Research And Implementation Of Express List Recognition Based On OCR

With the rapid development of the logistics industry, every day, lots of express document information needs to entry, in order to improve efficiency and reduce labor intensity, it’s an urgent problems to find an automatic way to input the text information in express document into computer.In this thesis, we design and realize a system of express list picture recognition based on OCR for a Courier, the following several aspects of the work were mainly finished:(1) A table line blanking methods was designed and implemented. The lines that exist in express list is seriously affected on extracting text blocks, this method can successfully remove the line in express list picture.(2) Image template matching was using in layout analysis stage, this method can accurately extract the text blocks of document images.(3) In this thesis, we combined projection histogram algorithm, multi-function fitting algorithm with post-merger strategy to design and implemented an character segmentation method based on multi segmentation strategy, and more successfully cut the text block into individual characters.(4) The training and recognition of the two-step was fully separated, and the classifier using offline training, the matrix parameters stored in the file for online recognition after training. The training set is read into memory only once, when the system is initialized, that greatly improved operational efficiency.(5)To improve the recognition rate of the system which uses a post-processing method based on database to calibrate the recognition results.