Method For Numeral Recognition Of Shopping Receipts Of POS Machine

By the improvement of science and the development of society, as the sales documentPOS ticket can not only reflect the level of consumption our country, but also indicate thetype of goods which catch the fancy of consumers, which is important for learning the trendin shopping and the values of goods. Although OCR is mature in our country and effectiverecognition of the shopping receipt is still many problems need us to solve. Every step in theprocess of recognition is very critical and is the decisive factor which influences the result.Collect different shopping receipts as samples and take pictures of the receipts whosesize and length are different by camera or scanner, then algorithms in the process of graying，binarization, tilt correction and smooth，de-noising will be improved, finally select the bestpictures as the sample.In this paper an improved segmentation method based on connected regions will be usedto segment images. In order to facilitate the analysis of experimental data the normalizationprocessing should be applied to the pictures for unifying the characters after segmentingimages. Select the number character in the mixed text by extracting the architectural feature ofthe numeral character and the amounts of pixels of the numeral character.The POS optimization algorithm in this paper is improved parameter optimization anduse optimized parameters to design vector machine classifier and recognize the print digits ofwhich can maximize the accuracy rate of recognition. It is found that the algorithm is thispaper is efficient according to experiment and by comparing with other result and it not onlyhas high recognition rate.，but also practical value.