Research On Handwritten Numeral Recognition Method Based On Imaging Technology

With the development of era, Increasingly developed science and technology, National trade has become increasingly close。Due to the close of trade, which led to the financial markets and prosperity。Essential is the notes in the process of national trade exchanges, Countless number of notes are processed daily, If these bills to deal with artificial virtually impossible to imagine。So, When people deal with these data the process will think of a problem。That can use the machine instead of manual? This produced a handwritten numeral recognition—Pattern Recognition; Handwritten numeral recognition is a very wide application in the trade market, countries in the course of trade, that hope haveintelligent tools for recognition of handwritten digits。 Scholars from various countries study handwritten numeral recognition for a very long time, Also achieved great success。Put forward a number of good algorithms, but so far, No algorithm can be accurate to recognize handwritten digits, there is main digital calligraphy varies from person to person, a variety of forms。In the past for a long time, Countries have experienced the bill management which is very backward, all kinds of recognition of handwritten data。Digital printing and archiving cost a lot of manpower, to change this situation has drawn increasing national attention using intelligent to manage。Handwritten numeral recognition is one of an important research topic in the character recognition。Although the types of numbers have only ten kinds, and strokes but also simple, seems to recognize that is not very difficult; however, in practice, some test results showed that, recognition accuracy of the numbers smaller than the correct rate of printed Chinese character recognition。Handwritten numeral recognition, the main difficulties is that its many variants, and digit recognition accuracy rate demandi much more than character accuracy rate。For example, in the data, if the1identified as7; although only a number of different, the results are completely different, resulting in great loss。Therefore, the study an efficient handwritten digit recognition system occupies an important position in the pattern recognition。 This thesis studys some of single handwritten numeral recognition problems, such as grasping digital image、feature extraction、feature selection、digital identification of and so on research。Digital image preprocessing:digital image of the gray、binarization of handwritten digits、smoothing of the handwritten numeral、 digital image correction、remove image interference part、digital image normalization、 digital image thinning。The core of this article:1、using SVM to classify the numbers, this can be a good distinction between the main features of each digital、only by grasping the main features of each digital, In favor of feature extraction and feature selection of handwritten digits;2、Image acquisition use video instead of the scanner, this paper thickness, color, and a pen of handwritten digits have no special requirements; when video capture image, we have taken a VFW-based software development package, because of using development kit very extensive in the images, audio application, It encapsulates a large number of video library functions, In the preparation process, as long as the understanding of the use of the library functions do not understand the function implementation process。3、when get a video capture of the image, we must pretreat it, the purpose of pretreatment is to get a clear picture, it is a good to get the main features of the digital, to improve access to digital recognition rate。4、establish the recognition of handwritten digit image-based technology, the use of template matching algorithm。Experiments show that2,3,5,8, the figures are accuration identification1,7,0,6,4,9,digital basic identification and accurate, The method improves the recognition rate of digital and real-time。...