这些小矩形的占空比并通过设定阈值的方式将其离散化成{0,1}中的其中一个值,这些值将作为选项的填涂空间信息特征,然后将n个小矩形的离散后的值相加作为表征选项整体填涂信息特征,再将这n+1个特征构成特征向量的形式,去构造选项识别的决策树,最后,用测试集测试决策树的准确率和速度。经过仿真测试,在权衡识别准确率和识别效率之后,得出选项切割的最佳个数和最佳离散化阈值,在该参数的设置下,决策树的识别性能具有满意的结果。该方法实现方便、简单、易于理解,并具有很高的准确率和很快的识别速度。In this paper, we combine the decision tree algorithm in machine learning and image processing technology, and propose a method for option recognition based on decision tree. This method firstly needs to extract the training set and test set from the answer card by manual annotation. The training set and the test set include both fill-in and unfilled options, and then cut the answer-card option in the training set into small rectangles of the same size. By calculating the duty cycle of these small rectangles, the value is discretized as one of the valuesin {0, 1}. These valueswill be used as the fill space information characteristic for the option, and then add the discrete valuesof the n small rectangles as the characterization options. And then the n＋l features constitute the form of eigenvectors to construct the decision tree of option recognition. Finally, the test set is used to test the accuracy and speed of the decision tree. After the simulation test, and weighing the recognition accuracy and recognition efficiency, we get the optimal number of cut and the best discretization threshold. Under the setting of this parameter, the recognition performance of the decision tree has satisfactory results. The method is convenient, simple, easy to understand, and has high accuracy and recognition speed.

关 键 词：机器学习 决策树 选项识别 特征提取 答题卡 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]