The Research On The Structure Of Card To Machine And Image Recognition And Counting System

With the rapid development of the socialist market economy, market circulation of banknotes needed for the growing. The notes handing of bank is very heavy. But the notes counting is still using manual counting currently. This method is labor intensive, low efficiency, and the accuracy and security are difficult to ensure.In response, Develop an automatically count on the note detection system can improve labor productivity, chang the traditional bank staff " eye of mental arithmetic operators" of the counting mode, free the bank staff from tedious manual and dry environment, make the banknote counting mechanization and automation. There is an urgent need for a status quo with China’s banking sector, to achieve fast and accurate automatic banknote counting system.The so-called note cards to count, is to identify the number and bagged bundles of bank notes to numbers; The so-called card to machine, is not open bags of paper money in the pocket of the case, symmetry by X-ray images scanned notes, and then use some algorithm notes tied to100%counting equipment. The ultimate goal is to identify X-ray scanning and statistical notes onto the imaging agent binding stripe imaging to determine the number of notes tied to the number and the bundle number.The main contents include the structural design and improvements, the development of imaging agent, the notes image preprocessing, the notes image recognition and counting in the ideal and actual condition.The notes image preprocessing include the quantitative increase gray and the threshold. For the characteristics of the topic image, comparing of several common image thresholding method, this paper is based on histogram statistics combined with the addition operation method. This method is simple and the results are very good. For the notes image recognition and counting notes in the ideal condition, using image connected component labeling can recognize and count. But in actual working conditions, as the paper money binding machine can also be a manual binding.Manual banding in the image will have tied stripes. And because the display of arbitrary, knotted fringe can appear anywhere in the image,or the imaging agent wear so that lack of stripes in the image. This will certainly affect the accuracy of counting system identification. In response, according to the characteristics of the image itself, This paper presents an image centroid distance between adjacent target stripe and the edge of the image notes outline the combination of multi-information fusion algorithm, to make up for missing the stripe, so the results are still accurate count.In addition, the article also notes placed on the Double A theoretical analysis for the card after the upgrade the computer system functions to do theoretical preliminary research.