Bank Note Numbers Recognition Algorithm Research Based On MCU

According to the uniqueness of banknotes numbers, notes number identification modules can be installed to a currency count machine, in the course of counting the bill marking each banknotes by the numbers, which records the whereabouts banknotes in circulation, finally, using of the Internet to achieve follow-up notes. This will not only serve the fight against financial crimes, safeguarding public role, and this number identification devices with the new type of currency count machine also has substantial commercial value and prospects for future development. This article has completed a set of system that based on SCM for notes numbers' collection, identification, storage's design and implementation. It mainly studied the SCM application for image segmentation, positioning and recognition algorithm. Since the goals of the very limited resources, it can not check errors when processing, unable to verify the reasonableness of the method. In accordance with the restrictive conditions of the objectives, we have used C + + Builder to simulate the algorithm, to certificate the feasibility of image processing and identify algorithm, finally transplant to MCU for offline operation. Completed the system hardware design, circuit connections and realized the module-to-module and inter-module to PC intergovernmental communications.System mainly consists of three parts. Image input devices (image sensor), image processor (MCU) and the results output devices (USB). The design consists of hardware and software aspects. First of all in terms of hardware through multi-comparative we choice CMOS image sensor OV7620 as image acquisition devices, the C8051F120 microcontroller as a high-speed processor, Cypress's SL811HS chip as a USB Host / Slave interface control module. The modules through the I~2C bus to transmit and control dates.On software aspect in the understanding of input device performance using the sensor's open window function to absorb the goal picture. Do A/D conversion, and other processed into the processor identification. After image segmentation, the pretreatment values, and so presents, point an algorithm by calculating the number and coordinates of pixel to determine the characteristic after rotation and then entered the different categorized branches for determining the current number. Finally, adopt of USB interface to output recognition results by the text mode.After extensive experiments proved that the design is reasonable to meet the embedded identification number identification devices' requirements to speed, recognition rate and cost-price. Prototype processing speed of 600-900 piece/ min, the recognition rate reached 96.5%, hardware materials of the target system costs about 300￥, performance and cost indicators have reached the design requirements.