Study On Machine Vision System Based On DSP

The image real-time processing is necessary in the industry examination based on machine and model identify. The system based on PC is not real-time, but expensive and it can't work smoothly in bad environment. In order to solve these problems, TMS320C6711 is used to process the image, a machine vision system of based on DSP is developed. After studying the history and status of DSP, a solution is proposed, which is based on "Image capturing -core DSP system "integrity scheme, breaks through the limits of the traditional equipment based on "Image Acquisition Card-PC-Execution device "computer vision processing to realize the optimization. The thesis begins with the development of the small system based on DSP. The most simplified DSP system, which focuses on the good stability and high efficiency according to the performance indexes, is composed of the interface of EMIF, extended E2PROM,SDRAM,SBSRAM outside. The real-time image capture and transmitting are realized in the methods of interrupt and DMA. The design of the electric power and the clock circle are introduced and the measures of resisting interference are presented. The image capture module of the system is based on SAA7111+CPLD+MCU, the system is controlled in such a way, that using the SCM simulating I2C to control SAA7111 instead of using the current I2C bus, so the system, holding more interfaces, is more flexible and improved in the ability of controlling the whole system. CPLD controls all the time serial and logic of all the chips, it also cut out the images according to the real need. The mechanism of adopting rotation working between two frames spares the instruments of FIFO and cuts the cost. The address storage method of "Cross-Continue"facilitates the DMA transmission of DSP in the way of the whole block. CPLD sends the digital signals transferred from the standard video signals to SRAM, and then transfer them to DSP, which processes the image through interrupt and DMA. In the following part, it introduces the loading software of the system, which can be accomplished on-line in DSP without other facilities, the internal logic controlling program of CPLD, the SCM dummy program of I2C bus. And finally, the optimized though of the software is presented linking the hardware structure of DSP. The above solution can meet the requirements of the system algorithm. A less-cost prototype can be built quickly if assort to output instruments. It is convenient to realize the objects automatic tracking system or fingerprint recognizing system easily, just changing the former sensor and programming the internal image processing software.