Design Of Intelligent Filtering System For Program Delivery Of Roadside Display Terminal

With the rapid development of China’s road traffic construction industry,roadside display devices have also developed from indicator to today’s intelligent display terminals,and the subsequent information security issues have also made information filtering technology a hotly discussed issue in engineering and academic circles.On the roadside intelligent display terminal,the efficiency and accuracy of the safety detection of the program information is an important criterion for measuring the safety of the roadside terminal.Most of the traditional security testing methods are manually audited.This method requires a certain amount of manpower and material resources to maintain,and in terms of accuracy,certain errors will occur with the working status of the auditors.Therefore,the research on the intelligent filtering system of program information is the threshold that the roadside terminal suppliers can’t get around.The program of the terminal is mainly composed of picture programs and text programs,and involves two fields of image processing and text analysis.This article focuses on two aspects: the extraction method of text information of picture programs and the filtering technology of sensitive information of text programs.First,the method of locating the text in the picture program is discussed.The traditional maximum stable extremal regions(MSER)algorithm and non-maximum suppression(NMS)algorithm are used to complete the initial positioning of the text.In view of the noise information generated after positioning,a binary classifier based on a convolutional neural network is introduced to complete the determination of whether the initially located area contains text,and all single text areas in the picture are obtained.The production process of the data set used to train the network is introduced,and the regional effects of each step are analyzed experimentally.Then,the method of extracting text from a single text area is discussed,and a firstlevel Chinese character classifier based on convolutional neural network is proposed to extract text.The data set required for training the network and the pre-processing of the data set are introduced.The experiment analyzes the effect of the model on text recognition after a certain number of trainings.Finally,the method of detecting sensitive words on text strings in text programs is discussed,and a sensitive word library based on a dictionary tree is proposed.The Chinese word segmentation technology is used to split the string into short words and enter the dictionary tree for matching.By processing the dictionary tree Experiments on the efficiency of sensitive words to verify the real-time performance of the sensitive word detection module built using a dictionary tree.