Research On Ticket Information Recognition System Based On Machine Vision

Image ticket information recognition technology plays an important role in ticket,movie ticket and invoice recognition,there is a wide range of practical requirements for the study of ticket image face information recognition technology.Aiming at the low efficiency of the traditional ticket checking method in railway service,which makes travel difficult for the vast number of passengers,this paper puts forward a fast and high recognition method for the receipt information identification.Firstly,this paper introduces present research situation of the key technology of image recognition system,On this basis,the algorithm is studied through image correction,binarization,location and segmentation of ticket information area and ticket character recognition in ticket image information recognition.Finally,the algorithm is implemented as a set of ticket information recognition system.The following are the main research contents of this paper:(1)In the research of the image direction correction algorithm of tickets,a direction detection algorithm based on the Radon transform in the clustering of feature areas is proposed.According to the simulation experiment,the algorithm can detect the direction of the train ticket image at any angle,and eliminate the interference of pixels in the complex background of the image and the direction feature area to the direction detection.The correct correction rate is 97.8%.(2)In the research of ticket surface information area location and extraction algorithm,a new algorithm based on global key information search is proposed to locate and cut the ticket area of interest.Based on the difference in the waveform of grayscale value of ticket row vector,the vector of the image whose grayscale value is 1 is differentiated by global search,and the longest segment whose derivative value is zero continuously is the key information of the image.Combined with the trial and error method,the threshold interval was determined and the region of interest of the ticket was extracted.Through the experiment,the accuracy of extraction is 98.86%.(3)In the research of ticket information recognition algorithm,the BP neural network algorithm based on variable learning rate is adopted.Through the simulation experiment of1396 characters available in 91 tickets,the experimental results show that the BP neuralnetwork algorithm with variable learning rate has better recognition performance in the identification system with a large sample size.(4)In the application of the system,a GUI visual interface is designed.Through the setting of different buttons,the purpose of human-computer interaction is realized.The paper realizes the ticket information recognition function of intelligent cloud service robot "xinjia",and emphatically introduces the application of ticket recognition algorithm in it.Through the above research,a set of ticket face information recognition system based on machine vision has been implemented and applied to the intelligent cloud service robot "Xinjia".After repeated debugging,the ticket identification accuracy reached 93%,and the identification speed was within 2~3s.It can meet the requirements of the daily ticket identification system for the identification accuracy and speed.The entire system works stably,and has a high recognition rate,and has a high practical value.