Research On Method Of Real-time Video Edge Detection Based On FPGA

As a vital technology of image processing,edge detection plays a pivotal role in the fields of target extraction,image segmentation and other fields.In the process of real-time video edge detection,the accuracy and real-time performance of edge detection algorithm are highly required.Field Programmable Gate Array(FPGA),with highly parallel computing capability and pipelined structure,can achieve high-speed processing of a great number of image data,ensuring real-time image processing so that FPGA attracts much attention in the field of edge detection.In order to solve the problems of weak noise immunity,low detection accuracy and wide edge of traditional Sobel algorithm,an improved Sobel edge detection algorithm is proposed in this thesis,and a real-time video edge detection system is designed based on Artix-7 XC7A35 T FPGA.This thesis focuses on the realization of the improved Sobel edge detection algorithm based on FPGA,and introduces the design scheme of each module of real-time video edge detection system using FPGA.In view of the weak noise immunity of traditional Sobel edge detection algorithm,median filtering is used in this thesis to eliminate the noise in the image before edge detection.The results show that the edge detection algorithm with median filter can effectively filter the noise in the image while preserving the edge.Aiming at the low accuracy problems of the traditional Sobel algorithm,such as false edge,edge loss and edge width,this thesis extends the edge detection direction from horizontal and vertical direction to four direction and eight direction,and introduces the non-maximum suppression module and the adaptive threshold module.The results show that the detection accuracy of the optimized Sobel algorithm is improved obviously.The FPGA real-time video edge detection system designed in this thesis includes OV5640 image acquisition module and image data cache module,edge detection module and High Definition Multimedia Interface(HDMI)display module.The edge detection effect of the improved Sobel algorithm is evaluated by edge retention index(ERI),signal-to-noise ratio(SNR),single edge responsiveness and edge continuity.The results show that the edge preservation index of the improved Sobel algorithm with better anti-noise performance is close to 1.The single edge response index and edge continuity index are increased by 37% and 43% on average respectively.The improved Sobel algorithm improves the accuracy of edge detection effectively.The test using hardware platform shows that the FPGA edge detection system based on the improved Sobel algorithm can process 60 frames of real-time video images in 1 second quickly so that the system can be applied widely.