Monetary Recognition Based On Statistical Theory Study

With the development of economic and society, increasing the number of the products of currency recognition has been used, playing more important role in work and life. To adapt the new Anti-counterfeiting technology and more advanced counterfeit technology, the characters of real currency should be all-round studied, and the multi-sensors system should be used in currency recognition.A new molde referred in this thesis is uniform and systemic for describing all character of currency. This molde is based on the attribute of currency obying normal distribution. Some alogrithms using probability fuction is proposed for comprehensive utilization of character of currency. Them offers the theoritical basis and bussiness framwork for commercial currency recognition products.First, the major background and progress of currency recognition is briefly introduced. And then, based on theory of normal distribution, vector norm and operator of sets, this thesis proposed a way that can evaluate the validity, precise, independence of sensor and sensor's attribute for distinguishing the diverse currency. This way offers the subject creation for the selection of sensor. The model of currency save the information of currency attribute by storing the parameter of probability function, which can effectively store the information of currency, decrease the memory for storing information and improve the speed of searching. Based this model, "the algorithm of recognition currency based on probability", "the algorithm of recognition currency based on distance", "the algorithm of recognition currency based on information fusion of multi-sensor" and "the algorithm for recognition locomotory currency" is proposed. The advantages and disadvantages of these algorithms and the reliability of algorithms are studied in this thesis. Finaly, the experiment is depicted for verifying the hypothesis that the characters of currency obey normal distribution.