Study On Edge Detection Algorithm Of Computer Vision

As the development of visual inspection technology, it has been used more and more widely. Visual inspection is used more and more in industry, agriculture, defense, medical, transportation, sports, entertainment and so on. At the same time, people's demands of inspection accuracy are higher and higher. While in the visual inspection system, a core technology is the image edge extraction. The edge contains the important information of the image, and is also one of the basic features of the image. The effects of edge extraction will directly affect the segmentation and recognition performance of the image, and eventually affect inspection accuracy. So edge extraction is a very important part of the machine vision and image processing. As the initial stage of the visual system, edge inspection is usually considered a non-well-conditioned problem. Images actually processed are generally mixed with noise, the edges and noise are high-frequency signal, therefore as suppress noise will affect the edge position. Therefore, noise suppression and precise location of the edge is difficult to be satisfied at the same time. The key of the image edge extraction is to solve the conflict between noise suppression and retention of the edge of the original image.This article is based on analysis and research on some of the existing edge inspection techniques, aiming at deficiencies and defects of them, and puts forward an improved algorithm in order to improve the effect of the edge extraction.In this paper, firstly ,some of the classical edge inspection operators are introduced, then the advantages and disadvantages of these operators in the theory and methods and their application situations are analyzed. All kinds of algorithms are achieved by programming and are in contrast with simulation experiments.Secondly, Canny criteria and Canny algorithm are mainly studied and discussed. Canny algorithm reduces the impact of noise by designing optimal filters and convolution with images, and then calculates the gradient magnitude and direction. After Non-maxima suppression, double-threshold method is used for processing and edges are connected in the end.Moreover, the defects and deficiencies of traditional Canny algorithm are analyzed, on the basis of this an improved algorithm is put forward. This algorithm is improved by aiming at the defects in steps of the traditional Canny operator. The improved algorithm uses mixed non-linear filtering instead of the traditional linear Gaussian filtering, preserves edge information well while eliminate noise, and improves the calculation of the gradient amplitude and non-maxima suppression algorithm. And based on gradient magnitude histogram, the high and low threshold of algorithm is automatically identified.The basic theory and implementation of improved Canny algorithm is introduced in detail. Programming and Simulation is by matlab7.0. By being analyzed and compared with simulation results of traditional Canny operator, it's shown that the improved algorithm has better performance of noise suppression and edge detail information extraction. And also is of the high degree of automation and a more effective edge inspection method.