摘要
为提高运维效率,针对视频中的分会场文字信息,采用计算机视觉技术,识别出分会场的名称,以便实现轮询视频的自动检测。提出一种基于隐马尔可夫模型(hidden Markov model,HMM)的轮询视频分会场名称识别算法,利用分会场名称中相邻单个文字的相关性,提高分会场名称的识别准确率。对每帧视频图像,采用微分二值化(differentiable binarization,DB)算法定位文字区域,提取单个文字的分块特征,并通过计算欧式距离进行单字识别。考虑分会场名称中相邻文字之间的相关性,构建HMM,实现相邻文字之间的关联,并采用Viterbi算法计算分会场名称识别结果。试验数据表明,在采用较低维数的特征向量时,本研究提出的分会场名称识别算法具有较高的识别率和较强的抗噪性能。
In order to improve the maintenance efficiency,aiming at the character information of branch venue in the video,the technique of computer vision was applied to identify the branch venue name so that the polling video could be detected automatically.A recognition algorithm of the branch venue name based on hidden Markov model(HMM) was proposed for polling video,which could make full use of the correlation between adjacent single characters in branch venue name to improve the recognition accuracy.For each video frame,differentiable binarization(DB) algorithm was used to locate text region and achieves single character recognition by extracting block features of single character and calculating Euclidean distance.Considering the correlation between the adjacent characters in branch venue name,a HMM was constructed to realize the association of adjacent characters,and Viterbi algorithm was used to calculate the recognition result of the branch venue name.The experimental data showed that the proposed algorithm achieved high recognition rate and strong anti-noise performance when the dimension of the feature vector was low.
作者
何子亨
孙丽丽
左修洋
刘鸿雁
王雨晨
车四四
王朔
HE Ziheng;SUN Lili;ZUO Xiuyang;LIU Hongyan;WANG Yuchen;CHE Sisi;WANG Shuo(Information&Telecommunications Company,State Grid Shandong Electric Power Company,Jinan 250001,Shandong,China;School of Information Science and Engineering,Shandong University,Qingdao 266000,Shandong,China)
出处
《山东大学学报（工学版）》
CAS
CSCD
北大核心
2022年第6期183-190,共8页
Journal of Shandong University（Engineering Science）
基金
国网山东省电力公司科技资助项目(520627210004)。
