Research And Application Of Testing Technology Based On Thegraph Theory

At present,most of the work of invoice management company or institution or the traditional artificial way.Use the computer to process the most common VAT invoice on the market at present automaticly,and the implementation of the software to save manpower and material resources consumption and capital investment,can improve the work efficiently.On after a scan invoice image binarization processing,tilt correction algorithm is presented.Put forward the invoice image binarization locally adaptive threshold correction method.For image skew correction,the improved method based on direction of the White Run-Length Direction.Due to its own sparse connectivity and can share the weights of characteristic,its train learning difficulty do not add much,but its recognition ability is greatly increased,the parallel connection makes it especially suitable for image recognition.Invoice for the realization of the information identification based on images saved as the medium,the image processing quality directly affect the success rate of recognition.Based on image processing,the elimination of noise of image,image color channel normalization and adjust the image in terms of location,size,after preprocessing images to identify the training sample set as text.Using word complex index and geometric invariants as text features,to this kind of characteristic parameter extraction method for improved.We deeply learn the Support Vector Machine(SVM)in detail,this paper introduces the network structure and deduction process,through the analysis of recognition results to prove that this algorithm than the traditional OCR recognition technology has more superiority in character recognition.