Research On Bus Station Information Recognition Based On Computer Vision Technology

As an international metropolis,the popularization of artificial intelligence and digitalization is the only way for Shanghai to develop into a smart city.However,inefficient management models are still used in many aspects.For example,buses,which are important public transportation vehicles,may have various abnormalities on the electronic screens at each station.At present,manual inspections are almost required to find these abnormalities.Each station is far away,manual inspection is inefficient,and it is difficult to detect problems in time.In this paper,computer vision technology is used to detect and analyze the electronic screen of the bus station,so as to automatically find a variety of abnormal phenomena.The main research contents of this paper are as follows:（1）It realizes the automatic detection of the site screen with black screen exception.Firstly,the pixel distribution of the screen area under various conditions is studied by using the traditional pixel-based image analysis method,and the calculation formulas of the black screen probability of the small screen and the large screen are summarized respectively.Then the formula is used to pre-classify large screen pictures to obtain a dataset,and the loss function of Inception V3 is improved.Then the classification model of black-screen pictures is trained through the improved Inception V3.Finally,the model is applied to realize automatic detection of black screen anomalies.（2）It realizes the automatic detection of glass breakage of screen and advertising screen.Firstly,the broken glass image is made and taken by Crawler script and Physical props,and then the data set is enhanced by various methods.The residual module of YOLOv5 is also proposed by two-way Atrous Convolution.Finally,the model is applied to realize the automatic detection of broken glass anomalies.（3）It realizes the automatic analysis of the site screen where the screen is stuck.Firstly,the picture facing the screen is obtained by preprocessing methods such as perspective transformation,and then the text information on the screen is obtained by applying text detection and text recognition technology,and then the text is structured in combination with the position information to obtain the bus stop information.Finally,the screen stuck judgment mechanism is customized for analysis,and the automatic detection of screen stuck abnormality is realized.