Research And Application Of PCI-E HD Video Capture System Based On FPGA

Video capture system has been applied in machine vision widely. Because the high-definition video stream is continuous and large in amount, traditional capture system used SDRAM cache to avoid dropped frames, which increases system complexity. This paper presents a scheme of a non-SDRAM capture system by improving video streaming mechanism,which simplifies the hardware architecture of system;completes the construction and test of improved capture system; studys deeply of the image enhancement and noise reduction algorithms to enhance image quality. The main contents are as follows:1.Overview the background and significance of this research. Explain the present development situation of video capture and PCIE technology, and put forward the main research content of this article;2.Analyze the traditional video streaming mechanism, propose a improvement strategy by constructing double commonbuffers and demonstrate through the time of frame processing. Analysis of the overall system architecture on this basis and propose a design scheme which sets FPGA (EP2C5Q208) as the core control chip and PEX8311 as the PCIE interface chip;3.Design and complete the system hardware circuit which include the local bus control modu、PCIE interface module and peripheral configuration module. Propose a DMA transmission scheme by hardware and give optimization measures;4.Design and complete the logic control in FPGA by Verilog language, which include the data acquisition module, the cache module and interface chip communication module. Made simulation of these modules with Modelsim. Test the speed of DMA transmission, resulted in 120.5MB/s which meets the system requirement and study of the the DMA transfer characteristics;5.Based on the deeply study of WDM model, complete the key driver routines and the upper application which collects the video stream by ping-pong operation of the double commonbuffer and displays in real-time by DirectDraw technology. After testing, the system captured the uncompressed HD video stream(1080P,30fps) with on SDRAM;6.For the defect of small dynamic range of images, study deeply in image enhancement algorithm based on Retinex theory and improve the gray-mapping algorithm. Propose a filtering algorithm based on VMF algorithm and actual noise characteristics, which eliminates the stripes noise in YUV color images effectively and dramatically increases in efficiency.