The Design Of Image Processing Plateform Based On Embedded FPGA

With the development of digital image processing technology, industrial image processing platform has been widly used, more and more companies take advantage of industrial image processing platform for monitoring and testing. With vigorous innovation of production equipments in tabacco industry, these equipments have realized continuous and high speed flow line production, and these automatic deveices can not avoide making errors due to different reasons. Cigarette sealing devices may miss cigarette in the processing of sealing. This will greatly damage the reputation of the tobacco factory and finally result unpredictable economic loss, so measures must be taken to solve the problem. It is good to know the number of the cigarette in the box in real time, because people can take effective measures. Artificial statistical methods can make real-time detection of the cigarette number in the box, but the quality and efficiency are relatively low. This design of image processing platform in the paper can guarantee the quality and efficiency.The design of the image processing plateform consists of three parts,the first part is a Gigabit Ethernet camera, the second part is an FPGA-based image acquisition and pre-processing board, and the third part is the PC for further image processing and image saving. In the board part, this paper design a SOPC using SOPC Builder based on FPGA, then the user software using Nios II IDE with C language, and the software of the image preprocessing. In the PC part, this paper design the software of image processing using C language, using a custom image enhancement algorithms to process the image data, and an adaptive threshold for image binarization based on local standard deviation, using SSE method and the method to reduce the amount of computation to optimize the local method to strike the local standard deviation of the whole image. The PC image receiving thread can receive each frame sent from the board based on FPGA with (ram lossing rate of zero percent, and the image processing algorithms designed can accurately detect the number of cigarette in the box, the rate of detection error is as low as 0.016%.The image processing platform designed in this paper can not only be used to detect the number of the cigarette in the box, but also can be applied in other areas with some modifications according to the specific requirements.