| Summary: | 碩士 === 大葉大學 === 資訊工程學系碩士班 === 105 === This study mainly focuses on the functions of support vector machine which is used to recognize the numerals on the two-copy cash register uniform-invoice. The research point is on the image pictures of two-copy cash register uniform-invoice. By using the technology of image processing, the numerals on the uniform-invoice will be preprocessed in order to analyze and position. Then, based on the support vector machine, we can undertake the comparison and recognition to achieve the targeted effects of this study.
In this study, the invoice image input by the system are all color images. First, these images will be grayed. Then, they are processed by means of to filter most of the noise to present cleaner binary images. By horizontal and vertical projection, the horizontal pixel points and vertical pixel points in the images will accumulate. And the numeral areas of uniform-invoice will be captured by cross-matching. The captured numerals will then be segmented in sequence to be indicated respectively. Because of the serious damages to some captured numerals, they will be inflated and eroded in a uniform way. Accordingly, the support vector machine follows the above process to treat the training pictures. After processing, the result will be put in the support vector machine to capture and classify the characteristics. Lastly, the input uniform-invoice images will be compared and recognized by the finished classified characteristic modules.
|
|---|
