Embeded Numeral Instrument Recognition System Based On DSP

This paper is the use of image processing technology to realize the numeral instrumentation automatic recognition. Nowadays, the numeral instrument is widely used in many fields. There are two kinds of numeral instrument, the one is digital numeral instrument and the other is mechanic numeral instrument. Because the function of digital numeral instrument is usual comparatively stronger, auto recognition of this instrument is scarcely any meanings. But for mechanic numeral instrument, which is widely used, simple in manufacture and cheap in price, the adjustment of this instrument in measurement department and factory commonly uses the methods of human eyes or optic-to-electron sampler. The efficiency of these two methods is usually low. Consequently, this paper is aimed at the mechanical numeral instrument for the development of automatic identification system.At present, most of digital image processing systems work by an image capturing card in PCI socket which work unstably. At the same time , the price is high and it can't work in bad circumstance and so on. On the other hand, because of quickly development of GSI and technology of embedded software and hardware, many powerful embedded processors, such as TI's DSP and Motorola's PowerPC, whose performance is higher and higher . But the price is lower and lower. Especially, a DSP chip includes abundance of peripheral interface and has very high programmed ability, which make the hardware and software of image processing system be implemented easily. But also, this embedded system conquers many deficiencies of the system based on PC. As a result of the increasing performance-price-ratio, adding in small area, low price and power consumption, embedded chips in the field of image processing occupies a place gradually.In this system, the multimedia processing chip TMS320DM642 which is made by TI corporation is applied to implement the multi-instrument numeral character recognition. This system can capture 4-route video signals at the same time, which cuts down the cost effectively. The captured data are processed and stored into local stored RAM. Firstly in this paper, the system is projected. Secondly, the hardware of system is designed, including power and monitor circuit, clock circuit and keyboard circuit ect. Thirdly, the software of system is programed, including boot-loader of system, distribution of RAM, initialization of DM642 , microcontroller(89C51) programming code , BP neural network arithmetic and so on. Lastly, this paper discusses the practice result and the prototype model is exhibited .