Research On Train Number Detection And Recognition Algorithm

With the development of national economy,the demand for rail transit was getting higher and higher.In order to ensure the safety of the rail transit,it was necessary to inspect the train regularly.The efficiency of manual detection method was hard to satisfy the need of rail transit,so that the automatic inspecting method was needed.To achieve this goal,we should obtain the train number firstly,so as to classify and summarize the results.Aiming at the scene of train maintenance,this thesis studied the positioning and recognition method based on image technology.The main contents included follows.Firstly,the positioning algorithm of train number was studied.Based on the analysis of train number structure and its characteristics,the edge information was selected as the key index of positioning.Since the edge blurring problem caused by brilliant light,a pretreatment method which enhanced the edge was designed.Base on the above,for the character width of train number are approximately equal,an improved stroke width transform algorithm was proposed to position the train numbers.Secondly,a train number recognition method was studied.For improving the speed and accuracy of the character recognition,this thesis transformed the plane image into a one-dimensional curve by skeleton extraction and weighted vertical projection.Based on this,an improved dynamic time warping algorithm was designed to recognize the characters.It greatly reduces the amount of calculation while ensuring the accuracy of recognition.Finally,the algorithm was tested with real train images.The results show that the accuracy of position algorithm could reach 96%,and the rate of character recognize reached 97.78%.