The Research And Implementation Of Paper Currency Number Acquisition And Recognition System Based On DSP

With the development of society, the paper currency, checks and other tickets which represent the Massive wealth, get more and more widespread application in people's life. These tickets bring us great convenience, but they also bring a great deal of insecurity, the events of forge and violence rob these tickets occurred unceasingly. Embedding the number recognition system into the counterfeit detector will help us record the serial number while detecting the currency paper, which will provides an effective way to identify the authenticity of these tickets and detect and prevent the robbery with violence. Moreover, recording the numbers of these tickets can help the country's financial sectors analyze the flow direction of these tickets, which will provide valuable and reliable information to help them make the correct policies.In view of the system's complexity, high-speed demands, the hardware platform based on DSP and CPLD is designed. This paper has divided the system's functional module and designed the system's working flow. A CIS sensor is chosen to acquisition signals. The peripheral circuit of CIS and the AD convert circuit is designed This paper designs the system's core circuit, which accomplishes the signals' reading and processing with the DSP as the core device. In view of the DSP memory space features and the insufficient I/O port features, this paper extends the DSP memory and the I/O port. This paper designs the system's timing and logical control function with CPLD as the core device, making the timing of CIS and AD matched accurately, enabling the DSP access the extend memory and collected the digital signals correctly, making the system work coordinately. The hardware platform can effectively achieve the image signals, which establishes the good foundation for the latter recognition.In the number recognizing part, this paper makes filtering and binarization and other preprocessing to the collected signals. According to the features of the image after binarization,this paper proposes an image segmentation method which is realized with line scanning and row scanning. A large number of experimental results show that this method is simple and effective. The character of the feature extraction methods and its impact on the characters recognition is analyzed, and chose the invariant moment, coefficients of wavelet transform, coefficients of Curvelet transform, the contour features as the feature of the characters. A comparison and analyze to these features is made with experimental data. At last the classifier is designed, and the characters of the BP neural network and the SVM is studied respectively. The classification to the four type features with these two classifier is made. After comparing the experimental data, this paper get the result: with the Curvelet transform as the feature extraction method and the BP neural network as the classifier, we can get the higher recognition rate. The final experimental results show that the software developed for recognizing the paper currency numbers can obtain a higher recognition rate.This paper complete the construction of the hardware platform of the number recognition system, and the quality of the signals collected from the hardware platform is good enough to meet the requirements of the subsequent recognition processing. The result of the recognition part is also very well. The experiment results show that the design of this system is successful.