基于先验知识的里程碑事件时间识别方法  

Milestone event time identification method based on prior knowledge

作　　者：吕宗磊[1,2] 姬婷婷 LYU Zong-lei;JI Ting-ting(Information Technology Research Base of Civil Aviation Administration of China,Civil AviationUniversity of China,Tianjin 300300,China;College of Computer Science and Technology, Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学中国民航信息技术科研基地,天津300300 [2]中国民航大学计算机科学与技术学院,天津300300

出　　处：《计算机工程与设计》2019年第6期1696-1700,1733,共6页Computer Engineering and Design

基　　金：国家科技支撑计划课题基金项目(2014BAJ04B02)

摘　　要：传统手写数字识别方法在实际应用中没有使用上下文信息,针对机场里程碑事件中的手写数字识别问题,提出基于先验知识的里程碑事件时间识别方法。在传统概率神经网络的基础上加入筛选层,删除不符合时间数据特点的结果,利用机场里程碑事件发生时间的先验知识对识别结果进行选择。在真实的里程碑事件发生时间的识别实验中,所提方法对手写体数字的识别正确率可以达到99%,而BP神经网络对手写体数字的识别正确率仅94.5%,实验验证了所提方法的有效性。In the practical application of handwritten numerals, traditional methods do not use context information. In the view of the problem of handwritten numerals recognition in airport milestone events, a method of time recognition of milestone events based on prior knowledge was proposed. A screening layer was added on the basis of the traditional probabilistic neural network, and the results that did not meet the characteristics of the time data were deleted, and the prior knowledge of the airport milestone event was used to select the recognition result. In the recognition experiment of real milestone events, the accuracy of the proposed method is 99%, while the recognition accuracy of handwritten digits in BP neural network is only 94.5%. Experimental results show the validity of the proposed method.

关 键 词：手写数字识别 神经网络 先验知识 机场里程碑事件 概率神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]