The Research&Application Of The Character Recogniton Algorithm For The Invoice Date

With the rapid development of second-hand car industry in China,the major car manufacturers have opened a second-hand car business.The used car businesses use second-hand car information management system to manage car sales information,date information needs to be stored in the system.In the past,the conversion of image character information to text information was manually entered,which not only increased the workload of operators,but also was inefficient and prone to make mistake.Therefore,it is significant to study the auto identification of character information in image.In this thesis,for the specific needs of date recognition in the machine-printed second-hand car sales invoice which proposed by Pui Information Technology(Beijing)Limited,the key method of date character recognition was studied and the related business system was developed.The main work includes the following aspects:(1)According to the problem of invoice date character location,concrete analysis of the invoice layout and date character position.Finally,we use a location algorithm which based on Hough transform line detection,the invoice date character can be positioned effectively.(2)In the character image pre-processing,according to the characteristics of invoice image use the weighted average method for gray;use the simple statistic method and the maximum between class variance method for binary;use median filtering algorithm for image denoising.In addition,according to the system needs use statistical projection method to cut away unnecessary portion,finally obtain a bitmap only contains date characters.(3)According to character segmentation and extraction problem,put forward the adaptive character segmentation and extraction algorithm,the average correct rate of which reaches 99%;In order to solve the date character recognition problem,use the template matching method for character recognition.Analyze and judge the final recognition results.Apply the above algorithm,using VC6.0 to design and develop the second-hand car invoice date character recognition subsystem.At present,the system has been delivered for use,for scanning acquisition car invoice recognition correct rate reaches 85%,and the performance is stable,well meet the business needs of the system.