The Research On High Speed Train Number Recognition Algorithm Based On OpenCV

In the past ten years,my country's high-speed railway has been introduced from scratch,digested,absorbed,and re-innovated.Until today,high-speed rail has become a bright business card made in China.The high-speed railway brings us convenience,speed,and comfort,while a group of people silently guard their safety.As a technology for real-time monitoring of the health status of high-speed trains,TEDS(Trouble of moving EMU Detection System)is a realtime monitoring technology.It can identify which train is faulty by identifying the number of the train,thus establishing the train.One-to-one correspondence between the train number and the fault location.The existing TEDS system relies on the train numbers sprayed on both sides of the train to identify the vehicle,thereby realizing the fault diagnosis of the EMU through automated image comparison.As the sole symbol of the train's identity,accurate vehicle number recognition is the prerequisite for the TEDS system to work properly,indicating the importance of train number recognition in the TEDS system.In this paper,the excellent features of the computer vision technology library(OpenCV)are used to pre-process the image of the high-speed train number data obtained,and then to locate the number,and after the number character segmentation,the automatic recognition of the number is finally achieved.The content of this article is as follows:(1)Train number image preprocessing operation,because of the noise and interference of the obtained Train number image,the quality of the train number image is improved through a series of algorithms such as image graying,image enhancement,and image denoising.(2)The combination of mathematical morphology and projection method is used to accurately locate the high-speed train number.The whole positioning process is divided into two steps.First,the mathematical morphology technology is used to find the candidate area of the number,that is,the rough positioning of the high-speed train number;Then,the projection method is used to accurately locate the train number,and the non-carriage number characters that are disturbed are removed,and an accurate train number image is finally obtained.(3)The projection method divides the train number characters to obtain six separate train number character images.The pixel size of the segmented trian number characters is different,and the segmented characters need to be normalized to finally obtain a train number character image of the same size.(4)Convolutional neural network has a good effect in the field of image processing.The input layer,convolutional layer,pooling layer,and output layer of the convolutional neural network are designed in combination with the character characteristics of the train number.The network structure is optimized in the experiment,and the EMU number can be quickly and accurately identified.The final test results show that the individual recognition accuracy of the ten Arabic numeral train number character images can be as high as 99.8%.