Design Of Train Number Automatic Recognition System Based On Image Processing

The railway transport has play a very important role in the process of our country economic construction.With the continuous growth of railway transportation,the information management of the railway industry should also be put on the agenda.The automatic recognition system of train numbers is an important part of the information management system.The automatic recognition system of train numbers can effectively replace human resources,make up for the shortcomings that are prone to errors when manually reading the number,and save financial and material resources.At present,the most widely used train number recognition system is based on radio frequency technology,but this system has the disadvantage of large hardware loss,a higher identification error rate,and increases the overall maintenance cost.In order to save costs and improve recognition efficiency,train number recognition based on image processing has become a research focus in recent years.The train number recognition system is based on computer vision and recognizes specific train number characters.The system collects the train car number in real time through the front-end camera,transmits it to the computer for further processing of the collected image,and locates and recognizes the train number.The main research contents of the paper are as follows:(1)Introducing the image acquisition system: By comparing the train number recognition system based on RFID technology with the image acquisition system of this article,the image acquisition hardware part used in this article is determined,and the installation position of the camera and the installation of other hardware devices are given in combination with the actual situation to ensure the information of the entire car It is not missed,the car number details will not be lost and clear images can be collected.(2)Image preprocessing: The image pre-processing stage can eliminate the noise in the collected image.By performing gray transformation in different ways on the collected train images,and then selecting the gray transformation enhancement to analyze the experimental results to enhance the effect of the gray image,and then denoising by means of mean filtering,median filtering and Wiener filtering.After analyzing the effect,the median filter was finally selected to denoise the image,which laid the foundation for the subsequent recognition steps.(3)Image binarization: The binarized image can simplify the difficulty of subsequent recognition.Through the experimental analysis of several binarization methods,it is found that the global threshold method is difficult to select the threshold.Although the Otsu method can adapt the threshold,it is not ideal for the image processing in this article.Finally,the Niblack method was used to perform the binarized image has achieved good results.(4)Train number positioning: By performing morphological operations on the binarization of the train number,the maximum position of the connected domain is determined as the position of the number,but it is found through experiments that there are many interferences on the car during positioning and positioning errors will occur.The train mark is also located in the position of the train number,it will be removed before the train number is recognized.(5)Character segmentation and recognition of train numbers: There are widespread character fractures in train numbers.This situation brings great difficulties to character segmentation.Through the experiment of character segmentation,it is found that the error rate of character segmentation is extremely high and poor robustness.The overall recognition of the train number characters will be performed,reducing the error rate of the system and improving the recognition efficiency.(6)Train speed detection based on image processing: processing several frames of vehicle number images that are continuously collected,marking the center position of train numbers in different frames,and fitting the trajectory of the train by the ratio of image pixels to the actual field of view size,Finally,solve the train speed.